xref: /linux/net/mptcp/protocol.h (revision 1e15510b71c99c6e49134d756df91069f7d18141)
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 
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 	struct_group(status,
153 		u16 suboptions;
154 		u16 use_map:1,
155 		    dsn64:1,
156 		    data_fin:1,
157 		    use_ack:1,
158 		    ack64:1,
159 		    mpc_map:1,
160 		    reset_reason:4,
161 		    reset_transient:1,
162 		    echo:1,
163 		    backup:1,
164 		    deny_join_id0:1,
165 		    __unused:2;
166 	);
167 	u8	join_id;
168 	u32	token;
169 	u32	nonce;
170 	u64	thmac;
171 	u8	hmac[MPTCPOPT_HMAC_LEN];
172 	struct mptcp_addr_info addr;
173 	struct mptcp_rm_list rm_list;
174 	u64	ahmac;
175 	u64	fail_seq;
176 };
177 
mptcp_option(u8 subopt,u8 len,u8 nib,u8 field)178 static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field)
179 {
180 	return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) |
181 		     ((nib & 0xF) << 8) | field);
182 }
183 
184 enum mptcp_pm_status {
185 	MPTCP_PM_ADD_ADDR_RECEIVED,
186 	MPTCP_PM_ADD_ADDR_SEND_ACK,
187 	MPTCP_PM_RM_ADDR_RECEIVED,
188 	MPTCP_PM_ESTABLISHED,
189 	MPTCP_PM_SUBFLOW_ESTABLISHED,
190 	MPTCP_PM_ALREADY_ESTABLISHED,	/* persistent status, set after ESTABLISHED event */
191 	MPTCP_PM_MPC_ENDPOINT_ACCOUNTED /* persistent status, set after MPC local address is
192 					 * accounted int id_avail_bitmap
193 					 */
194 };
195 
196 enum mptcp_pm_type {
197 	MPTCP_PM_TYPE_KERNEL = 0,
198 	MPTCP_PM_TYPE_USERSPACE,
199 
200 	__MPTCP_PM_TYPE_NR,
201 	__MPTCP_PM_TYPE_MAX = __MPTCP_PM_TYPE_NR - 1,
202 };
203 
204 /* Status bits below MPTCP_PM_ALREADY_ESTABLISHED need pm worker actions */
205 #define MPTCP_PM_WORK_MASK ((1 << MPTCP_PM_ALREADY_ESTABLISHED) - 1)
206 
207 enum mptcp_addr_signal_status {
208 	MPTCP_ADD_ADDR_SIGNAL,
209 	MPTCP_ADD_ADDR_ECHO,
210 	MPTCP_RM_ADDR_SIGNAL,
211 };
212 
213 /* max value of mptcp_addr_info.id */
214 #define MPTCP_PM_MAX_ADDR_ID		U8_MAX
215 
216 struct mptcp_pm_data {
217 	struct mptcp_addr_info local;
218 	struct mptcp_addr_info remote;
219 	struct list_head anno_list;
220 	struct list_head userspace_pm_local_addr_list;
221 
222 	spinlock_t	lock;		/*protects the whole PM data */
223 
224 	u8		addr_signal;
225 	bool		server_side;
226 	bool		work_pending;
227 	bool		accept_addr;
228 	bool		accept_subflow;
229 	bool		remote_deny_join_id0;
230 	u8		add_addr_signaled;
231 	u8		add_addr_accepted;
232 	u8		local_addr_used;
233 	u8		pm_type;
234 	u8		subflows;
235 	u8		status;
236 	DECLARE_BITMAP(id_avail_bitmap, MPTCP_PM_MAX_ADDR_ID + 1);
237 	struct mptcp_rm_list rm_list_tx;
238 	struct mptcp_rm_list rm_list_rx;
239 };
240 
241 struct mptcp_pm_local {
242 	struct mptcp_addr_info	addr;
243 	u8			flags;
244 	int			ifindex;
245 };
246 
247 struct mptcp_pm_addr_entry {
248 	struct list_head	list;
249 	struct mptcp_addr_info	addr;
250 	u8			flags;
251 	int			ifindex;
252 	struct socket		*lsk;
253 };
254 
255 struct mptcp_data_frag {
256 	struct list_head list;
257 	u64 data_seq;
258 	u16 data_len;
259 	u16 offset;
260 	u16 overhead;
261 	u16 already_sent;
262 	struct page *page;
263 };
264 
265 /* MPTCP connection sock */
266 struct mptcp_sock {
267 	/* inet_connection_sock must be the first member */
268 	struct inet_connection_sock sk;
269 	u64		local_key;		/* protected by the first subflow socket lock
270 						 * lockless access read
271 						 */
272 	u64		remote_key;		/* same as above */
273 	u64		write_seq;
274 	u64		bytes_sent;
275 	u64		snd_nxt;
276 	u64		bytes_received;
277 	u64		ack_seq;
278 	atomic64_t	rcv_wnd_sent;
279 	u64		rcv_data_fin_seq;
280 	u64		bytes_retrans;
281 	u64		bytes_consumed;
282 	int		rmem_fwd_alloc;
283 	int		snd_burst;
284 	int		old_wspace;
285 	u64		recovery_snd_nxt;	/* in recovery mode accept up to this seq;
286 						 * recovery related fields are under data_lock
287 						 * protection
288 						 */
289 	u64		bytes_acked;
290 	u64		snd_una;
291 	u64		wnd_end;
292 	u32		last_data_sent;
293 	u32		last_data_recv;
294 	u32		last_ack_recv;
295 	unsigned long	timer_ival;
296 	u32		token;
297 	int		rmem_released;
298 	unsigned long	flags;
299 	unsigned long	cb_flags;
300 	bool		recovery;		/* closing subflow write queue reinjected */
301 	bool		can_ack;
302 	bool		fully_established;
303 	bool		rcv_data_fin;
304 	bool		snd_data_fin_enable;
305 	bool		rcv_fastclose;
306 	bool		use_64bit_ack; /* Set when we received a 64-bit DSN */
307 	bool		csum_enabled;
308 	bool		allow_infinite_fallback;
309 	u8		pending_state; /* A subflow asked to set this sk_state,
310 					* protected by the msk data lock
311 					*/
312 	u8		mpc_endpoint_id;
313 	u8		recvmsg_inq:1,
314 			cork:1,
315 			nodelay:1,
316 			fastopening:1,
317 			in_accept_queue:1,
318 			free_first:1,
319 			rcvspace_init:1;
320 	u32		notsent_lowat;
321 	int		keepalive_cnt;
322 	int		keepalive_idle;
323 	int		keepalive_intvl;
324 	struct work_struct work;
325 	struct sk_buff  *ooo_last_skb;
326 	struct rb_root  out_of_order_queue;
327 	struct sk_buff_head receive_queue;
328 	struct list_head conn_list;
329 	struct list_head rtx_queue;
330 	struct mptcp_data_frag *first_pending;
331 	struct list_head join_list;
332 	struct sock	*first; /* The mptcp ops can safely dereference, using suitable
333 				 * ONCE annotation, the subflow outside the socket
334 				 * lock as such sock is freed after close().
