xref: /linux/net/mptcp/protocol.h (revision 3d041393)
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 	u16	suboptions;
153 	u32	token;
154 	u32	nonce;
155 	u16	use_map:1,
156 		dsn64:1,
157 		data_fin:1,
158 		use_ack:1,
159 		ack64:1,
160 		mpc_map:1,
161 		reset_reason:4,
162 		reset_transient:1,
163 		echo:1,
164 		backup:1,
165 		deny_join_id0:1,
166 		__unused:2;
167 	u8	join_id;
168 	u64	thmac;
169 	u8	hmac[MPTCPOPT_HMAC_LEN];
170 	struct mptcp_addr_info addr;
171 	struct mptcp_rm_list rm_list;
172 	u64	ahmac;
173 	u64	fail_seq;
174 };
175 
mptcp_option(u8 subopt,u8 len,u8 nib,u8 field)176 static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field)
177 {
178 	return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) |
179 		     ((nib & 0xF) << 8) | field);
180 }
181 
182 enum mptcp_pm_status {
183 	MPTCP_PM_ADD_ADDR_RECEIVED,
184 	MPTCP_PM_ADD_ADDR_SEND_ACK,
185 	MPTCP_PM_RM_ADDR_RECEIVED,
186 	MPTCP_PM_ESTABLISHED,
187 	MPTCP_PM_SUBFLOW_ESTABLISHED,
188 	MPTCP_PM_ALREADY_ESTABLISHED,	/* persistent status, set after ESTABLISHED event */
189 	MPTCP_PM_MPC_ENDPOINT_ACCOUNTED /* persistent status, set after MPC local address is
190 					 * accounted int id_avail_bitmap
191 					 */
192 };
193 
194 enum mptcp_pm_type {
195 	MPTCP_PM_TYPE_KERNEL = 0,
196 	MPTCP_PM_TYPE_USERSPACE,
197 
198 	__MPTCP_PM_TYPE_NR,
199 	__MPTCP_PM_TYPE_MAX = __MPTCP_PM_TYPE_NR - 1,
200 };
201 
202 /* Status bits below MPTCP_PM_ALREADY_ESTABLISHED need pm worker actions */
203 #define MPTCP_PM_WORK_MASK ((1 << MPTCP_PM_ALREADY_ESTABLISHED) - 1)
204 
205 enum mptcp_addr_signal_status {
206 	MPTCP_ADD_ADDR_SIGNAL,
207 	MPTCP_ADD_ADDR_ECHO,
208 	MPTCP_RM_ADDR_SIGNAL,
209 };
210 
211 /* max value of mptcp_addr_info.id */
212 #define MPTCP_PM_MAX_ADDR_ID		U8_MAX
213 
214 struct mptcp_pm_data {
215 	struct mptcp_addr_info local;
216 	struct mptcp_addr_info remote;
217 	struct list_head anno_list;
218 	struct list_head userspace_pm_local_addr_list;
219 
220 	spinlock_t	lock;		/*protects the whole PM data */
221 
222 	u8		addr_signal;
223 	bool		server_side;
224 	bool		work_pending;
225 	bool		accept_addr;
226 	bool		accept_subflow;
227 	bool		remote_deny_join_id0;
228 	u8		add_addr_signaled;
229 	u8		add_addr_accepted;
230 	u8		local_addr_used;
231 	u8		pm_type;
232 	u8		subflows;
233 	u8		status;
234 	DECLARE_BITMAP(id_avail_bitmap, MPTCP_PM_MAX_ADDR_ID + 1);
235 	struct mptcp_rm_list rm_list_tx;
236 	struct mptcp_rm_list rm_list_rx;
237 };
238 
239 struct mptcp_pm_local {
240 	struct mptcp_addr_info	addr;
241 	u8			flags;
242 	int			ifindex;
243 };
244 
245 struct mptcp_pm_addr_entry {
246 	struct list_head	list;
247 	struct mptcp_addr_info	addr;
248 	u8			flags;
249 	int			ifindex;
250 	struct socket		*lsk;
251 };
252 
253 struct mptcp_data_frag {
254 	struct list_head list;
255 	u64 data_seq;
256 	u16 data_len;
257 	u16 offset;
258 	u16 overhead;
259 	u16 already_sent;
260 	struct page *page;
261 };
262 
263 /* MPTCP connection sock */
264 struct mptcp_sock {
265 	/* inet_connection_sock must be the first member */
266 	struct inet_connection_sock sk;
267 	u64		local_key;		/* protected by the first subflow socket lock
268 						 * lockless access read
269 						 */
270 	u64		remote_key;		/* same as above */
271 	u64		write_seq;
272 	u64		bytes_sent;
273 	u64		snd_nxt;
274 	u64		bytes_received;
275 	u64		ack_seq;
276 	atomic64_t	rcv_wnd_sent;
277 	u64		rcv_data_fin_seq;
278 	u64		bytes_retrans;
279 	u64		bytes_consumed;
280 	int		rmem_fwd_alloc;
281 	int		snd_burst;
282 	int		old_wspace;
283 	u64		recovery_snd_nxt;	/* in recovery mode accept up to this seq;
284 						 * recovery related fields are under data_lock
285 						 * protection
286 						 */
287 	u64		bytes_acked;
288 	u64		snd_una;
289 	u64		wnd_end;
290 	u32		last_data_sent;
291 	u32		last_data_recv;
292 	u32		last_ack_recv;
293 	unsigned long	timer_ival;
294 	u32		token;
295 	int		rmem_released;
296 	unsigned long	flags;
297 	unsigned long	cb_flags;
298 	bool		recovery;		/* closing subflow write queue reinjected */
299 	bool		can_ack;
300 	bool		fully_established;
301 	bool		rcv_data_fin;
302 	bool		snd_data_fin_enable;
303 	bool		rcv_fastclose;
304 	bool		use_64bit_ack; /* Set when we received a 64-bit DSN */
305 	bool		csum_enabled;
306 	bool		allow_infinite_fallback;
307 	u8		pending_state; /* A subflow asked to set this sk_state,
308 					* protected by the msk data lock
309 					*/
310 	u8		mpc_endpoint_id;
311 	u8		recvmsg_inq:1,
312 			cork:1,
313 			nodelay:1,
314 			fastopening:1,
315 			in_accept_queue:1,
316 			free_first:1,
317 			rcvspace_init:1;
318 	u32		notsent_lowat;
319 	int		keepalive_cnt;
320 	int		keepalive_idle;
321 	int		keepalive_intvl;
322 	struct work_struct work;
323 	struct sk_buff  *ooo_last_skb;
324 	struct rb_root  out_of_order_queue;
325 	struct sk_buff_head receive_queue;
326 	struct list_head conn_list;
327 	struct list_head rtx_queue;
328 	struct mptcp_data_frag *first_pending;
329 	struct list_head join_list;
330 	struct sock	*first; /* The mptcp ops can safely dereference, using suitable
331 				 * ONCE annotation, the subflow outside the socket
332 				 * lock as such sock is freed after close().
