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 bool allow_subflows;
334
335 u32 subflow_id;
336 u32 setsockopt_seq;
337 char ca_name[TCP_CA_NAME_MAX];
338
339 spinlock_t fallback_lock; /* protects fallback,
340 * allow_infinite_fallback and
341 * allow_join
342 */
343 };
344
345 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock)
346 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock)
347
348 #define mptcp_for_each_subflow(__msk, __subflow) \
349 list_for_each_entry(__subflow, &((__msk)->conn_list), node)
350 #define mptcp_for_each_subflow_safe(__msk, __subflow, __tmp) \
351 list_for_each_entry_safe(__subflow, __tmp, &((__msk)->conn_list), node)
352
msk_owned_by_me(const struct mptcp_sock * msk)353 static inline void msk_owned_by_me(const struct mptcp_sock *msk)
354 {
355 sock_owned_by_me((const struct sock *)msk);
356 }
357
358 #define mptcp_sk(ptr) container_of_const(ptr, struct mptcp_sock, sk.icsk_inet.sk)
359
360 /* the msk socket don't use the backlog, also account for the bulk
361 * free memory
362 */
__mptcp_rmem(const struct sock * sk)363 static inline int __mptcp_rmem(const struct sock *sk)
364 {
365 return atomic_read(&sk->sk_rmem_alloc) - READ_ONCE(mptcp_sk(sk)->rmem_released);
366 }
367
mptcp_win_from_space(const struct sock * sk,int space)368 static inline int mptcp_win_from_space(const struct sock *sk, int space)
369 {
370 return __tcp_win_from_space(mptcp_sk(sk)->scaling_ratio, space);
371 }
372
__mptcp_space(const struct sock * sk)373 static inline int __mptcp_space(const struct sock *sk)
374 {
375 return mptcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - __mptcp_rmem(sk));
376 }
377
mptcp_send_head(const struct sock * sk)378 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk)
379 {
380 const struct mptcp_sock *msk = mptcp_sk(sk);
381
382 return READ_ONCE(msk->first_pending);
383 }
384
mptcp_send_next(struct sock * sk)385 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk)
386 {
387 struct mptcp_sock *msk = mptcp_sk(sk);
388 struct mptcp_data_frag *cur;
389
390 cur = msk->first_pending;
391 return list_is_last(&cur->list, &msk->rtx_queue) ? NULL :
392 list_next_entry(cur, list);
393 }
394
mptcp_pending_tail(const struct sock * sk)395 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk)
396 {
397 const struct mptcp_sock *msk = mptcp_sk(sk);
398
399 if (!msk->first_pending)
400 return NULL;
401
402 if (WARN_ON_ONCE(list_empty(&msk->rtx_queue)))
403 return NULL;
404
405 return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list);
406 }
407
mptcp_rtx_head(struct sock * sk)408 static inline struct mptcp_data_frag *mptcp_rtx_head(struct sock *sk)
409 {
410 struct mptcp_sock *msk = mptcp_sk(sk);
411
412 if (msk->snd_una == READ_ONCE(msk->snd_nxt))
413 return NULL;
414
415 return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list);
416 }
417
418 struct csum_pseudo_header {
419 __be64 data_seq;
420 __be32 subflow_seq;
421 __be16 data_len;
422 __sum16 csum;
423 };
424
425 struct mptcp_subflow_request_sock {
426 struct tcp_request_sock sk;
427 u16 mp_capable : 1,
428 mp_join : 1,
429 backup : 1,
430 request_bkup : 1,
431 csum_reqd : 1,
432 allow_join_id0 : 1;
433 u8 local_id;
434 u8 remote_id;
435 u64 local_key;
436 u64 idsn;
437 u32 token;
438 u32 ssn_offset;
439 u64 thmac;
440 u32 local_nonce;
441 u32 remote_nonce;
442 struct mptcp_sock *msk;
443 struct hlist_nulls_node token_node;
444 };
445
446 static inline struct mptcp_subflow_request_sock *
mptcp_subflow_rsk(const struct request_sock * rsk)447 mptcp_subflow_rsk(const struct request_sock *rsk)
448 {
449 return (struct mptcp_subflow_request_sock *)rsk;
450 }
451
452 enum mptcp_data_avail {
453 MPTCP_SUBFLOW_NODATA,
454 MPTCP_SUBFLOW_DATA_AVAIL,
455 };
456
457 struct mptcp_delegated_action {
