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
15 #define MPTCP_SUPPORTED_VERSION 1
16
17 /* MPTCP option bits */
18 #define OPTION_MPTCP_MPC_SYN BIT(0)
19 #define OPTION_MPTCP_MPC_SYNACK BIT(1)
20 #define OPTION_MPTCP_MPC_ACK BIT(2)
21 #define OPTION_MPTCP_MPJ_SYN BIT(3)
22 #define OPTION_MPTCP_MPJ_SYNACK BIT(4)
23 #define OPTION_MPTCP_MPJ_ACK BIT(5)
24 #define OPTION_MPTCP_ADD_ADDR BIT(6)
25 #define OPTION_MPTCP_RM_ADDR BIT(7)
26 #define OPTION_MPTCP_FASTCLOSE BIT(8)
27 #define OPTION_MPTCP_PRIO BIT(9)
28 #define OPTION_MPTCP_RST BIT(10)
29 #define OPTION_MPTCP_DSS BIT(11)
30 #define OPTION_MPTCP_FAIL BIT(12)
31
32 #define OPTION_MPTCP_CSUMREQD BIT(13)
33
34 #define OPTIONS_MPTCP_MPC (OPTION_MPTCP_MPC_SYN | OPTION_MPTCP_MPC_SYNACK | \
35 OPTION_MPTCP_MPC_ACK)
36 #define OPTIONS_MPTCP_MPJ (OPTION_MPTCP_MPJ_SYN | OPTION_MPTCP_MPJ_SYNACK | \
37 OPTION_MPTCP_MPJ_ACK)
38
39 /* MPTCP option subtypes */
40 #define MPTCPOPT_MP_CAPABLE 0
41 #define MPTCPOPT_MP_JOIN 1
42 #define MPTCPOPT_DSS 2
43 #define MPTCPOPT_ADD_ADDR 3
44 #define MPTCPOPT_RM_ADDR 4
45 #define MPTCPOPT_MP_PRIO 5
46 #define MPTCPOPT_MP_FAIL 6
47 #define MPTCPOPT_MP_FASTCLOSE 7
48 #define MPTCPOPT_RST 8
49
50 /* MPTCP suboption lengths */
51 #define TCPOLEN_MPTCP_MPC_SYN 4
52 #define TCPOLEN_MPTCP_MPC_SYNACK 12
53 #define TCPOLEN_MPTCP_MPC_ACK 20
54 #define TCPOLEN_MPTCP_MPC_ACK_DATA 22
55 #define TCPOLEN_MPTCP_MPJ_SYN 12
56 #define TCPOLEN_MPTCP_MPJ_SYNACK 16
57 #define TCPOLEN_MPTCP_MPJ_ACK 24
58 #define TCPOLEN_MPTCP_DSS_BASE 4
59 #define TCPOLEN_MPTCP_DSS_ACK32 4
60 #define TCPOLEN_MPTCP_DSS_ACK64 8
61 #define TCPOLEN_MPTCP_DSS_MAP32 10
62 #define TCPOLEN_MPTCP_DSS_MAP64 14
63 #define TCPOLEN_MPTCP_DSS_CHECKSUM 2
64 #define TCPOLEN_MPTCP_ADD_ADDR 16
65 #define TCPOLEN_MPTCP_ADD_ADDR_PORT 18
66 #define TCPOLEN_MPTCP_ADD_ADDR_BASE 8
67 #define TCPOLEN_MPTCP_ADD_ADDR_BASE_PORT 10
68 #define TCPOLEN_MPTCP_ADD_ADDR6 28
69 #define TCPOLEN_MPTCP_ADD_ADDR6_PORT 30
70 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE 20
71 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE_PORT 22
72 #define TCPOLEN_MPTCP_PORT_LEN 2
73 #define TCPOLEN_MPTCP_PORT_ALIGN 2
74 #define TCPOLEN_MPTCP_RM_ADDR_BASE 3
75 #define TCPOLEN_MPTCP_PRIO 3
76 #define TCPOLEN_MPTCP_PRIO_ALIGN 4
77 #define TCPOLEN_MPTCP_FASTCLOSE 12
78 #define TCPOLEN_MPTCP_RST 4
79 #define TCPOLEN_MPTCP_FAIL 12
80
81 #define TCPOLEN_MPTCP_MPC_ACK_DATA_CSUM (TCPOLEN_MPTCP_DSS_CHECKSUM + TCPOLEN_MPTCP_MPC_ACK_DATA)
82
83 /* MPTCP MP_JOIN flags */
84 #define MPTCPOPT_BACKUP BIT(0)
85 #define MPTCPOPT_HMAC_LEN 20
86 #define MPTCPOPT_THMAC_LEN 8
87
88 /* MPTCP MP_CAPABLE flags */
89 #define MPTCP_VERSION_MASK (0x0F)
90 #define MPTCP_CAP_CHECKSUM_REQD BIT(7)
91 #define MPTCP_CAP_EXTENSIBILITY BIT(6)
92 #define MPTCP_CAP_DENY_JOIN_ID0 BIT(5)
93 #define MPTCP_CAP_HMAC_SHA256 BIT(0)
94 #define MPTCP_CAP_FLAG_MASK (0x1F)
95
96 /* MPTCP DSS flags */
97 #define MPTCP_DSS_DATA_FIN BIT(4)
98 #define MPTCP_DSS_DSN64 BIT(3)
99 #define MPTCP_DSS_HAS_MAP BIT(2)
100 #define MPTCP_DSS_ACK64 BIT(1)
101 #define MPTCP_DSS_HAS_ACK BIT(0)
102 #define MPTCP_DSS_FLAG_MASK (0x1F)
103
104 /* MPTCP ADD_ADDR flags */
105 #define MPTCP_ADDR_ECHO BIT(0)
106
107 /* MPTCP MP_PRIO flags */
108 #define MPTCP_PRIO_BKUP BIT(0)
109
110 /* MPTCP TCPRST flags */
111 #define MPTCP_RST_TRANSIENT BIT(0)
112
113 /* MPTCP socket flags */
114 #define MPTCP_DATA_READY 0
115 #define MPTCP_NOSPACE 1
116 #define MPTCP_WORK_RTX 2
117 #define MPTCP_WORK_EOF 3
118 #define MPTCP_FALLBACK_DONE 4
119 #define MPTCP_WORK_CLOSE_SUBFLOW 5
120 #define MPTCP_PUSH_PENDING 6
121 #define MPTCP_CLEAN_UNA 7
122 #define MPTCP_ERROR_REPORT 8
123 #define MPTCP_RETRANSMIT 9
124 #define MPTCP_WORK_SYNC_SETSOCKOPT 10
125 #define MPTCP_CONNECTED 11
126
before64(__u64 seq1,__u64 seq2)127 static inline bool before64(__u64 seq1, __u64 seq2)