333 				 */
334 	struct mptcp_pm_data	pm;
335 	struct mptcp_sched_ops	*sched;
336 	struct {
337 		u32	space;	/* bytes copied in last measurement window */
338 		u32	copied; /* bytes copied in this measurement window */
339 		u64	time;	/* start time of measurement window */
340 		u64	rtt_us; /* last maximum rtt of subflows */
341 	} rcvq_space;
342 	u8		scaling_ratio;
343 
344 	u32		subflow_id;
345 	u32		setsockopt_seq;
346 	char		ca_name[TCP_CA_NAME_MAX];
347 };
348 
349 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock)
350 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock)
351 
352 #define mptcp_for_each_subflow(__msk, __subflow)			\
353 	list_for_each_entry(__subflow, &((__msk)->conn_list), node)
354 #define mptcp_for_each_subflow_safe(__msk, __subflow, __tmp)			\
355 	list_for_each_entry_safe(__subflow, __tmp, &((__msk)->conn_list), node)
356 
357 extern struct genl_family mptcp_genl_family;
358 
msk_owned_by_me(const struct mptcp_sock * msk)359 static inline void msk_owned_by_me(const struct mptcp_sock *msk)
360 {
361 	sock_owned_by_me((const struct sock *)msk);
362 }
363 
364 #ifdef CONFIG_DEBUG_NET
365 /* MPTCP-specific: we might (indirectly) call this helper with the wrong sk */
366 #undef tcp_sk
367 #define tcp_sk(ptr) ({								\
368 	typeof(ptr) _ptr = (ptr);						\
369 	WARN_ON(_ptr->sk_protocol != IPPROTO_TCP);				\
370 	container_of_const(_ptr, struct tcp_sock, inet_conn.icsk_inet.sk);	\
371 })
372 #define mptcp_sk(ptr) ({						\
373 	typeof(ptr) _ptr = (ptr);					\
374 	WARN_ON(_ptr->sk_protocol != IPPROTO_MPTCP);			\
375 	container_of_const(_ptr, struct mptcp_sock, sk.icsk_inet.sk);	\
376 })
377 
378 #else /* !CONFIG_DEBUG_NET */
379 #define mptcp_sk(ptr) container_of_const(ptr, struct mptcp_sock, sk.icsk_inet.sk)
380 #endif
381 
382 /* the msk socket don't use the backlog, also account for the bulk
383  * free memory
384  */
__mptcp_rmem(const struct sock * sk)385 static inline int __mptcp_rmem(const struct sock *sk)
386 {
387 	return atomic_read(&sk->sk_rmem_alloc) - READ_ONCE(mptcp_sk(sk)->rmem_released);
388 }
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) - __mptcp_rmem(sk));
403 }
404 
mptcp_send_head(const struct sock * sk)405 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk)
406 {
407 	const struct mptcp_sock *msk = mptcp_sk(sk);
408 
409 	return READ_ONCE(msk->first_pending);
410 }
411 
mptcp_send_next(struct sock * sk)412 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk)
413 {
414 	struct mptcp_sock *msk = mptcp_sk(sk);
415 	struct mptcp_data_frag *cur;
416 
417 	cur = msk->first_pending;
418 	return list_is_last(&cur->list, &msk->rtx_queue) ? NULL :
419 						     list_next_entry(cur, list);
420 }
421 
mptcp_pending_tail(const struct sock * sk)422 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk)
423 {
424 	const struct mptcp_sock *msk = mptcp_sk(sk);
425 
426 	if (!msk->first_pending)
427 		return NULL;
428 
429 	if (WARN_ON_ONCE(list_empty(&msk->rtx_queue)))
430 		return NULL;
431 
432 	return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list);
433 }
434 
mptcp_rtx_head(struct sock * sk)435 static inline struct mptcp_data_frag *mptcp_rtx_head(struct sock *sk)
436 {
437 	struct mptcp_sock *msk = mptcp_sk(sk);
438 
439 	if (msk->snd_una == msk->snd_nxt)
440 		return NULL;
441 
442 	return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list);
443 }
444 
445 struct csum_pseudo_header {
446 	__be64 data_seq;
447 	__be32 subflow_seq;
448 	__be16 data_len;
449 	__sum16 csum;
450 };
451 
452 struct mptcp_subflow_request_sock {
453 	struct	tcp_request_sock sk;
454 	u16	mp_capable : 1,
455 		mp_join : 1,
456 		backup : 1,
457 		request_bkup : 1,
458 		csum_reqd : 1,
459 		allow_join_id0 : 1;
460 	u8	local_id;
461 	u8	remote_id;
462 	u64	local_key;
463 	u64	idsn;
464 	u32	token;
465 	u32	ssn_offset;
466 	u64	thmac;
467 	u32	local_nonce;
468 	u32	remote_nonce;
469 	struct mptcp_sock	*msk;
470 	struct hlist_nulls_node token_node;
471 };
472 
473 static inline struct mptcp_subflow_request_sock *
mptcp_subflow_rsk(const struct request_sock * rsk)474 mptcp_subflow_rsk(const struct request_sock *rsk)
475 {
476 	return (struct mptcp_subflow_request_sock *)rsk;
477 }
478 
479 struct mptcp_delegated_action {
480 	struct napi_struct napi;
481 	struct list_head head;
482 };
483 
484 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions);
485 
486 #define MPTCP_DELEGATE_SCHEDULED	0
487 #define MPTCP_DELEGATE_SEND		1
488 #define MPTCP_DELEGATE_ACK		2
489 #define MPTCP_DELEGATE_SNDBUF		3
490 
491 #define MPTCP_DELEGATE_ACTIONS_MASK	(~BIT(MPTCP_DELEGATE_SCHEDULED))
492 /* MPTCP subflow context */
493 struct mptcp_subflow_context {
494 	struct	list_head node;/* conn_list of subflows */
495 
496 	struct_group(reset,
497 
498 	unsigned long avg_pacing_rate; /* protected by msk socket lock */
499 	u64	local_key;
500 	u64	remote_key;
501 	u64	idsn;
502 	u64	map_seq;
503 	u32	snd_isn;
504 	u32	token;
505 	u32	rel_write_seq;
506 	u32	map_subflow_seq;
507 	u32	ssn_offset;
508 	u32	map_data_len;
509 	__wsum	map_data_csum;
510 	u32	map_csum_len;
511 	u32	request_mptcp : 1,  /* send MP_CAPABLE */