335 				 */
336 	struct mptcp_pm_data	pm;
337 	struct mptcp_sched_ops	*sched;
338 	struct {
339 		u32	space;	/* bytes copied in last measurement window */
340 		u32	copied; /* bytes copied in this measurement window */
341 		u64	time;	/* start time of measurement window */
342 		u64	rtt_us; /* last maximum rtt of subflows */
343 	} rcvq_space;
344 	u8		scaling_ratio;
345 
346 	u32		subflow_id;
347 	u32		setsockopt_seq;
348 	char		ca_name[TCP_CA_NAME_MAX];
349 };
350 
351 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock)
352 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock)
353 
354 #define mptcp_for_each_subflow(__msk, __subflow)			\
355 	list_for_each_entry(__subflow, &((__msk)->conn_list), node)
356 #define mptcp_for_each_subflow_safe(__msk, __subflow, __tmp)			\
357 	list_for_each_entry_safe(__subflow, __tmp, &((__msk)->conn_list), node)
358 
359 extern struct genl_family mptcp_genl_family;
360 
msk_owned_by_me(const struct mptcp_sock * msk)361 static inline void msk_owned_by_me(const struct mptcp_sock *msk)
362 {
363 	sock_owned_by_me((const struct sock *)msk);
364 }
365 
366 #ifdef CONFIG_DEBUG_NET
367 /* MPTCP-specific: we might (indirectly) call this helper with the wrong sk */
368 #undef tcp_sk
369 #define tcp_sk(ptr) ({								\
370 	typeof(ptr) _ptr = (ptr);						\
371 	WARN_ON(_ptr->sk_protocol != IPPROTO_TCP);				\
372 	container_of_const(_ptr, struct tcp_sock, inet_conn.icsk_inet.sk);	\
373 })
374 #define mptcp_sk(ptr) ({						\
375 	typeof(ptr) _ptr = (ptr);					\
376 	WARN_ON(_ptr->sk_protocol != IPPROTO_MPTCP);			\
377 	container_of_const(_ptr, struct mptcp_sock, sk.icsk_inet.sk);	\
378 })
379 
380 #else /* !CONFIG_DEBUG_NET */
381 #define mptcp_sk(ptr) container_of_const(ptr, struct mptcp_sock, sk.icsk_inet.sk)
382 #endif
383 
384 /* the msk socket don't use the backlog, also account for the bulk
385  * free memory
386  */
__mptcp_rmem(const struct sock * sk)387 static inline int __mptcp_rmem(const struct sock *sk)
388 {
389 	return atomic_read(&sk->sk_rmem_alloc) - READ_ONCE(mptcp_sk(sk)->rmem_released);
390 }
391 
mptcp_win_from_space(const struct sock * sk,int space)392 static inline int mptcp_win_from_space(const struct sock *sk, int space)
393 {
394 	return __tcp_win_from_space(mptcp_sk(sk)->scaling_ratio, space);
395 }
396 
mptcp_space_from_win(const struct sock * sk,int win)397 static inline int mptcp_space_from_win(const struct sock *sk, int win)
398 {
399 	return __tcp_space_from_win(mptcp_sk(sk)->scaling_ratio, win);
400 }
401 
__mptcp_space(const struct sock * sk)402 static inline int __mptcp_space(const struct sock *sk)
403 {
404 	return mptcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - __mptcp_rmem(sk));
405 }
406 
mptcp_send_head(const struct sock * sk)407 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk)
408 {
409 	const struct mptcp_sock *msk = mptcp_sk(sk);
410 
411 	return READ_ONCE(msk->first_pending);
412 }
413 
mptcp_send_next(struct sock * sk)414 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk)
415 {
416 	struct mptcp_sock *msk = mptcp_sk(sk);
417 	struct mptcp_data_frag *cur;
418 
419 	cur = msk->first_pending;
420 	return list_is_last(&cur->list, &msk->rtx_queue) ? NULL :
421 						     list_next_entry(cur, list);
422 }
423 
mptcp_pending_tail(const struct sock * sk)424 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk)
425 {
426 	const struct mptcp_sock *msk = mptcp_sk(sk);
427 
428 	if (!msk->first_pending)
429 		return NULL;
430 
431 	if (WARN_ON_ONCE(list_empty(&msk->rtx_queue)))
432 		return NULL;
433 
434 	return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list);
435 }
436 
mptcp_rtx_head(struct sock * sk)437 static inline struct mptcp_data_frag *mptcp_rtx_head(struct sock *sk)
438 {
439 	struct mptcp_sock *msk = mptcp_sk(sk);
440 
441 	if (msk->snd_una == msk->snd_nxt)
442 		return NULL;
443 
444 	return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list);
445 }
446 
447 struct csum_pseudo_header {
448 	__be64 data_seq;
449 	__be32 subflow_seq;
450 	__be16 data_len;
451 	__sum16 csum;
452 };
453 
454 struct mptcp_subflow_request_sock {
455 	struct	tcp_request_sock sk;
456 	u16	mp_capable : 1,
457 		mp_join : 1,
458 		backup : 1,
459 		request_bkup : 1,
460 		csum_reqd : 1,
461 		allow_join_id0 : 1;
462 	u8	local_id;
463 	u8	remote_id;
464 	u64	local_key;
465 	u64	idsn;
466 	u32	token;
467 	u32	ssn_offset;
468 	u64	thmac;
469 	u32	local_nonce;
470 	u32	remote_nonce;
471 	struct mptcp_sock	*msk;
472 	struct hlist_nulls_node token_node;
473 };
474 
475 static inline struct mptcp_subflow_request_sock *
mptcp_subflow_rsk(const struct request_sock * rsk)476 mptcp_subflow_rsk(const struct request_sock *rsk)
477 {
478 	return (struct mptcp_subflow_request_sock *)rsk;
479 }
480 
481 struct mptcp_delegated_action {
482 	struct napi_struct napi;
483 	struct list_head head;
484 };
485 
486 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions);
487 
488 #define MPTCP_DELEGATE_SCHEDULED	0
489 #define MPTCP_DELEGATE_SEND		1
490 #define MPTCP_DELEGATE_ACK		2
491 #define MPTCP_DELEGATE_SNDBUF		3
492 
493 #define MPTCP_DELEGATE_ACTIONS_MASK	(~BIT(MPTCP_DELEGATE_SCHEDULED))
494 /* MPTCP subflow context */
