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 u16 suboptions;
153 u32 token;
154 u32 nonce;
155 u16 use_map:1,
156 dsn64:1,
157 data_fin:1,
158 use_ack:1,
159 ack64:1,
160 mpc_map:1,
161 reset_reason:4,
162 reset_transient:1,
163 echo:1,
164 backup:1,
165 deny_join_id0:1,
166 __unused:2;
167 u8 join_id;
168 u64 thmac;
169 u8 hmac[MPTCPOPT_HMAC_LEN];
170 struct mptcp_addr_info addr;
171 struct mptcp_rm_list rm_list;
172 u64 ahmac;
173 u64 fail_seq;
174 };
175
mptcp_option(u8 subopt,u8 len,u8 nib,u8 field)176 static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field)
177 {
178 return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) |
179 ((nib & 0xF) << 8) | field);
180 }
181
182 enum mptcp_pm_status {
183 MPTCP_PM_ADD_ADDR_RECEIVED,
184 MPTCP_PM_ADD_ADDR_SEND_ACK,
185 MPTCP_PM_RM_ADDR_RECEIVED,
186 MPTCP_PM_ESTABLISHED,
187 MPTCP_PM_SUBFLOW_ESTABLISHED,
188 MPTCP_PM_ALREADY_ESTABLISHED, /* persistent status, set after ESTABLISHED event */
189 MPTCP_PM_MPC_ENDPOINT_ACCOUNTED /* persistent status, set after MPC local address is
190 * accounted int id_avail_bitmap
191 */
192 };
193
194 enum mptcp_pm_type {
195 MPTCP_PM_TYPE_KERNEL = 0,
196 MPTCP_PM_TYPE_USERSPACE,
197
198 __MPTCP_PM_TYPE_NR,
199 __MPTCP_PM_TYPE_MAX = __MPTCP_PM_TYPE_NR - 1,
200 };
201
202 /* Status bits below MPTCP_PM_ALREADY_ESTABLISHED need pm worker actions */
203 #define MPTCP_PM_WORK_MASK ((1 << MPTCP_PM_ALREADY_ESTABLISHED) - 1)
204
205 enum mptcp_addr_signal_status {
206 MPTCP_ADD_ADDR_SIGNAL,
207 MPTCP_ADD_ADDR_ECHO,
208 MPTCP_RM_ADDR_SIGNAL,
209 };
210
211 /* max value of mptcp_addr_info.id */
212 #define MPTCP_PM_MAX_ADDR_ID U8_MAX
213
214 struct mptcp_pm_data {
215 struct mptcp_addr_info local;
216 struct mptcp_addr_info remote;
217 struct list_head anno_list;
218 struct list_head userspace_pm_local_addr_list;
219
220 spinlock_t lock; /*protects the whole PM data */
221
222 u8 addr_signal;
223 bool server_side;
224 bool work_pending;
225 bool accept_addr;
226 bool accept_subflow;
227 bool remote_deny_join_id0;
228 u8 add_addr_signaled;
229 u8 add_addr_accepted;
230 u8 local_addr_used;
231 u8 pm_type;
232 u8 subflows;
233 u8 status;
234 DECLARE_BITMAP(id_avail_bitmap, MPTCP_PM_MAX_ADDR_ID + 1);
235 struct mptcp_rm_list rm_list_tx;
236 struct mptcp_rm_list rm_list_rx;
237 };
238
239 struct mptcp_pm_addr_entry {
240 struct list_head list;
241 struct mptcp_addr_info addr;
242 u8 flags;
243 int ifindex;
244 struct socket *lsk;
245 };
246
247 struct mptcp_data_frag {
248 struct list_head list;
249 u64 data_seq;
250 u16 data_len;
251 u16 offset;
252 u16 overhead;
253 u16 already_sent;
254 struct page *page;
255 };
256
257 /* MPTCP connection sock */
258 struct mptcp_sock {
259 /* inet_connection_sock must be the first member */
260 struct inet_connection_sock sk;
261 u64 local_key;
262 u64 remote_key;
263 u64 write_seq;
264 u64 bytes_sent;
265 u64 snd_nxt;
266 u64 bytes_received;
267 u64 ack_seq;
268 atomic64_t rcv_wnd_sent;
269 u64 rcv_data_fin_seq;
270 u64 bytes_retrans;
271 int rmem_fwd_alloc;
272 int snd_burst;
273 int old_wspace;
274 u64 recovery_snd_nxt; /* in recovery mode accept up to this seq;
275 * recovery related fields are under data_lock
276 * protection
277 */
278 u64 bytes_acked;
279 u64 snd_una;
280 u64 wnd_end;
281 unsigned long timer_ival;
282 u32 token;
283 int rmem_released;
284 unsigned long flags;
285 unsigned long cb_flags;
286 bool recovery; /* closing subflow write queue reinjected */
287 bool can_ack;
288 bool fully_established;
289 bool rcv_data_fin;
290 bool snd_data_fin_enable;
291 bool rcv_fastclose;
292 bool use_64bit_ack; /* Set when we received a 64-bit DSN */
293 bool csum_enabled;
294 bool allow_infinite_fallback;
295 u8 pending_state; /* A subflow asked to set this sk_state,
296 * protected by the msk data lock
297 */
298 u8 mpc_endpoint_id;
299 u8 recvmsg_inq:1,
300 cork:1,
301 nodelay:1,
302 fastopening:1,
303 in_accept_queue:1,
304 free_first:1,
305 rcvspace_init:1;
306 int keepalive_cnt;
307 int keepalive_idle;
308 int keepalive_intvl;
309 struct work_struct work;
310 struct sk_buff *ooo_last_skb;
311 struct rb_root out_of_order_queue;
312 struct sk_buff_head receive_queue;
313 struct list_head conn_list;
314 struct list_head rtx_queue;
315 struct mptcp_data_frag *first_pending;
316 struct list_head join_list;
317 struct sock *first; /* The mptcp ops can safely dereference, using suitable
318 * ONCE annotation, the subflow outside the socket
319 * lock as such sock is freed after close().