128 {
129 return (__s64)(seq1 - seq2) < 0;
130 }
131
132 #define after64(seq2, seq1) before64(seq1, seq2)
133
134 struct mptcp_options_received {
135 u64 sndr_key;
136 u64 rcvr_key;
137 u64 data_ack;
138 u64 data_seq;
139 u32 subflow_seq;
140 u16 data_len;
141 __sum16 csum;
142 u16 suboptions;
143 u32 token;
144 u32 nonce;
145 u16 use_map:1,
146 dsn64:1,
147 data_fin:1,
148 use_ack:1,
149 ack64:1,
150 mpc_map:1,
151 reset_reason:4,
152 reset_transient:1,
153 echo:1,
154 backup:1,
155 deny_join_id0:1,
156 __unused:2;
157 u8 join_id;
158 u64 thmac;
159 u8 hmac[MPTCPOPT_HMAC_LEN];
160 struct mptcp_addr_info addr;
161 struct mptcp_rm_list rm_list;
162 u64 ahmac;
163 u64 fail_seq;
164 };
165
mptcp_option(u8 subopt,u8 len,u8 nib,u8 field)166 static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field)
167 {
168 return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) |
169 ((nib & 0xF) << 8) | field);
170 }
171
172 enum mptcp_pm_status {
173 MPTCP_PM_ADD_ADDR_RECEIVED,
174 MPTCP_PM_ADD_ADDR_SEND_ACK,
175 MPTCP_PM_RM_ADDR_RECEIVED,
176 MPTCP_PM_ESTABLISHED,
177 MPTCP_PM_ALREADY_ESTABLISHED, /* persistent status, set after ESTABLISHED event */
178 MPTCP_PM_SUBFLOW_ESTABLISHED,
179 };
180
181 enum mptcp_addr_signal_status {
182 MPTCP_ADD_ADDR_SIGNAL,
183 MPTCP_ADD_ADDR_ECHO,
184 MPTCP_RM_ADDR_SIGNAL,
185 };
186
187 struct mptcp_pm_data {
188 struct mptcp_addr_info local;
189 struct mptcp_addr_info remote;
190 struct list_head anno_list;
191
192 spinlock_t lock; /*protects the whole PM data */
193
194 u8 addr_signal;
195 bool server_side;
196 bool work_pending;
197 bool accept_addr;
198 bool accept_subflow;
199 bool remote_deny_join_id0;
200 u8 add_addr_signaled;
201 u8 add_addr_accepted;
202 u8 local_addr_used;
203 u8 subflows;
204 u8 status;
205 struct mptcp_rm_list rm_list_tx;
206 struct mptcp_rm_list rm_list_rx;
207 };
208
209 struct mptcp_data_frag {
210 struct list_head list;
211 u64 data_seq;
212 u16 data_len;
213 u16 offset;
214 u16 overhead;
215 u16 already_sent;
216 struct page *page;
217 };
218
219 /* MPTCP connection sock */
220 struct mptcp_sock {
221 /* inet_connection_sock must be the first member */
222 struct inet_connection_sock sk;
223 u64 local_key;
224 u64 remote_key;
225 u64 write_seq;
226 u64 snd_nxt;
227 u64 ack_seq;
228 u64 rcv_wnd_sent;
229 u64 rcv_data_fin_seq;
230 int wmem_reserved;
231 struct sock *last_snd;
232 int snd_burst;
233 int old_wspace;
234 u64 recovery_snd_nxt; /* in recovery mode accept up to this seq;
235 * recovery related fields are under data_lock
236 * protection
237 */
238 u64 snd_una;
239 u64 wnd_end;
240 unsigned long timer_ival;
241 u32 token;
242 int rmem_released;
243 unsigned long flags;
244 bool recovery; /* closing subflow write queue reinjected */
245 bool can_ack;
246 bool fully_established;
247 bool rcv_data_fin;
248 bool snd_data_fin_enable;
249 bool rcv_fastclose;
250 bool use_64bit_ack; /* Set when we received a 64-bit DSN */
251 bool csum_enabled;
252 spinlock_t join_list_lock;
253 struct work_struct work;
254 struct sk_buff *ooo_last_skb;
255 struct rb_root out_of_order_queue;
256 struct sk_buff_head receive_queue;
257 int tx_pending_data;
258 struct list_head conn_list;
259 struct list_head rtx_queue;
260 struct mptcp_data_frag *first_pending;
261 struct list_head join_list;
262 struct socket *subflow; /* outgoing connect/listener/!mp_capable */
263 struct sock *first;
264 struct mptcp_pm_data pm;
265 struct {
266 u32 space; /* bytes copied in last measurement window */
267 u32 copied; /* bytes copied in this measurement window */
268 u64 time; /* start time of measurement window */
269 u64 rtt_us; /* last maximum rtt of subflows */
270 } rcvq_space;
271
272 u32 setsockopt_seq;
273 char ca_name[TCP_CA_NAME_MAX];
274 };
275
276 #define mptcp_lock_sock(___sk, cb) do { \
277 struct sock *__sk = (___sk); /* silence macro reuse warning */ \