512 		request_join : 1,   /* send MP_JOIN */
513 		request_bkup : 1,
514 		mp_capable : 1,	    /* remote is MPTCP capable */
515 		mp_join : 1,	    /* remote is JOINing */
516 		fully_established : 1,	    /* path validated */
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 : 8;
536 	bool	data_avail;
537 	bool	scheduled;
538 	bool	pm_listener;	    /* a listener managed by the kernel PM? */
539 	u32	remote_nonce;
540 	u64	thmac;
541 	u32	local_nonce;
542 	u32	remote_token;
543 	union {
544 		u8	hmac[MPTCPOPT_HMAC_LEN]; /* MPJ subflow only */
545 		u64	iasn;	    /* initial ack sequence number, MPC subflows only */
546 	};
547 	s16	local_id;	    /* if negative not initialized yet */
548 	u8	remote_id;
549 	u8	reset_seen:1;
550 	u8	reset_transient:1;
551 	u8	reset_reason:4;
552 	u8	stale_count;
553 
554 	u32	subflow_id;
555 
556 	long	delegated_status;
557 	unsigned long	fail_tout;
558 
559 	);
560 
561 	struct	list_head delegated_node;   /* link into delegated_action, protected by local BH */
562 
563 	u32	setsockopt_seq;
564 	u32	stale_rcv_tstamp;
565 	int     cached_sndbuf;	    /* sndbuf size when last synced with the msk sndbuf,
566 				     * protected by the msk socket lock
567 				     */
568 
569 	struct	sock *tcp_sock;	    /* tcp sk backpointer */
570 	struct	sock *conn;	    /* parent mptcp_sock */
571 	const	struct inet_connection_sock_af_ops *icsk_af_ops;
572 	void	(*tcp_state_change)(struct sock *sk);
573 	void	(*tcp_error_report)(struct sock *sk);
574 
575 	struct	rcu_head rcu;
576 };
577 
578 static inline struct mptcp_subflow_context *
mptcp_subflow_ctx(const struct sock * sk)579 mptcp_subflow_ctx(const struct sock *sk)
580 {
581 	const struct inet_connection_sock *icsk = inet_csk(sk);
582 
583 	/* Use RCU on icsk_ulp_data only for sock diag code */
584 	return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data;
585 }
586 
587 static inline struct sock *
mptcp_subflow_tcp_sock(const struct mptcp_subflow_context * subflow)588 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow)
589 {
590 	return subflow->tcp_sock;
591 }
592 
593 static inline void
mptcp_subflow_ctx_reset(struct mptcp_subflow_context * subflow)594 mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow)
595 {
596 	memset(&subflow->reset, 0, sizeof(subflow->reset));
597 	subflow->request_mptcp = 1;
598 	WRITE_ONCE(subflow->local_id, -1);
599 }
600 
601 /* Convert reset reasons in MPTCP to enum sk_rst_reason type */
602 static inline enum sk_rst_reason
sk_rst_convert_mptcp_reason(u32 reason)603 sk_rst_convert_mptcp_reason(u32 reason)
604 {
605 	switch (reason) {
606 	case MPTCP_RST_EUNSPEC:
607 		return SK_RST_REASON_MPTCP_RST_EUNSPEC;
608 	case MPTCP_RST_EMPTCP:
609 		return SK_RST_REASON_MPTCP_RST_EMPTCP;
610 	case MPTCP_RST_ERESOURCE:
611 		return SK_RST_REASON_MPTCP_RST_ERESOURCE;
612 	case MPTCP_RST_EPROHIBIT:
613 		return SK_RST_REASON_MPTCP_RST_EPROHIBIT;
614 	case MPTCP_RST_EWQ2BIG:
615 		return SK_RST_REASON_MPTCP_RST_EWQ2BIG;
616 	case MPTCP_RST_EBADPERF:
617 		return SK_RST_REASON_MPTCP_RST_EBADPERF;
618 	case MPTCP_RST_EMIDDLEBOX:
619 		return SK_RST_REASON_MPTCP_RST_EMIDDLEBOX;
620 	default:
621 		/* It should not happen, or else errors may occur
622 		 * in MPTCP layer
623 		 */
624 		return SK_RST_REASON_ERROR;
625 	}
626 }
627 
628 static inline void
mptcp_send_active_reset_reason(struct sock * sk)629 mptcp_send_active_reset_reason(struct sock *sk)
630 {
631 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
632 	enum sk_rst_reason reason;
633 
634 	reason = sk_rst_convert_mptcp_reason(subflow->reset_reason);
635 	tcp_send_active_reset(sk, GFP_ATOMIC, reason);
636 }
637 
638 static inline u64
mptcp_subflow_get_map_offset(const struct mptcp_subflow_context * subflow)639 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow)
640 {
641 	return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq -
642 		      subflow->ssn_offset -
643 		      subflow->map_subflow_seq;
644 }
645 
646 static inline u64
mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context * subflow)647 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow)
648 {
649 	return subflow->map_seq + mptcp_subflow_get_map_offset(subflow);
650 }
651 
652 void mptcp_subflow_process_delegated(struct sock *ssk, long actions);
653 
mptcp_subflow_delegate(struct mptcp_subflow_context * subflow,int action)654 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action)
655 {
656 	long old, set_bits = BIT(MPTCP_DELEGATE_SCHEDULED) | BIT(action);
657 	struct mptcp_delegated_action *delegated;
658 	bool schedule;
659 
660 	/* the caller held the subflow bh socket lock */
661 	lockdep_assert_in_softirq();
662 
663 	/* The implied barrier pairs with tcp_release_cb_override()
664 	 * mptcp_napi_poll(), and ensures the below list check sees list
665 	 * updates done prior to delegated status bits changes
666 	 */
667 	old = set_mask_bits(&subflow->delegated_status, 0, set_bits);
668 	if (!(old & BIT(MPTCP_DELEGATE_SCHEDULED))) {
669 		if (WARN_ON_ONCE(!list_empty(&subflow->delegated_node)))
670 			return;
671 
672 		delegated = this_cpu_ptr(&mptcp_delegated_actions);
673 		schedule = list_empty(&delegated->head);