320 */
321 struct mptcp_pm_data pm;
322 struct mptcp_sched_ops *sched;
323 struct {
324 u32 space; /* bytes copied in last measurement window */
325 u32 copied; /* bytes copied in this measurement window */
326 u64 time; /* start time of measurement window */
327 u64 rtt_us; /* last maximum rtt of subflows */
328 } rcvq_space;
329 u8 scaling_ratio;
330
331 u32 subflow_id;
332 u32 setsockopt_seq;
333 char ca_name[TCP_CA_NAME_MAX];
334 };
335
336 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock)
337 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock)
338
339 #define mptcp_for_each_subflow(__msk, __subflow) \
340 list_for_each_entry(__subflow, &((__msk)->conn_list), node)
341 #define mptcp_for_each_subflow_safe(__msk, __subflow, __tmp) \
342 list_for_each_entry_safe(__subflow, __tmp, &((__msk)->conn_list), node)
343
msk_owned_by_me(const struct mptcp_sock * msk)344 static inline void msk_owned_by_me(const struct mptcp_sock *msk)
345 {
346 sock_owned_by_me((const struct sock *)msk);
347 }
348
349 #define mptcp_sk(ptr) container_of_const(ptr, struct mptcp_sock, sk.icsk_inet.sk)
350
351 /* the msk socket don't use the backlog, also account for the bulk
352 * free memory
353 */
__mptcp_rmem(const struct sock * sk)354 static inline int __mptcp_rmem(const struct sock *sk)
355 {
356 return atomic_read(&sk->sk_rmem_alloc) - READ_ONCE(mptcp_sk(sk)->rmem_released);
357 }
358
mptcp_win_from_space(const struct sock * sk,int space)359 static inline int mptcp_win_from_space(const struct sock *sk, int space)
360 {
361 return __tcp_win_from_space(mptcp_sk(sk)->scaling_ratio, space);
362 }
363
__mptcp_space(const struct sock * sk)364 static inline int __mptcp_space(const struct sock *sk)
365 {
366 return mptcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - __mptcp_rmem(sk));
367 }
368
mptcp_send_head(const struct sock * sk)369 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk)
370 {
371 const struct mptcp_sock *msk = mptcp_sk(sk);
372
373 return READ_ONCE(msk->first_pending);
374 }
375
mptcp_send_next(struct sock * sk)376 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk)
377 {
378 struct mptcp_sock *msk = mptcp_sk(sk);
379 struct mptcp_data_frag *cur;
380
381 cur = msk->first_pending;
382 return list_is_last(&cur->list, &msk->rtx_queue) ? NULL :
383 list_next_entry(cur, list);
384 }
385
mptcp_pending_tail(const struct sock * sk)386 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk)
387 {
388 const struct mptcp_sock *msk = mptcp_sk(sk);
389
390 if (!msk->first_pending)
391 return NULL;
392
393 if (WARN_ON_ONCE(list_empty(&msk->rtx_queue)))
394 return NULL;
395
396 return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list);
397 }
398
mptcp_rtx_head(struct sock * sk)399 static inline struct mptcp_data_frag *mptcp_rtx_head(struct sock *sk)
400 {
401 struct mptcp_sock *msk = mptcp_sk(sk);
402
403 if (msk->snd_una == READ_ONCE(msk->snd_nxt))
404 return NULL;
405
406 return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list);
407 }
408
409 struct csum_pseudo_header {
410 __be64 data_seq;
411 __be32 subflow_seq;
412 __be16 data_len;
413 __sum16 csum;
414 };
415
416 struct mptcp_subflow_request_sock {
417 struct tcp_request_sock sk;
418 u16 mp_capable : 1,
419 mp_join : 1,
420 backup : 1,
421 csum_reqd : 1,
422 allow_join_id0 : 1;
423 u8 local_id;
424 u8 remote_id;
425 u64 local_key;
426 u64 idsn;
427 u32 token;
428 u32 ssn_offset;
429 u64 thmac;
430 u32 local_nonce;
431 u32 remote_nonce;
432 struct mptcp_sock *msk;
433 struct hlist_nulls_node token_node;
434 };
435
436 static inline struct mptcp_subflow_request_sock *
mptcp_subflow_rsk(const struct request_sock * rsk)437 mptcp_subflow_rsk(const struct request_sock *rsk)
438 {
439 return (struct mptcp_subflow_request_sock *)rsk;
440 }
441
442 enum mptcp_data_avail {
443 MPTCP_SUBFLOW_NODATA,
444 MPTCP_SUBFLOW_DATA_AVAIL,
445 };
446
447 struct mptcp_delegated_action {
448 struct napi_struct napi;
449 struct list_head head;
450 };
451
452 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions);
453
454 #define MPTCP_DELEGATE_SCHEDULED 0
455 #define MPTCP_DELEGATE_SEND 1
456 #define MPTCP_DELEGATE_ACK 2
457 #define MPTCP_DELEGATE_SNDBUF 3
458
459 #define MPTCP_DELEGATE_ACTIONS_MASK (~BIT(MPTCP_DELEGATE_SCHEDULED))