278 might_sleep(); \
279 spin_lock_bh(&__sk->sk_lock.slock); \
280 if (__sk->sk_lock.owned) \
281 __lock_sock(__sk); \
282 cb; \
283 __sk->sk_lock.owned = 1; \
284 spin_unlock(&__sk->sk_lock.slock); \
285 mutex_acquire(&__sk->sk_lock.dep_map, 0, 0, _RET_IP_); \
286 local_bh_enable(); \
287 } while (0)
288
289 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock)
290 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock)
291
292 #define mptcp_for_each_subflow(__msk, __subflow) \
293 list_for_each_entry(__subflow, &((__msk)->conn_list), node)
294
msk_owned_by_me(const struct mptcp_sock * msk)295 static inline void msk_owned_by_me(const struct mptcp_sock *msk)
296 {
297 sock_owned_by_me((const struct sock *)msk);
298 }
299
mptcp_sk(const struct sock * sk)300 static inline struct mptcp_sock *mptcp_sk(const struct sock *sk)
301 {
302 return (struct mptcp_sock *)sk;
303 }
304
305 /* the msk socket don't use the backlog, also account for the bulk
306 * free memory
307 */
__mptcp_rmem(const struct sock * sk)308 static inline int __mptcp_rmem(const struct sock *sk)
309 {
310 return atomic_read(&sk->sk_rmem_alloc) - READ_ONCE(mptcp_sk(sk)->rmem_released);
311 }
312
__mptcp_space(const struct sock * sk)313 static inline int __mptcp_space(const struct sock *sk)
314 {
315 return tcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - __mptcp_rmem(sk));
316 }
317
mptcp_send_head(const struct sock * sk)318 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk)
319 {
320 const struct mptcp_sock *msk = mptcp_sk(sk);
321
322 return READ_ONCE(msk->first_pending);
323 }
324
mptcp_send_next(struct sock * sk)325 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk)
326 {
327 struct mptcp_sock *msk = mptcp_sk(sk);
328 struct mptcp_data_frag *cur;
329
330 cur = msk->first_pending;
331 return list_is_last(&cur->list, &msk->rtx_queue) ? NULL :
332 list_next_entry(cur, list);
333 }
334
mptcp_pending_tail(const struct sock * sk)335 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk)
336 {
337 struct mptcp_sock *msk = mptcp_sk(sk);
338
339 if (!msk->first_pending)
340 return NULL;
341
342 if (WARN_ON_ONCE(list_empty(&msk->rtx_queue)))
343 return NULL;
344
345 return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list);
346 }
347
mptcp_rtx_head(const struct sock * sk)348 static inline struct mptcp_data_frag *mptcp_rtx_head(const struct sock *sk)
349 {
350 struct mptcp_sock *msk = mptcp_sk(sk);
351
352 if (msk->snd_una == READ_ONCE(msk->snd_nxt))
353 return NULL;
354
355 return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list);
356 }
357
358 struct csum_pseudo_header {
359 __be64 data_seq;
360 __be32 subflow_seq;
361 __be16 data_len;
362 __sum16 csum;
363 };
364
365 struct mptcp_subflow_request_sock {
366 struct tcp_request_sock sk;
367 u16 mp_capable : 1,
368 mp_join : 1,
369 backup : 1,
370 csum_reqd : 1,
371 allow_join_id0 : 1;
372 u8 local_id;
373 u8 remote_id;
374 u64 local_key;
375 u64 idsn;
376 u32 token;
377 u32 ssn_offset;
378 u64 thmac;
379 u32 local_nonce;
380 u32 remote_nonce;
381 struct mptcp_sock *msk;
382 struct hlist_nulls_node token_node;
383 };
384
385 static inline struct mptcp_subflow_request_sock *
mptcp_subflow_rsk(const struct request_sock * rsk)386 mptcp_subflow_rsk(const struct request_sock *rsk)
387 {
388 return (struct mptcp_subflow_request_sock *)rsk;
389 }
390
391 enum mptcp_data_avail {
392 MPTCP_SUBFLOW_NODATA,
393 MPTCP_SUBFLOW_DATA_AVAIL,
394 };
395
396 struct mptcp_delegated_action {
397 struct napi_struct napi;
398 struct list_head head;
399 };
400
401 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions);
402
403 #define MPTCP_DELEGATE_SEND 0
404 #define MPTCP_DELEGATE_ACK 1
405
406 /* MPTCP subflow context */
407 struct mptcp_subflow_context {
408 struct list_head node;/* conn_list of subflows */
409 u64 local_key;
410 u64 remote_key;