674 		list_add_tail(&subflow->delegated_node, &delegated->head);
675 		sock_hold(mptcp_subflow_tcp_sock(subflow));
676 		if (schedule)
677 			napi_schedule(&delegated->napi);
678 	}
679 }
680 
681 static inline struct mptcp_subflow_context *
mptcp_subflow_delegated_next(struct mptcp_delegated_action * delegated)682 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated)
683 {
684 	struct mptcp_subflow_context *ret;
685 
686 	if (list_empty(&delegated->head))
687 		return NULL;
688 
689 	ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node);
690 	list_del_init(&ret->delegated_node);
691 	return ret;
692 }
693 
694 int mptcp_is_enabled(const struct net *net);
695 unsigned int mptcp_get_add_addr_timeout(const struct net *net);
696 int mptcp_is_checksum_enabled(const struct net *net);
697 int mptcp_allow_join_id0(const struct net *net);
698 unsigned int mptcp_stale_loss_cnt(const struct net *net);
699 unsigned int mptcp_close_timeout(const struct sock *sk);
700 int mptcp_get_pm_type(const struct net *net);
701 const char *mptcp_get_scheduler(const struct net *net);
702 
703 void mptcp_active_disable(struct sock *sk);
704 bool mptcp_active_should_disable(struct sock *ssk);
705 void mptcp_active_enable(struct sock *sk);
706 
707 void mptcp_get_available_schedulers(char *buf, size_t maxlen);
708 void __mptcp_subflow_fully_established(struct mptcp_sock *msk,
709 				       struct mptcp_subflow_context *subflow,
710 				       const struct mptcp_options_received *mp_opt);
711 bool __mptcp_retransmit_pending_data(struct sock *sk);
712 void mptcp_check_and_set_pending(struct sock *sk);
713 void __mptcp_push_pending(struct sock *sk, unsigned int flags);
714 bool mptcp_subflow_data_available(struct sock *sk);
715 void __init mptcp_subflow_init(void);
716 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how);
717 void mptcp_close_ssk(struct sock *sk, struct sock *ssk,
718 		     struct mptcp_subflow_context *subflow);
719 void __mptcp_subflow_send_ack(struct sock *ssk);
720 void mptcp_subflow_reset(struct sock *ssk);
721 void mptcp_subflow_queue_clean(struct sock *sk, struct sock *ssk);
722 void mptcp_sock_graft(struct sock *sk, struct socket *parent);
723 struct sock *__mptcp_nmpc_sk(struct mptcp_sock *msk);
724 bool __mptcp_close(struct sock *sk, long timeout);
725 void mptcp_cancel_work(struct sock *sk);
726 void __mptcp_unaccepted_force_close(struct sock *sk);
727 void mptcp_set_owner_r(struct sk_buff *skb, struct sock *sk);
728 void mptcp_set_state(struct sock *sk, int state);
729 
730 bool mptcp_addresses_equal(const struct mptcp_addr_info *a,
731 			   const struct mptcp_addr_info *b, bool use_port);
732 void mptcp_local_address(const struct sock_common *skc, struct mptcp_addr_info *addr);
733 
734 /* called with sk socket lock held */
735 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_pm_local *local,
736 			    const struct mptcp_addr_info *remote);
737 int mptcp_subflow_create_socket(struct sock *sk, unsigned short family,
738 				struct socket **new_sock);
739 void mptcp_info2sockaddr(const struct mptcp_addr_info *info,
740 			 struct sockaddr_storage *addr,
741 			 unsigned short family);
742 struct mptcp_sched_ops *mptcp_sched_find(const char *name);
743 int mptcp_register_scheduler(struct mptcp_sched_ops *sched);
744 void mptcp_unregister_scheduler(struct mptcp_sched_ops *sched);
745 void mptcp_sched_init(void);
746 int mptcp_init_sched(struct mptcp_sock *msk,
747 		     struct mptcp_sched_ops *sched);
748 void mptcp_release_sched(struct mptcp_sock *msk);
749 void mptcp_subflow_set_scheduled(struct mptcp_subflow_context *subflow,
750 				 bool scheduled);
751 struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk);
752 struct sock *mptcp_subflow_get_retrans(struct mptcp_sock *msk);
753 int mptcp_sched_get_send(struct mptcp_sock *msk);
754 int mptcp_sched_get_retrans(struct mptcp_sock *msk);
755 
mptcp_data_avail(const struct mptcp_sock * msk)756 static inline u64 mptcp_data_avail(const struct mptcp_sock *msk)
757 {
758 	return READ_ONCE(msk->bytes_received) - READ_ONCE(msk->bytes_consumed);
759 }
760 
mptcp_epollin_ready(const struct sock * sk)761 static inline bool mptcp_epollin_ready(const struct sock *sk)
762 {
763 	/* mptcp doesn't have to deal with small skbs in the receive queue,
764 	 * at it can always coalesce them
765 	 */
766 	return (mptcp_data_avail(mptcp_sk(sk)) >= sk->sk_rcvlowat) ||
767 	       (mem_cgroup_sockets_enabled && sk->sk_memcg &&
768 		mem_cgroup_under_socket_pressure(sk->sk_memcg)) ||
769 	       READ_ONCE(tcp_memory_pressure);
770 }
771 
772 int mptcp_set_rcvlowat(struct sock *sk, int val);
773 
__tcp_can_send(const struct sock * ssk)774 static inline bool __tcp_can_send(const struct sock *ssk)
775 {
776 	/* only send if our side has not closed yet */
777 	return ((1 << inet_sk_state_load(ssk)) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT));
778 }
779 
__mptcp_subflow_active(struct mptcp_subflow_context * subflow)780 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow)
781 {
782 	/* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */
783 	if (subflow->request_join && !subflow->fully_established)