458 struct napi_struct napi;
459 struct list_head head;
460 };
461
462 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions);
463
464 #define MPTCP_DELEGATE_SCHEDULED 0
465 #define MPTCP_DELEGATE_SEND 1
466 #define MPTCP_DELEGATE_ACK 2
467 #define MPTCP_DELEGATE_SNDBUF 3
468
469 #define MPTCP_DELEGATE_ACTIONS_MASK (~BIT(MPTCP_DELEGATE_SCHEDULED))
470 /* MPTCP subflow context */
471 struct mptcp_subflow_context {
472 struct list_head node;/* conn_list of subflows */
473
474 struct_group(reset,
475
476 unsigned long avg_pacing_rate; /* protected by msk socket lock */
477 u64 local_key;
478 u64 remote_key;
479 u64 idsn;
480 u64 map_seq;
481 u32 snd_isn;
482 u32 token;
483 u32 rel_write_seq;
484 u32 map_subflow_seq;
485 u32 ssn_offset;
486 u32 map_data_len;
487 __wsum map_data_csum;
488 u32 map_csum_len;
489 u32 request_mptcp : 1, /* send MP_CAPABLE */
490 request_join : 1, /* send MP_JOIN */
491 request_bkup : 1,
492 mp_capable : 1, /* remote is MPTCP capable */
493 mp_join : 1, /* remote is JOINing */
494 fully_established : 1, /* path validated */
495 pm_notified : 1, /* PM hook called for established status */
496 conn_finished : 1,
497 map_valid : 1,
498 map_csum_reqd : 1,
499 map_data_fin : 1,
500 mpc_map : 1,
501 backup : 1,
502 send_mp_prio : 1,
503 send_mp_fail : 1,
504 send_fastclose : 1,
505 send_infinite_map : 1,
506 remote_key_valid : 1, /* received the peer key from */
507 disposable : 1, /* ctx can be free at ulp release time */
508 stale : 1, /* unable to snd/rcv data, do not use for xmit */
509 valid_csum_seen : 1, /* at least one csum validated */
510 is_mptfo : 1, /* subflow is doing TFO */
511 close_event_done : 1, /* has done the post-closed part */
512 __unused : 9;
513 enum mptcp_data_avail data_avail;
514 bool scheduled;
515 bool pm_listener; /* a listener managed by the kernel PM? */
516 u32 remote_nonce;
517 u64 thmac;
518 u32 local_nonce;
519 u32 remote_token;
520 union {
521 u8 hmac[MPTCPOPT_HMAC_LEN]; /* MPJ subflow only */
522 u64 iasn; /* initial ack sequence number, MPC subflows only */
523 };
524 s16 local_id; /* if negative not initialized yet */
525 u8 remote_id;
526 u8 reset_seen:1;
527 u8 reset_transient:1;
528 u8 reset_reason:4;
529 u8 stale_count;
530
531 u32 subflow_id;
532
533 long delegated_status;
534 unsigned long fail_tout;
535
536 );
537
538 struct list_head delegated_node; /* link into delegated_action, protected by local BH */
539
540 u32 setsockopt_seq;
541 u32 stale_rcv_tstamp;
542 int cached_sndbuf; /* sndbuf size when last synced with the msk sndbuf,
543 * protected by the msk socket lock
544 */
545
546 struct sock *tcp_sock; /* tcp sk backpointer */
547 struct sock *conn; /* parent mptcp_sock */
548 const struct inet_connection_sock_af_ops *icsk_af_ops;
549 void (*tcp_state_change)(struct sock *sk);
550 void (*tcp_error_report)(struct sock *sk);
551
552 struct rcu_head rcu;
553 };
554
555 static inline struct mptcp_subflow_context *
mptcp_subflow_ctx(const struct sock * sk)556 mptcp_subflow_ctx(const struct sock *sk)
557 {
558 struct inet_connection_sock *icsk = inet_csk(sk);
559
560 /* Use RCU on icsk_ulp_data only for sock diag code */
561 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data;
562 }
563
564 static inline struct sock *
mptcp_subflow_tcp_sock(const struct mptcp_subflow_context * subflow)565 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow)
566 {
567 return subflow->tcp_sock;
568 }
569
570 static inline void
mptcp_subflow_ctx_reset(struct mptcp_subflow_context * subflow)571 mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow)
572 {
573 memset(&subflow->reset, 0, sizeof(subflow->reset));
574 subflow->request_mptcp = 1;
575 WRITE_ONCE(subflow->local_id, -1);
576 }
577
578 static inline u64