460 /* MPTCP subflow context */
461 struct mptcp_subflow_context {
462 struct list_head node;/* conn_list of subflows */
463
464 struct_group(reset,
465
466 unsigned long avg_pacing_rate; /* protected by msk socket lock */
467 u64 local_key;
468 u64 remote_key;
469 u64 idsn;
470 u64 map_seq;
471 u32 snd_isn;
472 u32 token;
473 u32 rel_write_seq;
474 u32 map_subflow_seq;
475 u32 ssn_offset;
476 u32 map_data_len;
477 __wsum map_data_csum;
478 u32 map_csum_len;
479 u32 request_mptcp : 1, /* send MP_CAPABLE */
480 request_join : 1, /* send MP_JOIN */
481 request_bkup : 1,
482 mp_capable : 1, /* remote is MPTCP capable */
483 mp_join : 1, /* remote is JOINing */
484 fully_established : 1, /* path validated */
485 pm_notified : 1, /* PM hook called for established status */
486 conn_finished : 1,
487 map_valid : 1,
488 map_csum_reqd : 1,
489 map_data_fin : 1,
490 mpc_map : 1,
491 backup : 1,
492 send_mp_prio : 1,
493 send_mp_fail : 1,
494 send_fastclose : 1,
495 send_infinite_map : 1,
496 remote_key_valid : 1, /* received the peer key from */
497 disposable : 1, /* ctx can be free at ulp release time */
498 stale : 1, /* unable to snd/rcv data, do not use for xmit */
499 valid_csum_seen : 1, /* at least one csum validated */
500 is_mptfo : 1, /* subflow is doing TFO */
501 __unused : 10;
502 enum mptcp_data_avail data_avail;
503 bool scheduled;
504 u32 remote_nonce;
505 u64 thmac;
506 u32 local_nonce;
507 u32 remote_token;
508 union {
509 u8 hmac[MPTCPOPT_HMAC_LEN]; /* MPJ subflow only */
510 u64 iasn; /* initial ack sequence number, MPC subflows only */
511 };
512 s16 local_id; /* if negative not initialized yet */
513 u8 remote_id;
514 u8 reset_seen:1;
515 u8 reset_transient:1;
516 u8 reset_reason:4;
517 u8 stale_count;
518
519 u32 subflow_id;
520
521 long delegated_status;
522 unsigned long fail_tout;
523
524 );
525
526 struct list_head delegated_node; /* link into delegated_action, protected by local BH */
527
528 u32 setsockopt_seq;
529 u32 stale_rcv_tstamp;
530 int cached_sndbuf; /* sndbuf size when last synced with the msk sndbuf,
531 * protected by the msk socket lock
532 */
533
534 struct sock *tcp_sock; /* tcp sk backpointer */
535 struct sock *conn; /* parent mptcp_sock */
536 const struct inet_connection_sock_af_ops *icsk_af_ops;
537 void (*tcp_state_change)(struct sock *sk);
538 void (*tcp_error_report)(struct sock *sk);
539
540 struct rcu_head rcu;
541 };
542
543 static inline struct mptcp_subflow_context *
mptcp_subflow_ctx(const struct sock * sk)544 mptcp_subflow_ctx(const struct sock *sk)
545 {
546 struct inet_connection_sock *icsk = inet_csk(sk);
547
548 /* Use RCU on icsk_ulp_data only for sock diag code */
549 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data;
550 }
551
552 static inline struct sock *
mptcp_subflow_tcp_sock(const struct mptcp_subflow_context * subflow)553 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow)
554 {
555 return subflow->tcp_sock;
556 }
557
558 static inline void
mptcp_subflow_ctx_reset(struct mptcp_subflow_context * subflow)559 mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow)
560 {
561 memset(&subflow->reset, 0, sizeof(subflow->reset));
562 subflow->request_mptcp = 1;
563 WRITE_ONCE(subflow->local_id, -1);
564 }
565
566 static inline u64
mptcp_subflow_get_map_offset(const struct mptcp_subflow_context * subflow)567 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow)
568 {
569 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq -
570 subflow->ssn_offset -
571 subflow->map_subflow_seq;
572 }
573
574 static inline u64
mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context * subflow)575 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow)
576 {
577 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow);
578 }
579
580 void mptcp_subflow_process_delegated(struct sock *ssk, long actions);
581
mptcp_subflow_delegate(struct mptcp_subflow_context * subflow,int action)582 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action)
583 {
584 long old, set_bits = BIT(MPTCP_DELEGATE_SCHEDULED) | BIT(action);
585 struct mptcp_delegated_action *delegated;
586 bool schedule;
587
588 /* the caller held the subflow bh socket lock */
589 lockdep_assert_in_softirq();