411 u64 idsn;
412 u64 map_seq;
413 u32 snd_isn;
414 u32 token;
415 u32 rel_write_seq;
416 u32 map_subflow_seq;
417 u32 ssn_offset;
418 u32 map_data_len;
419 __wsum map_data_csum;
420 u32 map_csum_len;
421 u32 request_mptcp : 1, /* send MP_CAPABLE */
422 request_join : 1, /* send MP_JOIN */
423 request_bkup : 1,
424 mp_capable : 1, /* remote is MPTCP capable */
425 mp_join : 1, /* remote is JOINing */
426 fully_established : 1, /* path validated */
427 pm_notified : 1, /* PM hook called for established status */
428 conn_finished : 1,
429 map_valid : 1,
430 map_csum_reqd : 1,
431 map_data_fin : 1,
432 mpc_map : 1,
433 backup : 1,
434 send_mp_prio : 1,
435 send_mp_fail : 1,
436 rx_eof : 1,
437 can_ack : 1, /* only after processing the remote a key */
438 disposable : 1, /* ctx can be free at ulp release time */
439 stale : 1, /* unable to snd/rcv data, do not use for xmit */
440 valid_csum_seen : 1; /* at least one csum validated */
441 enum mptcp_data_avail data_avail;
442 u32 remote_nonce;
443 u64 thmac;
444 u32 local_nonce;
445 u32 remote_token;
446 u8 hmac[MPTCPOPT_HMAC_LEN];
447 u8 local_id;
448 u8 remote_id;
449 u8 reset_seen:1;
450 u8 reset_transient:1;
451 u8 reset_reason:4;
452 u8 stale_count;
453
454 long delegated_status;
455 struct list_head delegated_node; /* link into delegated_action, protected by local BH */
456
457 u32 setsockopt_seq;
458 u32 stale_rcv_tstamp;
459
460 struct sock *tcp_sock; /* tcp sk backpointer */
461 struct sock *conn; /* parent mptcp_sock */
462 const struct inet_connection_sock_af_ops *icsk_af_ops;
463 void (*tcp_data_ready)(struct sock *sk);
464 void (*tcp_state_change)(struct sock *sk);
465 void (*tcp_write_space)(struct sock *sk);
466 void (*tcp_error_report)(struct sock *sk);
467
468 struct rcu_head rcu;
469 };
470
471 static inline struct mptcp_subflow_context *
mptcp_subflow_ctx(const struct sock * sk)472 mptcp_subflow_ctx(const struct sock *sk)
473 {
474 struct inet_connection_sock *icsk = inet_csk(sk);
475
476 /* Use RCU on icsk_ulp_data only for sock diag code */
477 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data;
478 }
479
480 static inline struct sock *
mptcp_subflow_tcp_sock(const struct mptcp_subflow_context * subflow)481 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow)
482 {
483 return subflow->tcp_sock;
484 }
485
486 static inline u64
mptcp_subflow_get_map_offset(const struct mptcp_subflow_context * subflow)487 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow)
488 {
489 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq -
490 subflow->ssn_offset -
491 subflow->map_subflow_seq;
492 }
493
494 static inline u64
mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context * subflow)495 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow)
496 {
497 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow);
498 }
499
mptcp_add_pending_subflow(struct mptcp_sock * msk,struct mptcp_subflow_context * subflow)500 static inline void mptcp_add_pending_subflow(struct mptcp_sock *msk,
501 struct mptcp_subflow_context *subflow)
502 {
503 sock_hold(mptcp_subflow_tcp_sock(subflow));
504 spin_lock_bh(&msk->join_list_lock);
505 list_add_tail(&subflow->node, &msk->join_list);
506 spin_unlock_bh(&msk->join_list_lock);
507 }
508
509 void mptcp_subflow_process_delegated(struct sock *ssk);
510
mptcp_subflow_delegate(struct mptcp_subflow_context * subflow,int action)511 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action)
512 {
513 struct mptcp_delegated_action *delegated;
514 bool schedule;
515
516 /* the caller held the subflow bh socket lock */
517 lockdep_assert_in_softirq();
518
519 /* The implied barrier pairs with mptcp_subflow_delegated_done(), and
520 * ensures the below list check sees list updates done prior to status
521 * bit changes
522 */
523 if (!test_and_set_bit(action, &subflow->delegated_status)) {