mptcp_subflow_get_map_offset(const struct mptcp_subflow_context * subflow)579 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow)
580 {
581 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq -
582 subflow->ssn_offset -
583 subflow->map_subflow_seq;
584 }
585
586 static inline u64
mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context * subflow)587 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow)
588 {
589 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow);
590 }
591
592 void mptcp_subflow_process_delegated(struct sock *ssk, long actions);
593
mptcp_subflow_delegate(struct mptcp_subflow_context * subflow,int action)594 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action)
595 {
596 long old, set_bits = BIT(MPTCP_DELEGATE_SCHEDULED) | BIT(action);
597 struct mptcp_delegated_action *delegated;
598 bool schedule;
599
600 /* the caller held the subflow bh socket lock */
601 lockdep_assert_in_softirq();
602
603 /* The implied barrier pairs with tcp_release_cb_override()
604 * mptcp_napi_poll(), and ensures the below list check sees list
605 * updates done prior to delegated status bits changes
606 */
607 old = set_mask_bits(&subflow->delegated_status, 0, set_bits);
608 if (!(old & BIT(MPTCP_DELEGATE_SCHEDULED))) {
609 if (WARN_ON_ONCE(!list_empty(&subflow->delegated_node)))
610 return;
611
612 delegated = this_cpu_ptr(&mptcp_delegated_actions);
613 schedule = list_empty(&delegated->head);
614 list_add_tail(&subflow->delegated_node, &delegated->head);
615 sock_hold(mptcp_subflow_tcp_sock(subflow));
616 if (schedule)
617 napi_schedule(&delegated->napi);
618 }
619 }
620
621 static inline struct mptcp_subflow_context *
mptcp_subflow_delegated_next(struct mptcp_delegated_action * delegated)622 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated)
623 {
624 struct mptcp_subflow_context *ret;
625
626 if (list_empty(&delegated->head))
627 return NULL;
628
629 ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node);
630 list_del_init(&ret->delegated_node);
631 return ret;
632 }
633
634 int mptcp_is_enabled(const struct net *net);
635 unsigned int mptcp_get_add_addr_timeout(const struct net *net);
636 int mptcp_is_checksum_enabled(const struct net *net);
637 int mptcp_allow_join_id0(const struct net *net);
638 unsigned int mptcp_stale_loss_cnt(const struct net *net);
639 int mptcp_get_pm_type(const struct net *net);
640 const char *mptcp_get_scheduler(const struct net *net);
641 void __mptcp_subflow_fully_established(struct mptcp_sock *msk,
642 struct mptcp_subflow_context *subflow,
643 const struct mptcp_options_received *mp_opt);
644 bool __mptcp_retransmit_pending_data(struct sock *sk);
645 void mptcp_check_and_set_pending(struct sock *sk);
646 void __mptcp_push_pending(struct sock *sk, unsigned int flags);
647 bool mptcp_subflow_data_available(struct sock *sk);
648 void __init mptcp_subflow_init(void);
649 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how);
650 void mptcp_close_ssk(struct sock *sk, struct sock *ssk,
651 struct mptcp_subflow_context *subflow);
652 void __mptcp_subflow_send_ack(struct sock *ssk);
653 void mptcp_subflow_reset(struct sock *ssk);
654 void mptcp_subflow_queue_clean(struct sock *sk, struct sock *ssk);
655 void mptcp_sock_graft(struct sock *sk, struct socket *parent);
656 struct sock *__mptcp_nmpc_sk(struct mptcp_sock *msk);
657 bool __mptcp_close(struct sock *sk, long timeout);
658 void mptcp_cancel_work(struct sock *sk);
659 void __mptcp_unaccepted_force_close(struct sock *sk);
660 void mptcp_set_owner_r(struct sk_buff *skb, struct sock *sk);
661 void mptcp_set_state(struct sock *sk, int state);
662
663 bool mptcp_addresses_equal(const struct mptcp_addr_info *a,
664 const struct mptcp_addr_info *b, bool use_port);
665 void mptcp_local_address(const struct sock_common *skc, struct mptcp_addr_info *addr);