495 struct mptcp_subflow_context {
496 	struct	list_head node;/* conn_list of subflows */
497 
498 	struct_group(reset,
499 
500 	unsigned long avg_pacing_rate; /* protected by msk socket lock */
501 	u64	local_key;
502 	u64	remote_key;
503 	u64	idsn;
504 	u64	map_seq;
505 	u32	snd_isn;
506 	u32	token;
507 	u32	rel_write_seq;
508 	u32	map_subflow_seq;
509 	u32	ssn_offset;
510 	u32	map_data_len;
511 	__wsum	map_data_csum;
512 	u32	map_csum_len;
513 	u32	request_mptcp : 1,  /* send MP_CAPABLE */
514 		request_join : 1,   /* send MP_JOIN */
515 		request_bkup : 1,
516 		mp_capable : 1,	    /* remote is MPTCP capable */
517 		mp_join : 1,	    /* remote is JOINing */
518 		pm_notified : 1,    /* PM hook called for established status */
519 		conn_finished : 1,
520 		map_valid : 1,
521 		map_csum_reqd : 1,
522 		map_data_fin : 1,
523 		mpc_map : 1,
524 		backup : 1,
525 		send_mp_prio : 1,
526 		send_mp_fail : 1,
527 		send_fastclose : 1,
528 		send_infinite_map : 1,
529 		remote_key_valid : 1,        /* received the peer key from */
530 		disposable : 1,	    /* ctx can be free at ulp release time */
531 		stale : 1,	    /* unable to snd/rcv data, do not use for xmit */
532 		valid_csum_seen : 1,        /* at least one csum validated */
533 		is_mptfo : 1,	    /* subflow is doing TFO */
534 		close_event_done : 1,       /* has done the post-closed part */
535 		mpc_drop : 1,	    /* the MPC option has been dropped in a rtx */
536 		__unused : 9;
537 	bool	data_avail;
538 	bool	scheduled;
539 	bool	pm_listener;	    /* a listener managed by the kernel PM? */
540 	bool	fully_established;  /* path validated */
541 	u32	remote_nonce;
542 	u64	thmac;
543 	u32	local_nonce;
544 	u32	remote_token;
545 	union {
546 		u8	hmac[MPTCPOPT_HMAC_LEN]; /* MPJ subflow only */
547 		u64	iasn;	    /* initial ack sequence number, MPC subflows only */
548 	};
549 	s16	local_id;	    /* if negative not initialized yet */
550 	u8	remote_id;
551 	u8	reset_seen:1;
552 	u8	reset_transient:1;
553 	u8	reset_reason:4;
554 	u8	stale_count;
555 
556 	u32	subflow_id;
557 
558 	long	delegated_status;
559 	unsigned long	fail_tout;
560 
561 	);
562 
563 	struct	list_head delegated_node;   /* link into delegated_action, protected by local BH */
564 
565 	u32	setsockopt_seq;
566 	u32	stale_rcv_tstamp;
567 	int     cached_sndbuf;	    /* sndbuf size when last synced with the msk sndbuf,
568 				     * protected by the msk socket lock
569 				     */
570 
571 	struct	sock *tcp_sock;	    /* tcp sk backpointer */
572 	struct	sock *conn;	    /* parent mptcp_sock */
573 	const	struct inet_connection_sock_af_ops *icsk_af_ops;
574 	void	(*tcp_state_change)(struct sock *sk);
575 	void	(*tcp_error_report)(struct sock *sk);
576 
577 	struct	rcu_head rcu;
578 };
579 
580 static inline struct mptcp_subflow_context *
mptcp_subflow_ctx(const struct sock * sk)581 mptcp_subflow_ctx(const struct sock *sk)
582 {
583 	const struct inet_connection_sock *icsk = inet_csk(sk);
584 
585 	/* Use RCU on icsk_ulp_data only for sock diag code */
586 	return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data;
587 }
588 
589 static inline struct sock *
mptcp_subflow_tcp_sock(const struct mptcp_subflow_context * subflow)590 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow)
591 {
592 	return subflow->tcp_sock;
593 }
594 
595 static inline void
mptcp_subflow_ctx_reset(struct mptcp_subflow_context * subflow)596 mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow)
597 {
598 	memset(&subflow->reset, 0, sizeof(subflow->reset));
599 	subflow->request_mptcp = 1;
600 	WRITE_ONCE(subflow->local_id, -1);
601 }
602 
603 /* Convert reset reasons in MPTCP to enum sk_rst_reason type */
604 static inline enum sk_rst_reason
sk_rst_convert_mptcp_reason(u32 reason)605 sk_rst_convert_mptcp_reason(u32 reason)
606 {
607 	switch (reason) {
608 	case MPTCP_RST_EUNSPEC:
609 		return SK_RST_REASON_MPTCP_RST_EUNSPEC;
610 	case MPTCP_RST_EMPTCP:
611 		return SK_RST_REASON_MPTCP_RST_EMPTCP;
612 	case MPTCP_RST_ERESOURCE:
613 		return SK_RST_REASON_MPTCP_RST_ERESOURCE;
614 	case MPTCP_RST_EPROHIBIT:
615 		return SK_RST_REASON_MPTCP_RST_EPROHIBIT;
616 	case MPTCP_RST_EWQ2BIG:
617 		return SK_RST_REASON_MPTCP_RST_EWQ2BIG;
618 	case MPTCP_RST_EBADPERF:
619 		return SK_RST_REASON_MPTCP_RST_EBADPERF;
620 	case MPTCP_RST_EMIDDLEBOX:
621 		return SK_RST_REASON_MPTCP_RST_EMIDDLEBOX;
622 	default:
623 		/* It should not happen, or else errors may occur
624 		 * in MPTCP layer
625 		 */
626 		return SK_RST_REASON_ERROR;
627 	}
628 }
629 
630 static inline void
mptcp_send_active_reset_reason(struct sock * sk)631 mptcp_send_active_reset_reason(struct sock *sk)
632 {
633 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
634 	enum sk_rst_reason reason;
635 
636 	reason = sk_rst_convert_mptcp_reason(subflow->reset_reason);
637 	tcp_send_active_reset(sk, GFP_ATOMIC, reason);
638 }
639 
640 static inline u64
mptcp_subflow_get_map_offset(const struct mptcp_subflow_context * subflow)641 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow)
642 {
643 	return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq -
644 		      subflow->ssn_offset -
645 		      subflow->map_subflow_seq;
646 }
647 
648 static inline u64
mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context * subflow)649 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow)
650 {
651 	return subflow->map_seq + mptcp_subflow_get_map_offset(subflow);
652 }
653 
654 void mptcp_subflow_process_delegated(struct sock *ssk, long actions);
655 
mptcp_subflow_delegate(struct mptcp_subflow_context * subflow,int action)656 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action)
657 {
658 	long old, set_bits = BIT(MPTCP_DELEGATE_SCHEDULED) | BIT(action);
659 	struct mptcp_delegated_action *delegated;
660 	bool schedule;
661 
662 	/* the caller held the subflow bh socket lock */
663 	lockdep_assert_in_softirq();
664 
665 	/* The implied barrier pairs with tcp_release_cb_override()
666 	 * mptcp_napi_poll(), and ensures the below list check sees list
667 	 * updates done prior to delegated status bits changes
668 	 */
669 	old = set_mask_bits(&subflow->delegated_status, 0, set_bits);
670 	if (!(old & BIT(MPTCP_DELEGATE_SCHEDULED))) {
671 		if (WARN_ON_ONCE(!list_empty(&subflow->delegated_node)))
672 			return;
673 
674 		delegated = this_cpu_ptr(&mptcp_delegated_actions);
675 		schedule = list_empty(&delegated->head);
676 		list_add_tail(&subflow->delegated_node, &delegated->head);
677 		sock_hold(mptcp_subflow_tcp_sock(subflow));
678 		if (schedule)
679 			napi_schedule(&delegated->napi);
680 	}
681 }
682 
683 static inline struct mptcp_subflow_context *
mptcp_subflow_delegated_next(struct mptcp_delegated_action * delegated)684 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated)
685 {
686 	struct mptcp_subflow_context *ret;
687 
688 	if (list_empty(&delegated->head))
689 		return NULL;
690 
691 	ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node);
692 	list_del_init(&ret->delegated_node);
693 	return ret;
694 }
695 
696 int mptcp_is_enabled(const struct net *net);
697 unsigned int mptcp_get_add_addr_timeout(const struct net *net);
698 int mptcp_is_checksum_enabled(const struct net *net);
699 int mptcp_allow_join_id0(const struct net *net);
700 unsigned int mptcp_stale_loss_cnt(const struct net *net);
701 unsigned int mptcp_close_timeout(const struct sock *sk);
702 int mptcp_get_pm_type(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_owner_r(struct sk_buff *skb, struct sock *sk);
730 void mptcp_set_state(struct sock *sk, int state);
731 
732 bool mptcp_addresses_equal(const struct mptcp_addr_info *a,
733 			   const struct mptcp_addr_info *b, bool use_port);
734 void mptcp_local_address(const struct sock_common *skc, struct mptcp_addr_info *addr);
735 
736 /* called with sk socket lock held */
737 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_pm_local *local,
738 			    const struct mptcp_addr_info *remote);
739 int mptcp_subflow_create_socket(struct sock *sk, unsigned short family,
740 				struct socket **new_sock);
741 void mptcp_info2sockaddr(const struct mptcp_addr_info *info,
742 			 struct sockaddr_storage *addr,
743 			 unsigned short family);
744 struct mptcp_sched_ops *mptcp_sched_find(const char *name);
745 int mptcp_register_scheduler(struct mptcp_sched_ops *sched);
746 void mptcp_unregister_scheduler(struct mptcp_sched_ops *sched);
747 void mptcp_sched_init(void);
748 int mptcp_init_sched(struct mptcp_sock *msk,
749 		     struct mptcp_sched_ops *sched);
750 void mptcp_release_sched(struct mptcp_sock *msk);
751 void mptcp_subflow_set_scheduled(struct mptcp_subflow_context *subflow,
752 				 bool scheduled);
753 struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk);
754 struct sock *mptcp_subflow_get_retrans(struct mptcp_sock *msk);
755 int mptcp_sched_get_send(struct mptcp_sock *msk);
756 int mptcp_sched_get_retrans(struct mptcp_sock *msk);
757 
mptcp_data_avail(const struct mptcp_sock * msk)758 static inline u64 mptcp_data_avail(const struct mptcp_sock *msk)
759 {
760 	return READ_ONCE(msk->bytes_received) - READ_ONCE(msk->bytes_consumed);
761 }
762 
mptcp_epollin_ready(const struct sock * sk)763 static inline bool mptcp_epollin_ready(const struct sock *sk)
764 {
765 	u64 data_avail = mptcp_data_avail(mptcp_sk(sk));
766 
767 	if (!data_avail)
768 		return false;
769 
770 	/* mptcp doesn't have to deal with small skbs in the receive queue,
771 	 * as it can always coalesce them
772 	 */
773 	return (data_avail >= sk->sk_rcvlowat) ||
774 	       (mem_cgroup_sockets_enabled && sk->sk_memcg &&
775 		mem_cgroup_under_socket_pressure(sk->sk_memcg)) ||
776 	       READ_ONCE(tcp_memory_pressure);
777 }
778 
779 int mptcp_set_rcvlowat(struct sock *sk, int val);
780 
__tcp_can_send(const struct sock * ssk)781 static inline bool __tcp_can_send(const struct sock *ssk)
782 {
783 	/* only send if our side has not closed yet */
784 	return ((1 << inet_sk_state_load(ssk)) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT));
785 }
786 
__mptcp_subflow_active(struct mptcp_subflow_context * subflow)787 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow)
788 {
789 	/* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */
790 	if (subflow->request_join && !READ_ONCE(subflow->fully_established))
791 		return false;
792 
793 	return __tcp_can_send(mptcp_subflow_tcp_sock(subflow));
794 }
795 
796 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow);