784 		return false;
785 
786 	return __tcp_can_send(mptcp_subflow_tcp_sock(subflow));
787 }
788 
789 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow);
790 
791 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow);
792 
793 void mptcp_subflow_drop_ctx(struct sock *ssk);
794 
mptcp_subflow_tcp_fallback(struct sock * sk,struct mptcp_subflow_context * ctx)795 static inline void mptcp_subflow_tcp_fallback(struct sock *sk,
796 					      struct mptcp_subflow_context *ctx)
797 {
798 	sk->sk_data_ready = sock_def_readable;
799 	sk->sk_state_change = ctx->tcp_state_change;
800 	sk->sk_write_space = sk_stream_write_space;
801 	sk->sk_error_report = ctx->tcp_error_report;
802 
803 	inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops;
804 }
805 
806 void __init mptcp_proto_init(void);
807 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
808 int __init mptcp_proto_v6_init(void);
809 #endif
810 
811 struct sock *mptcp_sk_clone_init(const struct sock *sk,
812 				 const struct mptcp_options_received *mp_opt,
813 				 struct sock *ssk,
814 				 struct request_sock *req);
815 void mptcp_get_options(const struct sk_buff *skb,
816 		       struct mptcp_options_received *mp_opt);
817 
818 void mptcp_finish_connect(struct sock *sk);
819 void __mptcp_sync_state(struct sock *sk, int state);
820 void mptcp_reset_tout_timer(struct mptcp_sock *msk, unsigned long fail_tout);
821 
mptcp_stop_tout_timer(struct sock * sk)822 static inline void mptcp_stop_tout_timer(struct sock *sk)
823 {
824 	if (!inet_csk(sk)->icsk_mtup.probe_timestamp)
825 		return;
826 
827 	sk_stop_timer(sk, &sk->sk_timer);
828 	inet_csk(sk)->icsk_mtup.probe_timestamp = 0;
829 }
830 
mptcp_set_close_tout(struct sock * sk,unsigned long tout)831 static inline void mptcp_set_close_tout(struct sock *sk, unsigned long tout)
832 {
833 	/* avoid 0 timestamp, as that means no close timeout */
834 	inet_csk(sk)->icsk_mtup.probe_timestamp = tout ? : 1;
835 }
836 
mptcp_start_tout_timer(struct sock * sk)837 static inline void mptcp_start_tout_timer(struct sock *sk)
838 {
839 	mptcp_set_close_tout(sk, tcp_jiffies32);
840 	mptcp_reset_tout_timer(mptcp_sk(sk), 0);
841 }
842 
mptcp_is_fully_established(struct sock * sk)843 static inline bool mptcp_is_fully_established(struct sock *sk)
844 {
845 	return inet_sk_state_load(sk) == TCP_ESTABLISHED &&
846 	       READ_ONCE(mptcp_sk(sk)->fully_established);
847 }
848 
849 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk);
850 void mptcp_data_ready(struct sock *sk, struct sock *ssk);
851 bool mptcp_finish_join(struct sock *sk);
852 bool mptcp_schedule_work(struct sock *sk);
853 int mptcp_setsockopt(struct sock *sk, int level, int optname,
854 		     sockptr_t optval, unsigned int optlen);
855 int mptcp_getsockopt(struct sock *sk, int level, int optname,
856 		     char __user *optval, int __user *option);
857 
858 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq);
mptcp_expand_seq(u64 old_seq,u64 cur_seq,bool use_64bit)859 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit)
860 {
861 	if (use_64bit)
862 		return cur_seq;
863 
864 	return __mptcp_expand_seq(old_seq, cur_seq);
865 }
866 void __mptcp_check_push(struct sock *sk, struct sock *ssk);
867 void __mptcp_data_acked(struct sock *sk);
868 void __mptcp_error_report(struct sock *sk);
869 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)870 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk)
871 {
872 	return READ_ONCE(msk->snd_data_fin_enable) &&
873 	       READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt);
874 }
875 
mptcp_notsent_lowat(const struct sock * sk)876 static inline u32 mptcp_notsent_lowat(const struct sock *sk)
877 {
878 	struct net *net = sock_net(sk);
879 	u32 val;
880 
881 	val = READ_ONCE(mptcp_sk(sk)->notsent_lowat);
882 	return val ?: READ_ONCE(net->ipv4.sysctl_tcp_notsent_lowat);
883 }
884 
mptcp_stream_memory_free(const struct sock * sk,int wake)885 static inline bool mptcp_stream_memory_free(const struct sock *sk, int wake)
886 {
887 	const struct mptcp_sock *msk = mptcp_sk(sk);
888 	u32 notsent_bytes;
889 
890 	notsent_bytes = READ_ONCE(msk->write_seq) - READ_ONCE(msk->snd_nxt);
891 	return (notsent_bytes << wake) < mptcp_notsent_lowat(sk);
892 }
893 
__mptcp_stream_is_writeable(const struct sock * sk,int wake)894 static inline bool __mptcp_stream_is_writeable(const struct sock *sk, int wake)
895 {
896 	return mptcp_stream_memory_free(sk, wake) &&
897 	       __sk_stream_is_writeable(sk, wake);
898 }
899 
mptcp_write_space(struct sock * sk)900 static inline void mptcp_write_space(struct sock *sk)
901 {
902 	/* pairs with memory barrier in mptcp_poll */
903 	smp_mb();
904 	if (mptcp_stream_memory_free(sk, 1))
905 		sk_stream_write_space(sk);
906 }
907 
__mptcp_sync_sndbuf(struct sock * sk)908 static inline void __mptcp_sync_sndbuf(struct sock *sk)
909 {
910 	struct mptcp_subflow_context *subflow;
911 	int ssk_sndbuf, new_sndbuf;
912 
913 	if (sk->sk_userlocks & SOCK_SNDBUF_LOCK)
914 		return;
915 
916 	new_sndbuf = READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_wmem[0]);
917 	mptcp_for_each_subflow(mptcp_sk(sk), subflow) {