666
667 /* called with sk socket lock held */
668 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc,
669 const struct mptcp_addr_info *remote);
670 int mptcp_subflow_create_socket(struct sock *sk, unsigned short family,
671 struct socket **new_sock);
672 void mptcp_info2sockaddr(const struct mptcp_addr_info *info,
673 struct sockaddr_storage *addr,
674 unsigned short family);
675 struct mptcp_sched_ops *mptcp_sched_find(const char *name);
676 int mptcp_register_scheduler(struct mptcp_sched_ops *sched);
677 void mptcp_unregister_scheduler(struct mptcp_sched_ops *sched);
678 void mptcp_sched_init(void);
679 int mptcp_init_sched(struct mptcp_sock *msk,
680 struct mptcp_sched_ops *sched);
681 void mptcp_release_sched(struct mptcp_sock *msk);
682 void mptcp_subflow_set_scheduled(struct mptcp_subflow_context *subflow,
683 bool scheduled);
684 struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk);
685 struct sock *mptcp_subflow_get_retrans(struct mptcp_sock *msk);
686 int mptcp_sched_get_send(struct mptcp_sock *msk);
687 int mptcp_sched_get_retrans(struct mptcp_sock *msk);
688
mptcp_data_avail(const struct mptcp_sock * msk)689 static inline u64 mptcp_data_avail(const struct mptcp_sock *msk)
690 {
691 return READ_ONCE(msk->bytes_received) - READ_ONCE(msk->bytes_consumed);
692 }
693
mptcp_epollin_ready(const struct sock * sk)694 static inline bool mptcp_epollin_ready(const struct sock *sk)
695 {
696 u64 data_avail = mptcp_data_avail(mptcp_sk(sk));
697
698 if (!data_avail)
699 return false;
700
701 /* mptcp doesn't have to deal with small skbs in the receive queue,
702 * as it can always coalesce them
703 */
704 return (data_avail >= sk->sk_rcvlowat) ||
705 (mem_cgroup_sockets_enabled && sk->sk_memcg &&
706 mem_cgroup_under_socket_pressure(sk->sk_memcg)) ||
707 READ_ONCE(tcp_memory_pressure);
708 }
709
710 int mptcp_set_rcvlowat(struct sock *sk, int val);
711
__tcp_can_send(const struct sock * ssk)712 static inline bool __tcp_can_send(const struct sock *ssk)
713 {
714 /* only send if our side has not closed yet */
715 return ((1 << inet_sk_state_load(ssk)) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT));
716 }
717
__mptcp_subflow_active(struct mptcp_subflow_context * subflow)718 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow)
719 {
720 /* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */
721 if (subflow->request_join && !subflow->fully_established)
722 return false;
723
724 return __tcp_can_send(mptcp_subflow_tcp_sock(subflow));
725 }
726
727 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow);
728
729 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow);
730
731 void mptcp_subflow_drop_ctx(struct sock *ssk);
732
mptcp_subflow_tcp_fallback(struct sock * sk,struct mptcp_subflow_context * ctx)733 static inline void mptcp_subflow_tcp_fallback(struct sock *sk,
734 struct mptcp_subflow_context *ctx)
735 {
736 sk->sk_data_ready = sock_def_readable;
737 sk->sk_state_change = ctx->tcp_state_change;
738 sk->sk_write_space = sk_stream_write_space;
739 sk->sk_error_report = ctx->tcp_error_report;
740
741 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops;
742 }
743
744 void __init mptcp_proto_init(void);
745 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
746 int __init mptcp_proto_v6_init(void);
747 #endif
748
749 struct sock *mptcp_sk_clone_init(const struct sock *sk,
750 const struct mptcp_options_received *mp_opt,
751 struct sock *ssk,
752 struct request_sock *req);
753 void mptcp_get_options(const struct sk_buff *skb,
754 struct mptcp_options_received *mp_opt);
755
756 void mptcp_finish_connect(struct sock *sk);
757 void __mptcp_sync_state(struct sock *sk, int state);
758 void mptcp_reset_tout_timer(struct mptcp_sock *msk, unsigned long fail_tout);
