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