524 /* still on delegated list from previous scheduling */
525 if (!list_empty(&subflow->delegated_node))
526 return;
527
528 delegated = this_cpu_ptr(&mptcp_delegated_actions);
529 schedule = list_empty(&delegated->head);
530 list_add_tail(&subflow->delegated_node, &delegated->head);
531 sock_hold(mptcp_subflow_tcp_sock(subflow));
532 if (schedule)
533 napi_schedule(&delegated->napi);
534 }
535 }
536
537 static inline struct mptcp_subflow_context *
mptcp_subflow_delegated_next(struct mptcp_delegated_action * delegated)538 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated)
539 {
540 struct mptcp_subflow_context *ret;
541
542 if (list_empty(&delegated->head))
543 return NULL;
544
545 ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node);
546 list_del_init(&ret->delegated_node);
547 return ret;
548 }
549
mptcp_subflow_has_delegated_action(const struct mptcp_subflow_context * subflow)550 static inline bool mptcp_subflow_has_delegated_action(const struct mptcp_subflow_context *subflow)
551 {
552 return !!READ_ONCE(subflow->delegated_status);
553 }
554
mptcp_subflow_delegated_done(struct mptcp_subflow_context * subflow,int action)555 static inline void mptcp_subflow_delegated_done(struct mptcp_subflow_context *subflow, int action)
556 {
557 /* pairs with mptcp_subflow_delegate, ensures delegate_node is updated before
558 * touching the status bit
559 */
560 smp_wmb();
561 clear_bit(action, &subflow->delegated_status);
562 }
563
564 int mptcp_is_enabled(const struct net *net);
565 unsigned int mptcp_get_add_addr_timeout(const struct net *net);
566 int mptcp_is_checksum_enabled(const struct net *net);
567 int mptcp_allow_join_id0(const struct net *net);
568 unsigned int mptcp_stale_loss_cnt(const struct net *net);
569 void mptcp_subflow_fully_established(struct mptcp_subflow_context *subflow,
570 struct mptcp_options_received *mp_opt);
571 bool __mptcp_retransmit_pending_data(struct sock *sk);
572 void __mptcp_push_pending(struct sock *sk, unsigned int flags);
573 bool mptcp_subflow_data_available(struct sock *sk);
574 void __init mptcp_subflow_init(void);
575 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how);
576 void mptcp_close_ssk(struct sock *sk, struct sock *ssk,
577 struct mptcp_subflow_context *subflow);
578 void mptcp_subflow_send_ack(struct sock *ssk);
579 void mptcp_subflow_reset(struct sock *ssk);
580 void mptcp_sock_graft(struct sock *sk, struct socket *parent);
581 struct socket *__mptcp_nmpc_socket(const struct mptcp_sock *msk);
582
583 /* called with sk socket lock held */
584 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc,
585 const struct mptcp_addr_info *remote);
586 int mptcp_subflow_create_socket(struct sock *sk, struct socket **new_sock);
587 void mptcp_info2sockaddr(const struct mptcp_addr_info *info,
588 struct sockaddr_storage *addr,
589 unsigned short family);
590
__mptcp_subflow_active(struct mptcp_subflow_context * subflow)591 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow)
592 {
593 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
594
595 /* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */
596 if (subflow->request_join && !subflow->fully_established)
597 return false;
598
599 /* only send if our side has not closed yet */
600 return ((1 << ssk->sk_state) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT));
601 }
602
603 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow);
604
605 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow);
606
mptcp_subflow_tcp_fallback(struct sock * sk,struct mptcp_subflow_context * ctx)607 static inline void mptcp_subflow_tcp_fallback(struct sock *sk,
608 struct mptcp_subflow_context *ctx)
609 {
610 sk->sk_data_ready = ctx->tcp_data_ready;
611 sk->sk_state_change = ctx->tcp_state_change;
612 sk->sk_write_space = ctx->tcp_write_space;
613 sk->sk_error_report = ctx->tcp_error_report;