918 		ssk_sndbuf =  READ_ONCE(mptcp_subflow_tcp_sock(subflow)->sk_sndbuf);
919 
920 		subflow->cached_sndbuf = ssk_sndbuf;
921 		new_sndbuf += ssk_sndbuf;
922 	}
923 
924 	/* the msk max wmem limit is <nr_subflows> * tcp wmem[2] */
925 	WRITE_ONCE(sk->sk_sndbuf, new_sndbuf);
926 	mptcp_write_space(sk);
927 }
928 
929 /* The called held both the msk socket and the subflow socket locks,
930  * possibly under BH
931  */
__mptcp_propagate_sndbuf(struct sock * sk,struct sock * ssk)932 static inline void __mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
933 {
934 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
935 
936 	if (READ_ONCE(ssk->sk_sndbuf) != subflow->cached_sndbuf)
937 		__mptcp_sync_sndbuf(sk);
938 }
939 
940 /* the caller held only the subflow socket lock, either in process or
941  * BH context. Additionally this can be called under the msk data lock,
942  * so we can't acquire such lock here: let the delegate action acquires
943  * the needed locks in suitable order.
944  */
mptcp_propagate_sndbuf(struct sock * sk,struct sock * ssk)945 static inline void mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
946 {
947 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
948 
949 	if (likely(READ_ONCE(ssk->sk_sndbuf) == subflow->cached_sndbuf))
950 		return;
951 
952 	local_bh_disable();
953 	mptcp_subflow_delegate(subflow, MPTCP_DELEGATE_SNDBUF);
954 	local_bh_enable();
955 }
956 
957 void mptcp_destroy_common(struct mptcp_sock *msk, unsigned int flags);
958 
959 #define MPTCP_TOKEN_MAX_RETRIES	4
960 
961 void __init mptcp_token_init(void);
mptcp_token_init_request(struct request_sock * req)962 static inline void mptcp_token_init_request(struct request_sock *req)
963 {
964 	mptcp_subflow_rsk(req)->token_node.pprev = NULL;
965 }
966 
967 int mptcp_token_new_request(struct request_sock *req);
968 void mptcp_token_destroy_request(struct request_sock *req);
969 int mptcp_token_new_connect(struct sock *ssk);
970 void mptcp_token_accept(struct mptcp_subflow_request_sock *r,
971 			struct mptcp_sock *msk);
972 bool mptcp_token_exists(u32 token);
973 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token);
974 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot,
975 					 long *s_num);
976 void mptcp_token_destroy(struct mptcp_sock *msk);
977 
978 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn);
979 
980 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac);
981 __sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum);
982 
983 void __init mptcp_pm_init(void);
984 void mptcp_pm_data_init(struct mptcp_sock *msk);
985 void mptcp_pm_data_reset(struct mptcp_sock *msk);
986 int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info,
987 			struct mptcp_addr_info *addr);
988 int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info,
989 			 bool require_family,
990 			 struct mptcp_pm_addr_entry *entry);
991 bool mptcp_pm_addr_families_match(const struct sock *sk,
992 				  const struct mptcp_addr_info *loc,
993 				  const struct mptcp_addr_info *rem);
994 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
995 void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
996 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side);
997 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk);
998 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk);
999 void mptcp_pm_connection_closed(struct mptcp_sock *msk);
1000 void mptcp_pm_subflow_established(struct mptcp_sock *msk);
1001 bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk);
1002 void mptcp_pm_subflow_check_next(struct mptcp_sock *msk,
1003 				 const struct mptcp_subflow_context *subflow);
1004 void mptcp_pm_add_addr_received(const struct sock *ssk,
1005 				const struct mptcp_addr_info *addr);
1006 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk,
1007 			      const struct mptcp_addr_info *addr);
1008 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk);
1009 bool mptcp_pm_nl_is_init_remote_addr(struct mptcp_sock *msk,
1010 				     const struct mptcp_addr_info *remote);
1011 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk);
1012 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk,
1013 			       const struct mptcp_rm_list *rm_list);
1014 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup);
1015 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq);
1016 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk,
1017 				 struct mptcp_addr_info *addr,
1018 				 struct mptcp_addr_info *rem,
1019 				 u8 bkup);
1020 bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk,
1021 			      const struct mptcp_addr_info *addr);
1022 void mptcp_pm_free_anno_list(struct mptcp_sock *msk);
1023 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk);
1024 struct mptcp_pm_add_entry *
1025 mptcp_pm_del_add_timer(struct mptcp_sock *msk,
1026 		       const struct mptcp_addr_info *addr, bool check_id);
1027 struct mptcp_pm_add_entry *
1028 mptcp_lookup_anno_list_by_saddr(const struct mptcp_sock *msk,
1029 				const struct mptcp_addr_info *addr);
1030 int mptcp_pm_set_flags(struct sk_buff *skb, struct genl_info *info);
1031 int mptcp_pm_nl_set_flags(struct sk_buff *skb, struct genl_info *info);