759
mptcp_stop_tout_timer(struct sock * sk)760 static inline void mptcp_stop_tout_timer(struct sock *sk)
761 {
762 if (!inet_csk(sk)->icsk_mtup.probe_timestamp)
763 return;
764
765 sk_stop_timer(sk, &sk->sk_timer);
766 inet_csk(sk)->icsk_mtup.probe_timestamp = 0;
767 }
768
mptcp_set_close_tout(struct sock * sk,unsigned long tout)769 static inline void mptcp_set_close_tout(struct sock *sk, unsigned long tout)
770 {
771 /* avoid 0 timestamp, as that means no close timeout */
772 inet_csk(sk)->icsk_mtup.probe_timestamp = tout ? : 1;
773 }
774
mptcp_start_tout_timer(struct sock * sk)775 static inline void mptcp_start_tout_timer(struct sock *sk)
776 {
777 mptcp_set_close_tout(sk, tcp_jiffies32);
778 mptcp_reset_tout_timer(mptcp_sk(sk), 0);
779 }
780
mptcp_is_fully_established(struct sock * sk)781 static inline bool mptcp_is_fully_established(struct sock *sk)
782 {
783 return inet_sk_state_load(sk) == TCP_ESTABLISHED &&
784 READ_ONCE(mptcp_sk(sk)->fully_established);
785 }
786
787 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk);
788 void mptcp_data_ready(struct sock *sk, struct sock *ssk);
789 bool mptcp_finish_join(struct sock *sk);
790 bool mptcp_schedule_work(struct sock *sk);
791 int mptcp_setsockopt(struct sock *sk, int level, int optname,
792 sockptr_t optval, unsigned int optlen);
793 int mptcp_getsockopt(struct sock *sk, int level, int optname,
794 char __user *optval, int __user *option);
795
796 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq);
mptcp_expand_seq(u64 old_seq,u64 cur_seq,bool use_64bit)797 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit)
798 {
799 if (use_64bit)
800 return cur_seq;
801
802 return __mptcp_expand_seq(old_seq, cur_seq);
803 }
804 void __mptcp_check_push(struct sock *sk, struct sock *ssk);
805 void __mptcp_data_acked(struct sock *sk);
806 void __mptcp_error_report(struct sock *sk);
807 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)808 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk)
809 {
810 return READ_ONCE(msk->snd_data_fin_enable) &&
811 READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt);
812 }
813
mptcp_write_space(struct sock * sk)814 static inline void mptcp_write_space(struct sock *sk)
815 {
816 if (sk_stream_is_writeable(sk)) {
817 /* pairs with memory barrier in mptcp_poll */
818 smp_mb();
819 if (test_and_clear_bit(MPTCP_NOSPACE, &mptcp_sk(sk)->flags))
820 sk_stream_write_space(sk);
821 }
822 }
823
__mptcp_sync_sndbuf(struct sock * sk)824 static inline void __mptcp_sync_sndbuf(struct sock *sk)
825 {
826 struct mptcp_subflow_context *subflow;
827 int ssk_sndbuf, new_sndbuf;
828
829 if (sk->sk_userlocks & SOCK_SNDBUF_LOCK)
830 return;
831
832 new_sndbuf = sock_net(sk)->ipv4.sysctl_tcp_wmem[0];
833 mptcp_for_each_subflow(mptcp_sk(sk), subflow) {
834 ssk_sndbuf = READ_ONCE(mptcp_subflow_tcp_sock(subflow)->sk_sndbuf);
835
836 subflow->cached_sndbuf = ssk_sndbuf;
837 new_sndbuf += ssk_sndbuf;
838 }
839
840 /* the msk max wmem limit is <nr_subflows> * tcp wmem[2] */
841 WRITE_ONCE(sk->sk_sndbuf, new_sndbuf);
842 mptcp_write_space(sk);
843 }
844
845 /* The called held both the msk socket and the subflow socket locks,
846 * possibly under BH
847 */
__mptcp_propagate_sndbuf(struct sock * sk,struct sock * ssk)848 static inline void __mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
849 {
850 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
851
852 if (READ_ONCE(ssk->sk_sndbuf) != subflow->cached_sndbuf)
853 __mptcp_sync_sndbuf(sk);
854 }
855
856 /* the caller held only the subflow socket lock, either in process or
857 * BH context. Additionally this can be called under the msk data lock,
858 * so we can't acquire such lock here: let the delegate action acquires
859 * the needed locks in suitable order.