614
615 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops;
616 }
617
mptcp_has_another_subflow(struct sock * ssk)618 static inline bool mptcp_has_another_subflow(struct sock *ssk)
619 {
620 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk), *tmp;
621 struct mptcp_sock *msk = mptcp_sk(subflow->conn);
622
623 mptcp_for_each_subflow(msk, tmp) {
624 if (tmp != subflow)
625 return true;
626 }
627
628 return false;
629 }
630
631 void __init mptcp_proto_init(void);
632 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
633 int __init mptcp_proto_v6_init(void);
634 #endif
635
636 struct sock *mptcp_sk_clone(const struct sock *sk,
637 const struct mptcp_options_received *mp_opt,
638 struct request_sock *req);
639 void mptcp_get_options(const struct sk_buff *skb,
640 struct mptcp_options_received *mp_opt);
641
642 void mptcp_finish_connect(struct sock *sk);
643 void __mptcp_set_connected(struct sock *sk);
mptcp_is_fully_established(struct sock * sk)644 static inline bool mptcp_is_fully_established(struct sock *sk)
645 {
646 return inet_sk_state_load(sk) == TCP_ESTABLISHED &&
647 READ_ONCE(mptcp_sk(sk)->fully_established);
648 }
649 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk);
650 void mptcp_data_ready(struct sock *sk, struct sock *ssk);
651 bool mptcp_finish_join(struct sock *sk);
652 bool mptcp_schedule_work(struct sock *sk);
653 int mptcp_setsockopt(struct sock *sk, int level, int optname,
654 sockptr_t optval, unsigned int optlen);
655 int mptcp_getsockopt(struct sock *sk, int level, int optname,
656 char __user *optval, int __user *option);
657
658 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq);
mptcp_expand_seq(u64 old_seq,u64 cur_seq,bool use_64bit)659 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit)
660 {
661 if (use_64bit)
662 return cur_seq;
663
664 return __mptcp_expand_seq(old_seq, cur_seq);
665 }
666 void __mptcp_check_push(struct sock *sk, struct sock *ssk);
667 void __mptcp_data_acked(struct sock *sk);
668 void __mptcp_error_report(struct sock *sk);
669 void mptcp_subflow_eof(struct sock *sk);
670 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit);
671 void __mptcp_flush_join_list(struct mptcp_sock *msk);
mptcp_data_fin_enabled(const struct mptcp_sock * msk)672 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk)
673 {
674 return READ_ONCE(msk->snd_data_fin_enable) &&
675 READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt);
676 }
677
mptcp_propagate_sndbuf(struct sock * sk,struct sock * ssk)678 static inline bool mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
679 {
680 if ((sk->sk_userlocks & SOCK_SNDBUF_LOCK) || ssk->sk_sndbuf <= READ_ONCE(sk->sk_sndbuf))
681 return false;
682
683 WRITE_ONCE(sk->sk_sndbuf, ssk->sk_sndbuf);
684 return true;
685 }
686
mptcp_write_space(struct sock * sk)687 static inline void mptcp_write_space(struct sock *sk)
688 {
689 if (sk_stream_is_writeable(sk)) {
690 /* pairs with memory barrier in mptcp_poll */
691 smp_mb();
692 if (test_and_clear_bit(MPTCP_NOSPACE, &mptcp_sk(sk)->flags))
693 sk_stream_write_space(sk);
694 }
695 }
696
697 void mptcp_destroy_common(struct mptcp_sock *msk);
698
699 #define MPTCP_TOKEN_MAX_RETRIES 4
700
701 void __init mptcp_token_init(void);
mptcp_token_init_request(struct request_sock * req)702 static inline void mptcp_token_init_request(struct request_sock *req)
703 {
704 mptcp_subflow_rsk(req)->token_node.pprev = NULL;
705 }
706
707 int mptcp_token_new_request(struct request_sock *req);
708 void mptcp_token_destroy_request(struct request_sock *req);
709 int mptcp_token_new_connect(struct sock *sk);
710 void mptcp_token_accept(struct mptcp_subflow_request_sock *r,
711 struct mptcp_sock *msk);
712 bool mptcp_token_exists(u32 token);
713 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token);
714 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot,