1032 int mptcp_userspace_pm_set_flags(struct sk_buff *skb, struct genl_info *info);
1033 int mptcp_pm_announce_addr(struct mptcp_sock *msk,
1034 			   const struct mptcp_addr_info *addr,
1035 			   bool echo);
1036 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
1037 void mptcp_pm_remove_addrs(struct mptcp_sock *msk, struct list_head *rm_list);
1038 
1039 void mptcp_free_local_addr_list(struct mptcp_sock *msk);
1040 
1041 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk,
1042 		 const struct sock *ssk, gfp_t gfp);
1043 void mptcp_event_addr_announced(const struct sock *ssk, const struct mptcp_addr_info *info);
1044 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id);
1045 void mptcp_event_pm_listener(const struct sock *ssk,
1046 			     enum mptcp_event_type event);
1047 bool mptcp_userspace_pm_active(const struct mptcp_sock *msk);
1048 
1049 void __mptcp_fastopen_gen_msk_ackseq(struct mptcp_sock *msk, struct mptcp_subflow_context *subflow,
1050 				     const struct mptcp_options_received *mp_opt);
1051 void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subflow,
1052 					      struct request_sock *req);
1053 int mptcp_nl_fill_addr(struct sk_buff *skb,
1054 		       struct mptcp_pm_addr_entry *entry);
1055 
mptcp_pm_should_add_signal(struct mptcp_sock * msk)1056 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk)
1057 {
1058 	return READ_ONCE(msk->pm.addr_signal) &
1059 		(BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO));
1060 }
1061 
mptcp_pm_should_add_signal_addr(struct mptcp_sock * msk)1062 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk)
1063 {
1064 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL);
1065 }
1066 
mptcp_pm_should_add_signal_echo(struct mptcp_sock * msk)1067 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk)
1068 {
1069 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO);
1070 }
1071 
mptcp_pm_should_rm_signal(struct mptcp_sock * msk)1072 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk)
1073 {
1074 	return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL);
1075 }
1076 
mptcp_pm_is_userspace(const struct mptcp_sock * msk)1077 static inline bool mptcp_pm_is_userspace(const struct mptcp_sock *msk)
1078 {
1079 	return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_USERSPACE;
1080 }
1081 
mptcp_pm_is_kernel(const struct mptcp_sock * msk)1082 static inline bool mptcp_pm_is_kernel(const struct mptcp_sock *msk)
1083 {
1084 	return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_KERNEL;
1085 }
1086 
mptcp_add_addr_len(int family,bool echo,bool port)1087 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port)
1088 {
1089 	u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE;
1090 
1091 	if (family == AF_INET6)
1092 		len = TCPOLEN_MPTCP_ADD_ADDR6_BASE;
1093 	if (!echo)
1094 		len += MPTCPOPT_THMAC_LEN;
1095 	/* account for 2 trailing 'nop' options */
1096 	if (port)
1097 		len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN;
1098 
1099 	return len;
1100 }
1101 
mptcp_rm_addr_len(const struct mptcp_rm_list * rm_list)1102 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list)
1103 {
1104 	if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX)
1105 		return -EINVAL;
1106 
1107 	return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1;
1108 }
1109 
1110 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, const struct sk_buff *skb,
1111 			      unsigned int opt_size, unsigned int remaining,
1112 			      struct mptcp_addr_info *addr, bool *echo,
1113 			      bool *drop_other_suboptions);
1114 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining,
1115 			     struct mptcp_rm_list *rm_list);
1116 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
1117 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1118 int mptcp_userspace_pm_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1119 bool mptcp_pm_is_backup(struct mptcp_sock *msk, struct sock_common *skc);
1120 bool mptcp_pm_nl_is_backup(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1121 bool mptcp_userspace_pm_is_backup(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1122 int mptcp_pm_dump_addr(struct sk_buff *msg, struct netlink_callback *cb);
1123 int mptcp_pm_nl_dump_addr(struct sk_buff *msg,
1124 			  struct netlink_callback *cb);
1125 int mptcp_userspace_pm_dump_addr(struct sk_buff *msg,
1126 				 struct netlink_callback *cb);
1127 int mptcp_pm_get_addr(struct sk_buff *skb, struct genl_info *info);
1128 int mptcp_pm_nl_get_addr(struct sk_buff *skb, struct genl_info *info);
1129 int mptcp_userspace_pm_get_addr(struct sk_buff *skb,
1130 				struct genl_info *info);
1131 
subflow_get_local_id(const struct mptcp_subflow_context * subflow)1132 static inline u8 subflow_get_local_id(const struct mptcp_subflow_context *subflow)
1133 {
1134 	int local_id = READ_ONCE(subflow->local_id);
1135 
1136 	if (local_id < 0)
1137 		return 0;
1138 	return local_id;
1139 }
1140 
1141 void __init mptcp_pm_nl_init(void);
1142 void mptcp_pm_nl_work(struct mptcp_sock *msk);
1143 unsigned int mptcp_pm_get_add_addr_signal_max(const struct mptcp_sock *msk);