797 
798 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow);
799 
800 void mptcp_subflow_drop_ctx(struct sock *ssk);
801 
mptcp_subflow_tcp_fallback(struct sock * sk,struct mptcp_subflow_context * ctx)802 static inline void mptcp_subflow_tcp_fallback(struct sock *sk,
803 					      struct mptcp_subflow_context *ctx)
804 {
805 	sk->sk_data_ready = sock_def_readable;
806 	sk->sk_state_change = ctx->tcp_state_change;
807 	sk->sk_write_space = sk_stream_write_space;
808 	sk->sk_error_report = ctx->tcp_error_report;
809 
810 	inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops;
811 }
812 
813 void __init mptcp_proto_init(void);
814 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
815 int __init mptcp_proto_v6_init(void);
816 #endif
817 
818 struct sock *mptcp_sk_clone_init(const struct sock *sk,
819 				 const struct mptcp_options_received *mp_opt,
820 				 struct sock *ssk,
821 				 struct request_sock *req);
822 void mptcp_get_options(const struct sk_buff *skb,
823 		       struct mptcp_options_received *mp_opt);
824 
825 void mptcp_finish_connect(struct sock *sk);
826 void __mptcp_sync_state(struct sock *sk, int state);
827 void mptcp_reset_tout_timer(struct mptcp_sock *msk, unsigned long fail_tout);
828 
mptcp_stop_tout_timer(struct sock * sk)829 static inline void mptcp_stop_tout_timer(struct sock *sk)
830 {
831 	if (!inet_csk(sk)->icsk_mtup.probe_timestamp)
832 		return;
833 
834 	sk_stop_timer(sk, &sk->sk_timer);
835 	inet_csk(sk)->icsk_mtup.probe_timestamp = 0;
836 }
837 
mptcp_set_close_tout(struct sock * sk,unsigned long tout)838 static inline void mptcp_set_close_tout(struct sock *sk, unsigned long tout)
839 {
840 	/* avoid 0 timestamp, as that means no close timeout */
841 	inet_csk(sk)->icsk_mtup.probe_timestamp = tout ? : 1;
842 }
843 
mptcp_start_tout_timer(struct sock * sk)844 static inline void mptcp_start_tout_timer(struct sock *sk)
845 {
846 	mptcp_set_close_tout(sk, tcp_jiffies32);
847 	mptcp_reset_tout_timer(mptcp_sk(sk), 0);
848 }
849 
mptcp_is_fully_established(struct sock * sk)850 static inline bool mptcp_is_fully_established(struct sock *sk)
851 {
852 	return inet_sk_state_load(sk) == TCP_ESTABLISHED &&
853 	       READ_ONCE(mptcp_sk(sk)->fully_established);
854 }
855 
856 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk);
857 void mptcp_data_ready(struct sock *sk, struct sock *ssk);
858 bool mptcp_finish_join(struct sock *sk);
859 bool mptcp_schedule_work(struct sock *sk);
860 int mptcp_setsockopt(struct sock *sk, int level, int optname,
861 		     sockptr_t optval, unsigned int optlen);
862 int mptcp_getsockopt(struct sock *sk, int level, int optname,
863 		     char __user *optval, int __user *option);
864 
865 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq);
mptcp_expand_seq(u64 old_seq,u64 cur_seq,bool use_64bit)866 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit)
867 {
868 	if (use_64bit)
869 		return cur_seq;
870 
871 	return __mptcp_expand_seq(old_seq, cur_seq);
872 }
873 void __mptcp_check_push(struct sock *sk, struct sock *ssk);
874 void __mptcp_data_acked(struct sock *sk);
875 void __mptcp_error_report(struct sock *sk);
876 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)877 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk)
878 {
879 	return READ_ONCE(msk->snd_data_fin_enable) &&
880 	       READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt);
881 }
882 
mptcp_notsent_lowat(const struct sock * sk)883 static inline u32 mptcp_notsent_lowat(const struct sock *sk)
884 {
885 	struct net *net = sock_net(sk);
886 	u32 val;
887 
888 	val = READ_ONCE(mptcp_sk(sk)->notsent_lowat);
889 	return val ?: READ_ONCE(net->ipv4.sysctl_tcp_notsent_lowat);
890 }
891 
mptcp_stream_memory_free(const struct sock * sk,int wake)892 static inline bool mptcp_stream_memory_free(const struct sock *sk, int wake)
893 {
894 	const struct mptcp_sock *msk = mptcp_sk(sk);
895 	u32 notsent_bytes;
896 
897 	notsent_bytes = READ_ONCE(msk->write_seq) - READ_ONCE(msk->snd_nxt);
898 	return (notsent_bytes << wake) < mptcp_notsent_lowat(sk);
899 }
900 
__mptcp_stream_is_writeable(const struct sock * sk,int wake)901 static inline bool __mptcp_stream_is_writeable(const struct sock *sk, int wake)
902 {
903 	return mptcp_stream_memory_free(sk, wake) &&
904 	       __sk_stream_is_writeable(sk, wake);
905 }
906 
mptcp_write_space(struct sock * sk)907 static inline void mptcp_write_space(struct sock *sk)
908 {
909 	/* pairs with memory barrier in mptcp_poll */
910 	smp_mb();
911 	if (mptcp_stream_memory_free(sk, 1))
912 		sk_stream_write_space(sk);
913 }
914 
__mptcp_sync_sndbuf(struct sock * sk)915 static inline void __mptcp_sync_sndbuf(struct sock *sk)
916 {
917 	struct mptcp_subflow_context *subflow;
918 	int ssk_sndbuf, new_sndbuf;
919 
920 	if (sk->sk_userlocks & SOCK_SNDBUF_LOCK)
921 		return;
922 
923 	new_sndbuf = READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_wmem[0]);
924 	mptcp_for_each_subflow(mptcp_sk(sk), subflow) {
925 		ssk_sndbuf =  READ_ONCE(mptcp_subflow_tcp_sock(subflow)->sk_sndbuf);
926 
927 		subflow->cached_sndbuf = ssk_sndbuf;
928 		new_sndbuf += ssk_sndbuf;
929 	}
930 
931 	/* the msk max wmem limit is <nr_subflows> * tcp wmem[2] */