590
591 /* The implied barrier pairs with tcp_release_cb_override()
592 * mptcp_napi_poll(), and ensures the below list check sees list
593 * updates done prior to delegated status bits changes
594 */
595 old = set_mask_bits(&subflow->delegated_status, 0, set_bits);
596 if (!(old & BIT(MPTCP_DELEGATE_SCHEDULED))) {
597 if (WARN_ON_ONCE(!list_empty(&subflow->delegated_node)))
598 return;
599
600 delegated = this_cpu_ptr(&mptcp_delegated_actions);
601 schedule = list_empty(&delegated->head);
602 list_add_tail(&subflow->delegated_node, &delegated->head);
603 sock_hold(mptcp_subflow_tcp_sock(subflow));
604 if (schedule)
605 napi_schedule(&delegated->napi);
606 }
607 }
608
609 static inline struct mptcp_subflow_context *
mptcp_subflow_delegated_next(struct mptcp_delegated_action * delegated)610 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated)
611 {
612 struct mptcp_subflow_context *ret;
613
614 if (list_empty(&delegated->head))
615 return NULL;
616
617 ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node);
618 list_del_init(&ret->delegated_node);
619 return ret;
620 }
621
622 int mptcp_is_enabled(const struct net *net);
623 unsigned int mptcp_get_add_addr_timeout(const struct net *net);
624 int mptcp_is_checksum_enabled(const struct net *net);
625 int mptcp_allow_join_id0(const struct net *net);
626 unsigned int mptcp_stale_loss_cnt(const struct net *net);
627 int mptcp_get_pm_type(const struct net *net);
628 const char *mptcp_get_scheduler(const struct net *net);
629 void __mptcp_subflow_fully_established(struct mptcp_sock *msk,
630 struct mptcp_subflow_context *subflow,
631 const struct mptcp_options_received *mp_opt);
632 bool __mptcp_retransmit_pending_data(struct sock *sk);
633 void mptcp_check_and_set_pending(struct sock *sk);
634 void __mptcp_push_pending(struct sock *sk, unsigned int flags);
635 bool mptcp_subflow_data_available(struct sock *sk);
636 void __init mptcp_subflow_init(void);
637 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how);
638 void mptcp_close_ssk(struct sock *sk, struct sock *ssk,
639 struct mptcp_subflow_context *subflow);
640 void __mptcp_subflow_send_ack(struct sock *ssk);
641 void mptcp_subflow_reset(struct sock *ssk);
642 void mptcp_subflow_queue_clean(struct sock *sk, struct sock *ssk);
643 void mptcp_sock_graft(struct sock *sk, struct socket *parent);
644 struct sock *__mptcp_nmpc_sk(struct mptcp_sock *msk);
645 bool __mptcp_close(struct sock *sk, long timeout);
646 void mptcp_cancel_work(struct sock *sk);
647 void __mptcp_unaccepted_force_close(struct sock *sk);
648 void mptcp_set_owner_r(struct sk_buff *skb, struct sock *sk);
649 void mptcp_set_state(struct sock *sk, int state);
650
651 bool mptcp_addresses_equal(const struct mptcp_addr_info *a,
652 const struct mptcp_addr_info *b, bool use_port);
653 void mptcp_local_address(const struct sock_common *skc, struct mptcp_addr_info *addr);
654
655 /* called with sk socket lock held */
656 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc,
657 const struct mptcp_addr_info *remote);
658 int mptcp_subflow_create_socket(struct sock *sk, unsigned short family,
659 struct socket **new_sock);
660 void mptcp_info2sockaddr(const struct mptcp_addr_info *info,
661 struct sockaddr_storage *addr,
662 unsigned short family);
663 struct mptcp_sched_ops *mptcp_sched_find(const char *name);
664 int mptcp_register_scheduler(struct mptcp_sched_ops *sched);
665 void mptcp_unregister_scheduler(struct mptcp_sched_ops *sched);
666 void mptcp_sched_init(void);
667 int mptcp_init_sched(struct mptcp_sock *msk,
668 struct mptcp_sched_ops *sched);
669 void mptcp_release_sched(struct mptcp_sock *msk);
670 void mptcp_subflow_set_scheduled(struct mptcp_subflow_context *subflow,
671 bool scheduled);
672 struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk);
673 struct sock *mptcp_subflow_get_retrans(struct mptcp_sock *msk);
674 int mptcp_sched_get_send(struct mptcp_sock *msk);
675 int mptcp_sched_get_retrans(struct mptcp_sock *msk);
676
__tcp_can_send(const struct sock * ssk)677 static inline bool __tcp_can_send(const struct sock *ssk)
678 {
679 /* only send if our side has not closed yet */
680 return ((1 << inet_sk_state_load(ssk)) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT));