1144 unsigned int mptcp_pm_get_add_addr_accept_max(const struct mptcp_sock *msk);
1145 unsigned int mptcp_pm_get_subflows_max(const struct mptcp_sock *msk);
1146 unsigned int mptcp_pm_get_local_addr_max(const struct mptcp_sock *msk);
1147 
1148 /* called under PM lock */
__mptcp_pm_close_subflow(struct mptcp_sock * msk)1149 static inline void __mptcp_pm_close_subflow(struct mptcp_sock *msk)
1150 {
1151 	if (--msk->pm.subflows < mptcp_pm_get_subflows_max(msk))
1152 		WRITE_ONCE(msk->pm.accept_subflow, true);
1153 }
1154 
mptcp_pm_close_subflow(struct mptcp_sock * msk)1155 static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk)
1156 {
1157 	spin_lock_bh(&msk->pm.lock);
1158 	__mptcp_pm_close_subflow(msk);
1159 	spin_unlock_bh(&msk->pm.lock);
1160 }
1161 
1162 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk);
1163 
mptcp_get_ext(const struct sk_buff * skb)1164 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb)
1165 {
1166 	return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP);
1167 }
1168 
1169 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops);
1170 
__mptcp_check_fallback(const struct mptcp_sock * msk)1171 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk)
1172 {
1173 	return test_bit(MPTCP_FALLBACK_DONE, &msk->flags);
1174 }
1175 
mptcp_check_fallback(const struct sock * sk)1176 static inline bool mptcp_check_fallback(const struct sock *sk)
1177 {
1178 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1179 	struct mptcp_sock *msk = mptcp_sk(subflow->conn);
1180 
1181 	return __mptcp_check_fallback(msk);
1182 }
1183 
__mptcp_do_fallback(struct mptcp_sock * msk)1184 static inline void __mptcp_do_fallback(struct mptcp_sock *msk)
1185 {
1186 	if (__mptcp_check_fallback(msk)) {
1187 		pr_debug("TCP fallback already done (msk=%p)\n", msk);
1188 		return;
1189 	}
1190 	set_bit(MPTCP_FALLBACK_DONE, &msk->flags);
1191 }
1192 
__mptcp_has_initial_subflow(const struct mptcp_sock * msk)1193 static inline bool __mptcp_has_initial_subflow(const struct mptcp_sock *msk)
1194 {
1195 	struct sock *ssk = READ_ONCE(msk->first);
1196 
1197 	return ssk && ((1 << inet_sk_state_load(ssk)) &
1198 		       (TCPF_ESTABLISHED | TCPF_SYN_SENT |
1199 			TCPF_SYN_RECV | TCPF_LISTEN));
1200 }
1201 
mptcp_do_fallback(struct sock * ssk)1202 static inline void mptcp_do_fallback(struct sock *ssk)
1203 {
1204 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
1205 	struct sock *sk = subflow->conn;
1206 	struct mptcp_sock *msk;
1207 
1208 	msk = mptcp_sk(sk);
1209 	__mptcp_do_fallback(msk);
1210 	if (READ_ONCE(msk->snd_data_fin_enable) && !(ssk->sk_shutdown & SEND_SHUTDOWN)) {
1211 		gfp_t saved_allocation = ssk->sk_allocation;
1212 
1213 		/* we are in a atomic (BH) scope, override ssk default for data
1214 		 * fin allocation
1215 		 */
1216 		ssk->sk_allocation = GFP_ATOMIC;
1217 		ssk->sk_shutdown |= SEND_SHUTDOWN;
1218 		tcp_shutdown(ssk, SEND_SHUTDOWN);
1219 		ssk->sk_allocation = saved_allocation;
1220 	}
1221 }
1222 
1223 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)\n", __func__, a)
1224 
mptcp_subflow_early_fallback(struct mptcp_sock * msk,struct mptcp_subflow_context * subflow)1225 static inline void mptcp_subflow_early_fallback(struct mptcp_sock *msk,
1226 						struct mptcp_subflow_context *subflow)
1227 {
1228 	pr_fallback(msk);
1229 	subflow->request_mptcp = 0;
1230 	__mptcp_do_fallback(msk);
1231 }
1232 
mptcp_check_infinite_map(struct sk_buff * skb)1233 static inline bool mptcp_check_infinite_map(struct sk_buff *skb)
1234 {
1235 	struct mptcp_ext *mpext;
1236 
1237 	mpext = skb ? mptcp_get_ext(skb) : NULL;
1238 	if (mpext && mpext->infinite_map)
1239 		return true;
1240 
1241 	return false;
1242 }
1243 
is_active_ssk(struct mptcp_subflow_context * subflow)1244 static inline bool is_active_ssk(struct mptcp_subflow_context *subflow)
1245 {
1246 	return (subflow->request_mptcp || subflow->request_join);
1247 }
1248 
subflow_simultaneous_connect(struct sock * sk)1249 static inline bool subflow_simultaneous_connect(struct sock *sk)
1250 {
1251 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1252 
1253 	return (1 << sk->sk_state) &
1254 	       (TCPF_ESTABLISHED | TCPF_FIN_WAIT1 | TCPF_FIN_WAIT2 | TCPF_CLOSING) &&
1255 	       is_active_ssk(subflow) &&
1256 	       !subflow->conn_finished;
1257 }
1258 
1259 #ifdef CONFIG_SYN_COOKIES
1260 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1261 				       struct sk_buff *skb);
1262 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1263 					struct sk_buff *skb);
1264 void __init mptcp_join_cookie_init(void);
1265 #else
1266 static inline void
subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock * subflow_req,struct sk_buff * skb)1267 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1268 				  struct sk_buff *skb) {}
1269 static inline bool
mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock * subflow_req,struct sk_buff * skb)1270 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1271 				   struct sk_buff *skb)
1272 {
1273 	return false;
1274 }
1275 
mptcp_join_cookie_init(void)1276 static inline void mptcp_join_cookie_init(void) {}
1277 #endif
1278 
1279 #endif /* __MPTCP_PROTOCOL_H */
1280