860 */
mptcp_propagate_sndbuf(struct sock * sk,struct sock * ssk)861 static inline void mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
862 {
863 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
864
865 if (likely(READ_ONCE(ssk->sk_sndbuf) == subflow->cached_sndbuf))
866 return;
867
868 local_bh_disable();
869 mptcp_subflow_delegate(subflow, MPTCP_DELEGATE_SNDBUF);
870 local_bh_enable();
871 }
872
873 void mptcp_destroy_common(struct mptcp_sock *msk, unsigned int flags);
874
875 #define MPTCP_TOKEN_MAX_RETRIES 4
876
877 void __init mptcp_token_init(void);
mptcp_token_init_request(struct request_sock * req)878 static inline void mptcp_token_init_request(struct request_sock *req)
879 {
880 mptcp_subflow_rsk(req)->token_node.pprev = NULL;
881 }
882
883 int mptcp_token_new_request(struct request_sock *req);
884 void mptcp_token_destroy_request(struct request_sock *req);
885 int mptcp_token_new_connect(struct sock *ssk);
886 void mptcp_token_accept(struct mptcp_subflow_request_sock *r,
887 struct mptcp_sock *msk);
888 bool mptcp_token_exists(u32 token);
889 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token);
890 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot,
891 long *s_num);
892 void mptcp_token_destroy(struct mptcp_sock *msk);
893
894 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn);
895
896 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac);
897 __sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum);
898
899 void __init mptcp_pm_init(void);
900 void mptcp_pm_data_init(struct mptcp_sock *msk);
901 void mptcp_pm_data_reset(struct mptcp_sock *msk);
902 int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info,
903 struct mptcp_addr_info *addr);
904 int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info,
905 bool require_family,
906 struct mptcp_pm_addr_entry *entry);
907 bool mptcp_pm_addr_families_match(const struct sock *sk,
908 const struct mptcp_addr_info *loc,
909 const struct mptcp_addr_info *rem);
910 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
911 void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
912 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side);
913 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk);
914 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk);
915 void mptcp_pm_connection_closed(struct mptcp_sock *msk);
916 void mptcp_pm_subflow_established(struct mptcp_sock *msk);
917 bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk);
918 void mptcp_pm_subflow_check_next(struct mptcp_sock *msk, const struct sock *ssk,
919 const struct mptcp_subflow_context *subflow);
920 void mptcp_pm_add_addr_received(const struct sock *ssk,
921 const struct mptcp_addr_info *addr);
922 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk,
923 const struct mptcp_addr_info *addr);
924 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk);
925 bool mptcp_pm_nl_is_init_remote_addr(struct mptcp_sock *msk,
926 const struct mptcp_addr_info *remote);
927 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk);
928 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk,
929 const struct mptcp_rm_list *rm_list);
930 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup);
931 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq);
932 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk,
933 struct mptcp_addr_info *addr,
934 struct mptcp_addr_info *rem,
935 u8 bkup);
936 bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk,
937 const struct mptcp_addr_info *addr);
938 void mptcp_pm_free_anno_list(struct mptcp_sock *msk);
939 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk);
940 struct mptcp_pm_add_entry *
941 mptcp_pm_del_add_timer(struct mptcp_sock *msk,
942 const struct mptcp_addr_info *addr, bool check_id);
943 struct mptcp_pm_add_entry *
944 mptcp_lookup_anno_list_by_saddr(const struct mptcp_sock *msk,
945 const struct mptcp_addr_info *addr);
946 int mptcp_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk,
947 unsigned int id,
948 u8 *flags, int *ifindex);
949 int mptcp_pm_nl_get_flags_and_ifindex_by_id(struct mptcp_sock *msk, unsigned int id,
950 u8 *flags, int *ifindex);
951 int mptcp_userspace_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk,
952 unsigned int id,
953 u8 *flags, int *ifindex);
954 int mptcp_pm_set_flags(struct net *net, struct nlattr *token,
955 struct mptcp_pm_addr_entry *loc,
956 struct mptcp_pm_addr_entry *rem, u8 bkup);
957 int mptcp_pm_nl_set_flags(struct net *net, struct mptcp_pm_addr_entry *addr, u8 bkup);
958 int mptcp_userspace_pm_set_flags(struct net *net, struct nlattr *token,
959 struct mptcp_pm_addr_entry *loc,