932 	WRITE_ONCE(sk->sk_sndbuf, new_sndbuf);
933 	mptcp_write_space(sk);
934 }
935 
936 /* The called held both the msk socket and the subflow socket locks,
937  * possibly under BH
938  */
__mptcp_propagate_sndbuf(struct sock * sk,struct sock * ssk)939 static inline void __mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
940 {
941 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
942 
943 	if (READ_ONCE(ssk->sk_sndbuf) != subflow->cached_sndbuf)
944 		__mptcp_sync_sndbuf(sk);
945 }
946 
947 /* the caller held only the subflow socket lock, either in process or
948  * BH context. Additionally this can be called under the msk data lock,
949  * so we can't acquire such lock here: let the delegate action acquires
950  * the needed locks in suitable order.
951  */
mptcp_propagate_sndbuf(struct sock * sk,struct sock * ssk)952 static inline void mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
953 {
954 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
955 
956 	if (likely(READ_ONCE(ssk->sk_sndbuf) == subflow->cached_sndbuf))
957 		return;
958 
959 	local_bh_disable();
960 	mptcp_subflow_delegate(subflow, MPTCP_DELEGATE_SNDBUF);
961 	local_bh_enable();
962 }
963 
964 void mptcp_destroy_common(struct mptcp_sock *msk, unsigned int flags);
965 
966 #define MPTCP_TOKEN_MAX_RETRIES	4
967 
968 void __init mptcp_token_init(void);
mptcp_token_init_request(struct request_sock * req)969 static inline void mptcp_token_init_request(struct request_sock *req)
970 {
971 	mptcp_subflow_rsk(req)->token_node.pprev = NULL;
972 }
973 
974 int mptcp_token_new_request(struct request_sock *req);
975 void mptcp_token_destroy_request(struct request_sock *req);
976 int mptcp_token_new_connect(struct sock *ssk);
977 void mptcp_token_accept(struct mptcp_subflow_request_sock *r,
978 			struct mptcp_sock *msk);
979 bool mptcp_token_exists(u32 token);
980 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token);
981 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot,
982 					 long *s_num);
983 void mptcp_token_destroy(struct mptcp_sock *msk);
984 
985 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn);
986 
987 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac);
988 __sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum);
989 
990 void __init mptcp_pm_init(void);
991 void mptcp_pm_data_init(struct mptcp_sock *msk);
992 void mptcp_pm_data_reset(struct mptcp_sock *msk);
993 int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info,
994 			struct mptcp_addr_info *addr);
995 int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info,
996 			 bool require_family,
997 			 struct mptcp_pm_addr_entry *entry);
998 bool mptcp_pm_addr_families_match(const struct sock *sk,
999 				  const struct mptcp_addr_info *loc,
1000 				  const struct mptcp_addr_info *rem);
1001 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
1002 void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
1003 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side);
1004 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk);
1005 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk);
1006 void mptcp_pm_connection_closed(struct mptcp_sock *msk);
1007 void mptcp_pm_subflow_established(struct mptcp_sock *msk);
1008 bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk);
1009 void mptcp_pm_subflow_check_next(struct mptcp_sock *msk,
1010 				 const struct mptcp_subflow_context *subflow);
1011 void mptcp_pm_add_addr_received(const struct sock *ssk,
1012 				const struct mptcp_addr_info *addr);
1013 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk,
1014 			      const struct mptcp_addr_info *addr);
1015 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk);
1016 bool mptcp_pm_nl_is_init_remote_addr(struct mptcp_sock *msk,
1017 				     const struct mptcp_addr_info *remote);
1018 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk);
1019 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk,
1020 			       const struct mptcp_rm_list *rm_list);
1021 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup);
1022 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq);
1023 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk,
1024 				 struct mptcp_addr_info *addr,
1025 				 struct mptcp_addr_info *rem,
1026 				 u8 bkup);
1027 bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk,
1028 			      const struct mptcp_addr_info *addr);
1029 void mptcp_pm_free_anno_list(struct mptcp_sock *msk);
1030 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk);
1031 struct mptcp_pm_add_entry *
1032 mptcp_pm_del_add_timer(struct mptcp_sock *msk,
1033 		       const struct mptcp_addr_info *addr, bool check_id);
1034 struct mptcp_pm_add_entry *
1035 mptcp_lookup_anno_list_by_saddr(const struct mptcp_sock *msk,
1036 				const struct mptcp_addr_info *addr);
1037 bool mptcp_lookup_subflow_by_saddr(const struct list_head *list,
1038 				   const struct mptcp_addr_info *saddr);
1039 bool mptcp_remove_anno_list_by_saddr(struct mptcp_sock *msk,
1040 				     const struct mptcp_addr_info *addr);
1041 int mptcp_pm_set_flags(struct sk_buff *skb, struct genl_info *info);
1042 int mptcp_pm_nl_set_flags(struct sk_buff *skb, struct genl_info *info);