681 }
682
__mptcp_subflow_active(struct mptcp_subflow_context * subflow)683 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow)
684 {
685 /* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */
686 if (subflow->request_join && !subflow->fully_established)
687 return false;
688
689 return __tcp_can_send(mptcp_subflow_tcp_sock(subflow));
690 }
691
692 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow);
693
694 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow);
695
696 void mptcp_subflow_drop_ctx(struct sock *ssk);
697
mptcp_subflow_tcp_fallback(struct sock * sk,struct mptcp_subflow_context * ctx)698 static inline void mptcp_subflow_tcp_fallback(struct sock *sk,
699 struct mptcp_subflow_context *ctx)
700 {
701 sk->sk_data_ready = sock_def_readable;
702 sk->sk_state_change = ctx->tcp_state_change;
703 sk->sk_write_space = sk_stream_write_space;
704 sk->sk_error_report = ctx->tcp_error_report;
705
706 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops;
707 }
708
709 void __init mptcp_proto_init(void);
710 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
711 int __init mptcp_proto_v6_init(void);
712 #endif
713
714 struct sock *mptcp_sk_clone_init(const struct sock *sk,
715 const struct mptcp_options_received *mp_opt,
716 struct sock *ssk,
717 struct request_sock *req);
718 void mptcp_get_options(const struct sk_buff *skb,
719 struct mptcp_options_received *mp_opt);
720
721 void mptcp_finish_connect(struct sock *sk);
722 void __mptcp_sync_state(struct sock *sk, int state);
723 void mptcp_reset_tout_timer(struct mptcp_sock *msk, unsigned long fail_tout);
724
mptcp_stop_tout_timer(struct sock * sk)725 static inline void mptcp_stop_tout_timer(struct sock *sk)
726 {
727 if (!inet_csk(sk)->icsk_mtup.probe_timestamp)
728 return;
729
730 sk_stop_timer(sk, &sk->sk_timer);
731 inet_csk(sk)->icsk_mtup.probe_timestamp = 0;
732 }
733
mptcp_set_close_tout(struct sock * sk,unsigned long tout)734 static inline void mptcp_set_close_tout(struct sock *sk, unsigned long tout)
735 {
736 /* avoid 0 timestamp, as that means no close timeout */
737 inet_csk(sk)->icsk_mtup.probe_timestamp = tout ? : 1;
738 }
739
mptcp_start_tout_timer(struct sock * sk)740 static inline void mptcp_start_tout_timer(struct sock *sk)
741 {
742 mptcp_set_close_tout(sk, tcp_jiffies32);
743 mptcp_reset_tout_timer(mptcp_sk(sk), 0);
744 }
745
mptcp_is_fully_established(struct sock * sk)746 static inline bool mptcp_is_fully_established(struct sock *sk)
747 {
748 return inet_sk_state_load(sk) == TCP_ESTABLISHED &&
749 READ_ONCE(mptcp_sk(sk)->fully_established);
750 }
751 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk);
752 void mptcp_data_ready(struct sock *sk, struct sock *ssk);
753 bool mptcp_finish_join(struct sock *sk);
754 bool mptcp_schedule_work(struct sock *sk);
755 int mptcp_setsockopt(struct sock *sk, int level, int optname,
756 sockptr_t optval, unsigned int optlen);
757 int mptcp_getsockopt(struct sock *sk, int level, int optname,
758 char __user *optval, int __user *option);
759
760 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq);
mptcp_expand_seq(u64 old_seq,u64 cur_seq,bool use_64bit)761 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit)
762 {
763 if (use_64bit)
764 return cur_seq;
765
766 return __mptcp_expand_seq(old_seq, cur_seq);
767 }
768 void __mptcp_check_push(struct sock *sk, struct sock *ssk);
769 void __mptcp_data_acked(struct sock *sk);
770 void __mptcp_error_report(struct sock *sk);
771 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)772 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk)
773 {
774 return READ_ONCE(msk->snd_data_fin_enable) &&
775 READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt);
776 }
777
mptcp_write_space(struct sock * sk)778 static inline void mptcp_write_space(struct sock *sk)
779 {
780 if (sk_stream_is_writeable(sk)) {
781 /* pairs with memory barrier in mptcp_poll */
782 smp_mb();
783 if (test_and_clear_bit(MPTCP_NOSPACE, &mptcp_sk(sk)->flags))
784 sk_stream_write_space(sk);
785 }
786 }
787
__mptcp_sync_sndbuf(struct sock * sk)788 static inline void __mptcp_sync_sndbuf(struct sock *sk)
789 {
790 struct mptcp_subflow_context *subflow;
791 int ssk_sndbuf, new_sndbuf;
792