960 struct mptcp_pm_addr_entry *rem, u8 bkup);
961 int mptcp_pm_announce_addr(struct mptcp_sock *msk,
962 const struct mptcp_addr_info *addr,
963 bool echo);
964 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
965 void mptcp_pm_remove_addrs(struct mptcp_sock *msk, struct list_head *rm_list);
966
967 void mptcp_free_local_addr_list(struct mptcp_sock *msk);
968 int mptcp_nl_cmd_announce(struct sk_buff *skb, struct genl_info *info);
969 int mptcp_nl_cmd_remove(struct sk_buff *skb, struct genl_info *info);
970 int mptcp_nl_cmd_sf_create(struct sk_buff *skb, struct genl_info *info);
971 int mptcp_nl_cmd_sf_destroy(struct sk_buff *skb, struct genl_info *info);
972
973 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk,
974 const struct sock *ssk, gfp_t gfp);
975 void mptcp_event_addr_announced(const struct sock *ssk, const struct mptcp_addr_info *info);
976 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id);
977 void mptcp_event_pm_listener(const struct sock *ssk,
978 enum mptcp_event_type event);
979 bool mptcp_userspace_pm_active(const struct mptcp_sock *msk);
980
981 void __mptcp_fastopen_gen_msk_ackseq(struct mptcp_sock *msk, struct mptcp_subflow_context *subflow,
982 const struct mptcp_options_received *mp_opt);
983 void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subflow,
984 struct request_sock *req);
985
mptcp_pm_should_add_signal(struct mptcp_sock * msk)986 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk)
987 {
988 return READ_ONCE(msk->pm.addr_signal) &
989 (BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO));
990 }
991
mptcp_pm_should_add_signal_addr(struct mptcp_sock * msk)992 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk)
993 {
994 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL);
995 }
996
mptcp_pm_should_add_signal_echo(struct mptcp_sock * msk)997 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk)
998 {
999 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO);
1000 }
1001
mptcp_pm_should_rm_signal(struct mptcp_sock * msk)1002 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk)
1003 {
1004 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL);
1005 }
1006
mptcp_pm_is_userspace(const struct mptcp_sock * msk)1007 static inline bool mptcp_pm_is_userspace(const struct mptcp_sock *msk)
1008 {
1009 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_USERSPACE;
1010 }
1011
mptcp_pm_is_kernel(const struct mptcp_sock * msk)1012 static inline bool mptcp_pm_is_kernel(const struct mptcp_sock *msk)
1013 {
1014 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_KERNEL;
1015 }
1016
mptcp_add_addr_len(int family,bool echo,bool port)1017 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port)
1018 {
1019 u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE;
1020
1021 if (family == AF_INET6)
1022 len = TCPOLEN_MPTCP_ADD_ADDR6_BASE;
1023 if (!echo)
1024 len += MPTCPOPT_THMAC_LEN;
1025 /* account for 2 trailing 'nop' options */
1026 if (port)
1027 len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN;
1028
1029 return len;
1030 }
1031
mptcp_rm_addr_len(const struct mptcp_rm_list * rm_list)1032 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list)
1033 {
1034 if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX)
1035 return -EINVAL;
1036
1037 return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1;
1038 }
1039
1040 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, const struct sk_buff *skb,
1041 unsigned int opt_size, unsigned int remaining,
1042 struct mptcp_addr_info *addr, bool *echo,
1043 bool *drop_other_suboptions);
1044 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining,
1045 struct mptcp_rm_list *rm_list);
1046 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
1047 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1048 int mptcp_userspace_pm_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1049 bool mptcp_pm_is_backup(struct mptcp_sock *msk, struct sock_common *skc);
1050 bool mptcp_pm_nl_is_backup(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1051 bool mptcp_userspace_pm_is_backup(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1052
subflow_get_local_id(const struct mptcp_subflow_context * subflow)1053 static inline u8 subflow_get_local_id(const struct mptcp_subflow_context *subflow)
1054 {
1055 int local_id = READ_ONCE(subflow->local_id);
1056
1057 if (local_id < 0)
1058 return 0;
1059 return local_id;
1060 }
1061
1062 void __init mptcp_pm_nl_init(void);
1063 void mptcp_pm_nl_work(struct mptcp_sock *msk);
1064 unsigned int mptcp_pm_get_add_addr_signal_max(const struct mptcp_sock *msk);
1065 unsigned int mptcp_pm_get_add_addr_accept_max(const struct mptcp_sock *msk);