1043 int mptcp_userspace_pm_set_flags(struct sk_buff *skb, struct genl_info *info);
1044 int mptcp_pm_announce_addr(struct mptcp_sock *msk,
1045 			   const struct mptcp_addr_info *addr,
1046 			   bool echo);
1047 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
1048 void mptcp_pm_remove_addr_entry(struct mptcp_sock *msk,
1049 				struct mptcp_pm_addr_entry *entry);
1050 
1051 void mptcp_free_local_addr_list(struct mptcp_sock *msk);
1052 
1053 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk,
1054 		 const struct sock *ssk, gfp_t gfp);
1055 void mptcp_event_addr_announced(const struct sock *ssk, const struct mptcp_addr_info *info);
1056 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id);
1057 void mptcp_event_pm_listener(const struct sock *ssk,
1058 			     enum mptcp_event_type event);
1059 bool mptcp_userspace_pm_active(const struct mptcp_sock *msk);
1060 
1061 void __mptcp_fastopen_gen_msk_ackseq(struct mptcp_sock *msk, struct mptcp_subflow_context *subflow,
1062 				     const struct mptcp_options_received *mp_opt);
1063 void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subflow,
1064 					      struct request_sock *req);
1065 int mptcp_nl_fill_addr(struct sk_buff *skb,
1066 		       struct mptcp_pm_addr_entry *entry);
1067 
mptcp_pm_should_add_signal(struct mptcp_sock * msk)1068 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk)
1069 {
1070 	return READ_ONCE(msk->pm.addr_signal) &
1071 		(BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO));
1072 }
1073 
mptcp_pm_should_add_signal_addr(struct mptcp_sock * msk)1074 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk)
1075 {
1076 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL);
1077 }
1078 
mptcp_pm_should_add_signal_echo(struct mptcp_sock * msk)1079 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk)
1080 {
1081 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO);
1082 }
1083 
mptcp_pm_should_rm_signal(struct mptcp_sock * msk)1084 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk)
1085 {
1086 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL);
1087 }
1088 
mptcp_pm_is_userspace(const struct mptcp_sock * msk)1089 static inline bool mptcp_pm_is_userspace(const struct mptcp_sock *msk)
1090 {
1091 	return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_USERSPACE;
1092 }
1093 
mptcp_pm_is_kernel(const struct mptcp_sock * msk)1094 static inline bool mptcp_pm_is_kernel(const struct mptcp_sock *msk)
1095 {
1096 	return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_KERNEL;
1097 }
1098 
mptcp_add_addr_len(int family,bool echo,bool port)1099 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port)
1100 {
1101 	u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE;
1102 
1103 	if (family == AF_INET6)
1104 		len = TCPOLEN_MPTCP_ADD_ADDR6_BASE;
1105 	if (!echo)
1106 		len += MPTCPOPT_THMAC_LEN;
1107 	/* account for 2 trailing 'nop' options */
1108 	if (port)
1109 		len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN;
1110 
1111 	return len;
1112 }
1113 
mptcp_rm_addr_len(const struct mptcp_rm_list * rm_list)1114 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list)
1115 {
1116 	if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX)
1117 		return -EINVAL;
1118 
1119 	return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1;
1120 }
1121 
1122 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, const struct sk_buff *skb,
1123 			      unsigned int opt_size, unsigned int remaining,
1124 			      struct mptcp_addr_info *addr, bool *echo,
1125 			      bool *drop_other_suboptions);
1126 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining,
1127 			     struct mptcp_rm_list *rm_list);
1128 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
1129 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1130 int mptcp_userspace_pm_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1131 bool mptcp_pm_is_backup(struct mptcp_sock *msk, struct sock_common *skc);
1132 bool mptcp_pm_nl_is_backup(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1133 bool mptcp_userspace_pm_is_backup(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1134 int mptcp_pm_dump_addr(struct sk_buff *msg, struct netlink_callback *cb);
1135 int mptcp_pm_nl_dump_addr(struct sk_buff *msg,
1136 			  struct netlink_callback *cb);
1137 int mptcp_userspace_pm_dump_addr(struct sk_buff *msg,
1138 				 struct netlink_callback *cb);
1139 int mptcp_pm_get_addr(struct sk_buff *skb, struct genl_info *info);
1140 int mptcp_pm_nl_get_addr(struct sk_buff *skb, struct genl_info *info);
1141 int mptcp_userspace_pm_get_addr(struct sk_buff *skb,
1142 				struct genl_info *info);
1143 
subflow_get_local_id(const struct mptcp_subflow_context * subflow)1144 static inline u8 subflow_get_local_id(const struct mptcp_subflow_context *subflow)
1145 {
1146 	int local_id = READ_ONCE(subflow->local_id);
1147 
1148 	if (local_id < 0)
1149 		return 0;
1150 	return local_id;
1151 }
1152 
1153 void __init mptcp_pm_nl_init(void);
1154 void mptcp_pm_nl_work(struct mptcp_sock *msk);
1155 unsigned int mptcp_pm_get_add_addr_signal_max(const struct mptcp_sock *msk);
1156 unsigned int mptcp_pm_get_add_addr_accept_max(const struct mptcp_sock *msk);