793 if (sk->sk_userlocks & SOCK_SNDBUF_LOCK)
794 return;
795
796 new_sndbuf = sock_net(sk)->ipv4.sysctl_tcp_wmem[0];
797 mptcp_for_each_subflow(mptcp_sk(sk), subflow) {
798 ssk_sndbuf = READ_ONCE(mptcp_subflow_tcp_sock(subflow)->sk_sndbuf);
799
800 subflow->cached_sndbuf = ssk_sndbuf;
801 new_sndbuf += ssk_sndbuf;
802 }
803
804 /* the msk max wmem limit is <nr_subflows> * tcp wmem[2] */
805 WRITE_ONCE(sk->sk_sndbuf, new_sndbuf);
806 mptcp_write_space(sk);
807 }
808
809 /* The called held both the msk socket and the subflow socket locks,
810 * possibly under BH
811 */
__mptcp_propagate_sndbuf(struct sock * sk,struct sock * ssk)812 static inline void __mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
813 {
814 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
815
816 if (READ_ONCE(ssk->sk_sndbuf) != subflow->cached_sndbuf)
817 __mptcp_sync_sndbuf(sk);
818 }
819
820 /* the caller held only the subflow socket lock, either in process or
821 * BH context. Additionally this can be called under the msk data lock,
822 * so we can't acquire such lock here: let the delegate action acquires
823 * the needed locks in suitable order.
824 */
mptcp_propagate_sndbuf(struct sock * sk,struct sock * ssk)825 static inline void mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
826 {
827 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
828
829 if (likely(READ_ONCE(ssk->sk_sndbuf) == subflow->cached_sndbuf))
830 return;
831
832 local_bh_disable();
833 mptcp_subflow_delegate(subflow, MPTCP_DELEGATE_SNDBUF);
834 local_bh_enable();
835 }
836
837 void mptcp_destroy_common(struct mptcp_sock *msk, unsigned int flags);
838
839 #define MPTCP_TOKEN_MAX_RETRIES 4
840
841 void __init mptcp_token_init(void);
mptcp_token_init_request(struct request_sock * req)842 static inline void mptcp_token_init_request(struct request_sock *req)
843 {
844 mptcp_subflow_rsk(req)->token_node.pprev = NULL;
845 }
846
847 int mptcp_token_new_request(struct request_sock *req);
848 void mptcp_token_destroy_request(struct request_sock *req);
849 int mptcp_token_new_connect(struct sock *ssk);
850 void mptcp_token_accept(struct mptcp_subflow_request_sock *r,
851 struct mptcp_sock *msk);
852 bool mptcp_token_exists(u32 token);
853 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token);
854 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot,
855 long *s_num);
856 void mptcp_token_destroy(struct mptcp_sock *msk);
857
858 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn);
859
860 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac);
861 __sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum);
862
863 void __init mptcp_pm_init(void);
864 void mptcp_pm_data_init(struct mptcp_sock *msk);
865 void mptcp_pm_data_reset(struct mptcp_sock *msk);
866 int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info,
867 struct mptcp_addr_info *addr);
868 int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info,
869 bool require_family,
870 struct mptcp_pm_addr_entry *entry);
871 bool mptcp_pm_addr_families_match(const struct sock *sk,
872 const struct mptcp_addr_info *loc,
873 const struct mptcp_addr_info *rem);
874 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
875 void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
876 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side);
877 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk);
878 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk);
879 void mptcp_pm_connection_closed(struct mptcp_sock *msk);
880 void mptcp_pm_subflow_established(struct mptcp_sock *msk);
881 bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk);
882 void mptcp_pm_subflow_check_next(struct mptcp_sock *msk, const struct sock *ssk,
883 const struct mptcp_subflow_context *subflow);
884 void mptcp_pm_add_addr_received(const struct sock *ssk,
885 const struct mptcp_addr_info *addr);
886 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk,
887 const struct mptcp_addr_info *addr);
888 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk);
889 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk);
890 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk,
891 const struct mptcp_rm_list *rm_list);
892 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup);