715 long *s_num);
716 void mptcp_token_destroy(struct mptcp_sock *msk);
717
718 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn);
719
720 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac);
721 __sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum);
722
723 void __init mptcp_pm_init(void);
724 void mptcp_pm_data_init(struct mptcp_sock *msk);
725 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
726 void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
727 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side);
728 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk, gfp_t gfp);
729 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk);
730 void mptcp_pm_connection_closed(struct mptcp_sock *msk);
731 void mptcp_pm_subflow_established(struct mptcp_sock *msk);
732 void mptcp_pm_subflow_closed(struct mptcp_sock *msk, u8 id);
733 void mptcp_pm_add_addr_received(struct mptcp_sock *msk,
734 const struct mptcp_addr_info *addr);
735 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk,
736 struct mptcp_addr_info *addr);
737 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk);
738 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk);
739 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk,
740 const struct mptcp_rm_list *rm_list);
741 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup);
742 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk,
743 struct mptcp_addr_info *addr,
744 u8 bkup);
745 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq);
746 void mptcp_pm_free_anno_list(struct mptcp_sock *msk);
747 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk);
748 struct mptcp_pm_add_entry *
749 mptcp_pm_del_add_timer(struct mptcp_sock *msk,
750 struct mptcp_addr_info *addr, bool check_id);
751 struct mptcp_pm_add_entry *
752 mptcp_lookup_anno_list_by_saddr(struct mptcp_sock *msk,
753 struct mptcp_addr_info *addr);
754 int mptcp_pm_get_flags_and_ifindex_by_id(struct net *net, unsigned int id,
755 u8 *flags, int *ifindex);
756
757 int mptcp_pm_announce_addr(struct mptcp_sock *msk,
758 const struct mptcp_addr_info *addr,
759 bool echo);
760 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
761 int mptcp_pm_remove_subflow(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
762
763 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk,
764 const struct sock *ssk, gfp_t gfp);
765 void mptcp_event_addr_announced(const struct mptcp_sock *msk, const struct mptcp_addr_info *info);
766 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id);
767
mptcp_pm_should_add_signal(struct mptcp_sock * msk)768 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk)
769 {
770 return READ_ONCE(msk->pm.addr_signal) &
771 (BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO));
772 }
773
mptcp_pm_should_add_signal_addr(struct mptcp_sock * msk)774 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk)
775 {
776 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL);
777 }
778
mptcp_pm_should_add_signal_echo(struct mptcp_sock * msk)779 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk)
780 {
781 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO);
782 }
783
mptcp_pm_should_rm_signal(struct mptcp_sock * msk)784 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk)
785 {
786 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL);
787 }
788
mptcp_add_addr_len(int family,bool echo,bool port)789 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port)
790 {
791 u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE;
792
793 if (family == AF_INET6)
794 len = TCPOLEN_MPTCP_ADD_ADDR6_BASE;
795 if (!echo)
796 len += MPTCPOPT_THMAC_LEN;
797 /* account for 2 trailing 'nop' options */
798 if (port)
799 len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN;
800
801 return len;
802 }
803