1066 unsigned int mptcp_pm_get_subflows_max(const struct mptcp_sock *msk);
1067 unsigned int mptcp_pm_get_local_addr_max(const struct mptcp_sock *msk);
1068
1069 /* called under PM lock */
__mptcp_pm_close_subflow(struct mptcp_sock * msk)1070 static inline void __mptcp_pm_close_subflow(struct mptcp_sock *msk)
1071 {
1072 if (--msk->pm.subflows < mptcp_pm_get_subflows_max(msk))
1073 WRITE_ONCE(msk->pm.accept_subflow, true);
1074 }
1075
mptcp_pm_close_subflow(struct mptcp_sock * msk)1076 static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk)
1077 {
1078 spin_lock_bh(&msk->pm.lock);
1079 __mptcp_pm_close_subflow(msk);
1080 spin_unlock_bh(&msk->pm.lock);
1081 }
1082
1083 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk);
1084 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk);
1085
mptcp_get_ext(const struct sk_buff * skb)1086 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb)
1087 {
1088 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP);
1089 }
1090
1091 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops);
1092
__mptcp_check_fallback(const struct mptcp_sock * msk)1093 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk)
1094 {
1095 return test_bit(MPTCP_FALLBACK_DONE, &msk->flags);
1096 }
1097
mptcp_check_fallback(const struct sock * sk)1098 static inline bool mptcp_check_fallback(const struct sock *sk)
1099 {
1100 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1101 struct mptcp_sock *msk = mptcp_sk(subflow->conn);
1102
1103 return __mptcp_check_fallback(msk);
1104 }
1105
__mptcp_try_fallback(struct mptcp_sock * msk)1106 static inline bool __mptcp_try_fallback(struct mptcp_sock *msk)
1107 {
1108 if (test_bit(MPTCP_FALLBACK_DONE, &msk->flags)) {
1109 pr_debug("TCP fallback already done (msk=%p)\n", msk);
1110 return true;
1111 }
1112 spin_lock_bh(&msk->fallback_lock);
1113 if (!msk->allow_infinite_fallback) {
1114 spin_unlock_bh(&msk->fallback_lock);
1115 return false;
1116 }
1117
1118 msk->allow_subflows = false;
1119 set_bit(MPTCP_FALLBACK_DONE, &msk->flags);
1120 spin_unlock_bh(&msk->fallback_lock);
1121 return true;
1122 }
1123
mptcp_try_fallback(struct sock * ssk)1124 static inline bool mptcp_try_fallback(struct sock *ssk)
1125 {
1126 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
1127 struct sock *sk = subflow->conn;
1128 struct mptcp_sock *msk;
1129
1130 msk = mptcp_sk(sk);
1131 if (!__mptcp_try_fallback(msk))
1132 return false;
1133 if (READ_ONCE(msk->snd_data_fin_enable) && !(ssk->sk_shutdown & SEND_SHUTDOWN)) {
1134 gfp_t saved_allocation = ssk->sk_allocation;
1135
1136 /* we are in a atomic (BH) scope, override ssk default for data
1137 * fin allocation
1138 */
1139 ssk->sk_allocation = GFP_ATOMIC;
1140 ssk->sk_shutdown |= SEND_SHUTDOWN;
1141 tcp_shutdown(ssk, SEND_SHUTDOWN);
1142 ssk->sk_allocation = saved_allocation;
1143 }
1144 return true;
1145 }
1146
1147 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)\n", __func__, a)
1148
mptcp_check_infinite_map(struct sk_buff * skb)1149 static inline bool mptcp_check_infinite_map(struct sk_buff *skb)
1150 {
1151 struct mptcp_ext *mpext;
1152
1153 mpext = skb ? mptcp_get_ext(skb) : NULL;
1154 if (mpext && mpext->infinite_map)
1155 return true;
1156
1157 return false;
1158 }
1159
is_active_ssk(struct mptcp_subflow_context * subflow)1160 static inline bool is_active_ssk(struct mptcp_subflow_context *subflow)
1161 {
1162 return (subflow->request_mptcp || subflow->request_join);
1163 }
1164
subflow_simultaneous_connect(struct sock * sk)1165 static inline bool subflow_simultaneous_connect(struct sock *sk)
1166 {
1167 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1168
1169 return (1 << sk->sk_state) &
1170 (TCPF_ESTABLISHED | TCPF_FIN_WAIT1 | TCPF_FIN_WAIT2 | TCPF_CLOSING) &&
1171 is_active_ssk(subflow) &&
1172 !subflow->conn_finished;
1173 }
1174
1175 #ifdef CONFIG_SYN_COOKIES
1176 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1177 struct sk_buff *skb);
1178 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1179 struct sk_buff *skb);
1180 void __init mptcp_join_cookie_init(void);
1181 #else
1182 static inline void
subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock * subflow_req,struct sk_buff * skb)1183 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1184 struct sk_buff *skb) {}
1185 static inline bool
mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock * subflow_req,struct sk_buff * skb)1186 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1187 struct sk_buff *skb)
1188 {
1189 return false;
1190 }
1191
mptcp_join_cookie_init(void)1192 static inline void mptcp_join_cookie_init(void) {}
1193 #endif
1194
1195 #endif /* __MPTCP_PROTOCOL_H */
1196