893 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq);
894 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk,
895 struct mptcp_addr_info *addr,
896 struct mptcp_addr_info *rem,
897 u8 bkup);
898 bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk,
899 const struct mptcp_addr_info *addr);
900 void mptcp_pm_free_anno_list(struct mptcp_sock *msk);
901 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk);
902 struct mptcp_pm_add_entry *
903 mptcp_pm_del_add_timer(struct mptcp_sock *msk,
904 const struct mptcp_addr_info *addr, bool check_id);
905 struct mptcp_pm_add_entry *
906 mptcp_lookup_anno_list_by_saddr(const struct mptcp_sock *msk,
907 const struct mptcp_addr_info *addr);
908 int mptcp_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk,
909 unsigned int id,
910 u8 *flags, int *ifindex);
911 int mptcp_pm_nl_get_flags_and_ifindex_by_id(struct mptcp_sock *msk, unsigned int id,
912 u8 *flags, int *ifindex);
913 int mptcp_userspace_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk,
914 unsigned int id,
915 u8 *flags, int *ifindex);
916 int mptcp_pm_set_flags(struct net *net, struct nlattr *token,
917 struct mptcp_pm_addr_entry *loc,
918 struct mptcp_pm_addr_entry *rem, u8 bkup);
919 int mptcp_pm_nl_set_flags(struct net *net, struct mptcp_pm_addr_entry *addr, u8 bkup);
920 int mptcp_userspace_pm_set_flags(struct net *net, struct nlattr *token,
921 struct mptcp_pm_addr_entry *loc,
922 struct mptcp_pm_addr_entry *rem, u8 bkup);
923 int mptcp_pm_announce_addr(struct mptcp_sock *msk,
924 const struct mptcp_addr_info *addr,
925 bool echo);
926 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
927 int mptcp_pm_remove_subflow(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
928 void mptcp_pm_remove_addrs(struct mptcp_sock *msk, struct list_head *rm_list);
929 void mptcp_pm_remove_addrs_and_subflows(struct mptcp_sock *msk,
930 struct list_head *rm_list);
931
932 void mptcp_free_local_addr_list(struct mptcp_sock *msk);
933 int mptcp_nl_cmd_announce(struct sk_buff *skb, struct genl_info *info);
934 int mptcp_nl_cmd_remove(struct sk_buff *skb, struct genl_info *info);
935 int mptcp_nl_cmd_sf_create(struct sk_buff *skb, struct genl_info *info);
936 int mptcp_nl_cmd_sf_destroy(struct sk_buff *skb, struct genl_info *info);
937
938 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk,
939 const struct sock *ssk, gfp_t gfp);
940 void mptcp_event_addr_announced(const struct sock *ssk, const struct mptcp_addr_info *info);
941 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id);
942 void mptcp_event_pm_listener(const struct sock *ssk,
943 enum mptcp_event_type event);
944 bool mptcp_userspace_pm_active(const struct mptcp_sock *msk);
945
946 void __mptcp_fastopen_gen_msk_ackseq(struct mptcp_sock *msk, struct mptcp_subflow_context *subflow,
947 const struct mptcp_options_received *mp_opt);
948 void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subflow,
949 struct request_sock *req);
950
mptcp_pm_should_add_signal(struct mptcp_sock * msk)951 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk)
952 {
953 return READ_ONCE(msk->pm.addr_signal) &
954 (BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO));
955 }
956
mptcp_pm_should_add_signal_addr(struct mptcp_sock * msk)957 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk)
958 {
959 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL);
960 }
961
mptcp_pm_should_add_signal_echo(struct mptcp_sock * msk)962 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk)
963 {
964 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO);
965 }
966
mptcp_pm_should_rm_signal(struct mptcp_sock * msk)967 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk)
968 {
969 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL);
970 }
971
mptcp_pm_is_userspace(const struct mptcp_sock * msk)972 static inline bool mptcp_pm_is_userspace(const struct mptcp_sock *msk)
973 {
974 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_USERSPACE;
975 }
976
mptcp_pm_is_kernel(const struct mptcp_sock * msk)977 static inline bool mptcp_pm_is_kernel(const struct mptcp_sock *msk)
978 {
979 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_KERNEL;
980 }
981