mptcp_rm_addr_len(const struct mptcp_rm_list * rm_list)804 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list)
805 {
806 if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX)
807 return -EINVAL;
808
809 return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1;
810 }
811
812 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, struct sk_buff *skb,
813 unsigned int opt_size, unsigned int remaining,
814 struct mptcp_addr_info *addr, bool *echo,
815 bool *port, bool *drop_other_suboptions);
816 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining,
817 struct mptcp_rm_list *rm_list);
818 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
819
820 void __init mptcp_pm_nl_init(void);
821 void mptcp_pm_nl_data_init(struct mptcp_sock *msk);
822 void mptcp_pm_nl_work(struct mptcp_sock *msk);
823 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk,
824 const struct mptcp_rm_list *rm_list);
825 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
826 unsigned int mptcp_pm_get_add_addr_signal_max(struct mptcp_sock *msk);
827 unsigned int mptcp_pm_get_add_addr_accept_max(struct mptcp_sock *msk);
828 unsigned int mptcp_pm_get_subflows_max(struct mptcp_sock *msk);
829 unsigned int mptcp_pm_get_local_addr_max(struct mptcp_sock *msk);
830
831 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk);
832 void mptcp_sockopt_sync_all(struct mptcp_sock *msk);
833
mptcp_get_ext(const struct sk_buff * skb)834 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb)
835 {
836 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP);
837 }
838
839 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops);
840
__mptcp_check_fallback(const struct mptcp_sock * msk)841 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk)
842 {
843 return test_bit(MPTCP_FALLBACK_DONE, &msk->flags);
844 }
845
mptcp_check_fallback(const struct sock * sk)846 static inline bool mptcp_check_fallback(const struct sock *sk)
847 {
848 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
849 struct mptcp_sock *msk = mptcp_sk(subflow->conn);
850
851 return __mptcp_check_fallback(msk);
852 }
853
__mptcp_do_fallback(struct mptcp_sock * msk)854 static inline void __mptcp_do_fallback(struct mptcp_sock *msk)
855 {
856 if (test_bit(MPTCP_FALLBACK_DONE, &msk->flags)) {
857 pr_debug("TCP fallback already done (msk=%p)", msk);
858 return;
859 }
860 set_bit(MPTCP_FALLBACK_DONE, &msk->flags);
861 }
862
mptcp_do_fallback(struct sock * sk)863 static inline void mptcp_do_fallback(struct sock *sk)
864 {
865 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
866 struct mptcp_sock *msk = mptcp_sk(subflow->conn);
867
868 __mptcp_do_fallback(msk);
869 }
870
871 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a)
872
subflow_simultaneous_connect(struct sock * sk)873 static inline bool subflow_simultaneous_connect(struct sock *sk)
874 {
875 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
876 struct sock *parent = subflow->conn;
877
878 return sk->sk_state == TCP_ESTABLISHED &&
879 !mptcp_sk(parent)->pm.server_side &&
880 !subflow->conn_finished;
881 }
882
883 #ifdef CONFIG_SYN_COOKIES
884 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
885 struct sk_buff *skb);
886 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
887 struct sk_buff *skb);
888 void __init mptcp_join_cookie_init(void);
889 #else
890 static inline void
subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock * subflow_req,struct sk_buff * skb)891 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
892 struct sk_buff *skb) {}
893 static inline bool
mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock * subflow_req,struct sk_buff * skb)894 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
895 struct sk_buff *skb)
896 {
897 return false;
898 }
899
mptcp_join_cookie_init(void)900 static inline void mptcp_join_cookie_init(void) {}
901 #endif
902
903 #endif /* __MPTCP_PROTOCOL_H */
904