1157 unsigned int mptcp_pm_get_subflows_max(const struct mptcp_sock *msk);
1158 unsigned int mptcp_pm_get_local_addr_max(const struct mptcp_sock *msk);
1159 
1160 /* called under PM lock */
__mptcp_pm_close_subflow(struct mptcp_sock * msk)1161 static inline void __mptcp_pm_close_subflow(struct mptcp_sock *msk)
1162 {
1163 	if (--msk->pm.subflows < mptcp_pm_get_subflows_max(msk))
1164 		WRITE_ONCE(msk->pm.accept_subflow, true);
1165 }
1166 
mptcp_pm_close_subflow(struct mptcp_sock * msk)1167 static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk)
1168 {
1169 	spin_lock_bh(&msk->pm.lock);
1170 	__mptcp_pm_close_subflow(msk);
1171 	spin_unlock_bh(&msk->pm.lock);
1172 }
1173 
1174 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk);
1175 
mptcp_get_ext(const struct sk_buff * skb)1176 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb)
1177 {
1178 	return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP);
1179 }
1180 
1181 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops);
1182 
__mptcp_check_fallback(const struct mptcp_sock * msk)1183 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk)
1184 {
1185 	return test_bit(MPTCP_FALLBACK_DONE, &msk->flags);
1186 }
1187 
mptcp_check_fallback(const struct sock * sk)1188 static inline bool mptcp_check_fallback(const struct sock *sk)
1189 {
1190 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1191 	struct mptcp_sock *msk = mptcp_sk(subflow->conn);
1192 
1193 	return __mptcp_check_fallback(msk);
1194 }
1195 
__mptcp_do_fallback(struct mptcp_sock * msk)1196 static inline void __mptcp_do_fallback(struct mptcp_sock *msk)
1197 {
1198 	if (__mptcp_check_fallback(msk)) {
1199 		pr_debug("TCP fallback already done (msk=%p)\n", msk);
1200 		return;
1201 	}
1202 	if (WARN_ON_ONCE(!READ_ONCE(msk->allow_infinite_fallback)))
1203 		return;
1204 	set_bit(MPTCP_FALLBACK_DONE, &msk->flags);
1205 }
1206 
__mptcp_has_initial_subflow(const struct mptcp_sock * msk)1207 static inline bool __mptcp_has_initial_subflow(const struct mptcp_sock *msk)
1208 {
1209 	struct sock *ssk = READ_ONCE(msk->first);
1210 
1211 	return ssk && ((1 << inet_sk_state_load(ssk)) &
1212 		       (TCPF_ESTABLISHED | TCPF_SYN_SENT |
1213 			TCPF_SYN_RECV | TCPF_LISTEN));
1214 }
1215 
mptcp_do_fallback(struct sock * ssk)1216 static inline void mptcp_do_fallback(struct sock *ssk)
1217 {
1218 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
1219 	struct sock *sk = subflow->conn;
1220 	struct mptcp_sock *msk;
1221 
1222 	msk = mptcp_sk(sk);
1223 	__mptcp_do_fallback(msk);
1224 	if (READ_ONCE(msk->snd_data_fin_enable) && !(ssk->sk_shutdown & SEND_SHUTDOWN)) {
1225 		gfp_t saved_allocation = ssk->sk_allocation;
1226 
1227 		/* we are in a atomic (BH) scope, override ssk default for data
1228 		 * fin allocation
1229 		 */
1230 		ssk->sk_allocation = GFP_ATOMIC;
1231 		ssk->sk_shutdown |= SEND_SHUTDOWN;
1232 		tcp_shutdown(ssk, SEND_SHUTDOWN);
1233 		ssk->sk_allocation = saved_allocation;
1234 	}
1235 }
1236 
1237 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)\n", __func__, a)
1238 
mptcp_subflow_early_fallback(struct mptcp_sock * msk,struct mptcp_subflow_context * subflow)1239 static inline void mptcp_subflow_early_fallback(struct mptcp_sock *msk,
1240 						struct mptcp_subflow_context *subflow)
1241 {
1242 	pr_fallback(msk);
1243 	subflow->request_mptcp = 0;
1244 	__mptcp_do_fallback(msk);
1245 }
1246 
mptcp_check_infinite_map(struct sk_buff * skb)1247 static inline bool mptcp_check_infinite_map(struct sk_buff *skb)
1248 {
1249 	struct mptcp_ext *mpext;
1250 
1251 	mpext = skb ? mptcp_get_ext(skb) : NULL;
1252 	if (mpext && mpext->infinite_map)
1253 		return true;
1254 
1255 	return false;
1256 }
1257 
is_active_ssk(struct mptcp_subflow_context * subflow)1258 static inline bool is_active_ssk(struct mptcp_subflow_context *subflow)
1259 {
1260 	return (subflow->request_mptcp || subflow->request_join);
1261 }
1262 
subflow_simultaneous_connect(struct sock * sk)1263 static inline bool subflow_simultaneous_connect(struct sock *sk)
1264 {
1265 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1266 
1267 	return (1 << sk->sk_state) &
1268 	       (TCPF_ESTABLISHED | TCPF_FIN_WAIT1 | TCPF_FIN_WAIT2 | TCPF_CLOSING) &&
1269 	       is_active_ssk(subflow) &&
1270 	       !subflow->conn_finished;
1271 }
1272 
1273 #ifdef CONFIG_SYN_COOKIES
1274 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1275 				       struct sk_buff *skb);
1276 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1277 					struct sk_buff *skb);
1278 void __init mptcp_join_cookie_init(void);
1279 #else
1280 static inline void
subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock * subflow_req,struct sk_buff * skb)1281 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1282 				  struct sk_buff *skb) {}
1283 static inline bool
mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock * subflow_req,struct sk_buff * skb)1284 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1285 				   struct sk_buff *skb)
1286 {
1287 	return false;
1288 }
1289 
mptcp_join_cookie_init(void)1290 static inline void mptcp_join_cookie_init(void) {}
1291 #endif
1292 
1293 #endif /* __MPTCP_PROTOCOL_H */
1294