mptcp_add_addr_len(int family,bool echo,bool port)982 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port)
983 {
984 u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE;
985
986 if (family == AF_INET6)
987 len = TCPOLEN_MPTCP_ADD_ADDR6_BASE;
988 if (!echo)
989 len += MPTCPOPT_THMAC_LEN;
990 /* account for 2 trailing 'nop' options */
991 if (port)
992 len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN;
993
994 return len;
995 }
996
mptcp_rm_addr_len(const struct mptcp_rm_list * rm_list)997 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list)
998 {
999 if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX)
1000 return -EINVAL;
1001
1002 return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1;
1003 }
1004
1005 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, const struct sk_buff *skb,
1006 unsigned int opt_size, unsigned int remaining,
1007 struct mptcp_addr_info *addr, bool *echo,
1008 bool *drop_other_suboptions);
1009 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining,
1010 struct mptcp_rm_list *rm_list);
1011 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
1012 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1013 int mptcp_userspace_pm_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1014
subflow_get_local_id(const struct mptcp_subflow_context * subflow)1015 static inline u8 subflow_get_local_id(const struct mptcp_subflow_context *subflow)
1016 {
1017 int local_id = READ_ONCE(subflow->local_id);
1018
1019 if (local_id < 0)
1020 return 0;
1021 return local_id;
1022 }
1023
1024 void __init mptcp_pm_nl_init(void);
1025 void mptcp_pm_nl_work(struct mptcp_sock *msk);
1026 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk,
1027 const struct mptcp_rm_list *rm_list);
1028 unsigned int mptcp_pm_get_add_addr_signal_max(const struct mptcp_sock *msk);
1029 unsigned int mptcp_pm_get_add_addr_accept_max(const struct mptcp_sock *msk);
1030 unsigned int mptcp_pm_get_subflows_max(const struct mptcp_sock *msk);
1031 unsigned int mptcp_pm_get_local_addr_max(const struct mptcp_sock *msk);
1032
1033 /* called under PM lock */
__mptcp_pm_close_subflow(struct mptcp_sock * msk)1034 static inline void __mptcp_pm_close_subflow(struct mptcp_sock *msk)
1035 {
1036 if (--msk->pm.subflows < mptcp_pm_get_subflows_max(msk))
1037 WRITE_ONCE(msk->pm.accept_subflow, true);
1038 }
1039
mptcp_pm_close_subflow(struct mptcp_sock * msk)1040 static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk)
1041 {
1042 spin_lock_bh(&msk->pm.lock);
1043 __mptcp_pm_close_subflow(msk);
1044 spin_unlock_bh(&msk->pm.lock);
1045 }
1046
1047 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk);
1048 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk);
1049
mptcp_get_ext(const struct sk_buff * skb)1050 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb)
1051 {
1052 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP);
1053 }
1054
1055 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops);
1056
__mptcp_check_fallback(const struct mptcp_sock * msk)1057 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk)
1058 {
1059 return test_bit(MPTCP_FALLBACK_DONE, &msk->flags);
1060 }
1061
mptcp_check_fallback(const struct sock * sk)1062 static inline bool mptcp_check_fallback(const struct sock *sk)
1063 {
1064 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1065 struct mptcp_sock *msk = mptcp_sk(subflow->conn);
1066
1067 return __mptcp_check_fallback(msk);
1068 }
1069
__mptcp_do_fallback(struct mptcp_sock * msk)1070 static inline void __mptcp_do_fallback(struct mptcp_sock *msk)
1071 {
1072 if (test_bit(MPTCP_FALLBACK_DONE, &msk->flags)) {
1073 pr_debug("TCP fallback already done (msk=%p)", msk);
1074 return;
1075 }
1076 set_bit(MPTCP_FALLBACK_DONE, &msk->flags);
1077 }
1078
mptcp_do_fallback(struct sock * ssk)1079 static inline void mptcp_do_fallback(struct sock *ssk)
1080 {
1081 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
1082 struct sock *sk = subflow->conn;
1083 struct mptcp_sock *msk;
1084
1085 msk = mptcp_sk(sk);
1086 __mptcp_do_fallback(msk);
1087 if (READ_ONCE(msk->snd_data_fin_enable) && !(ssk->sk_shutdown & SEND_SHUTDOWN)) {
1088 gfp_t saved_allocation = ssk->sk_allocation;
1089
1090 /* we are in a atomic (BH) scope, override ssk default for data
1091 * fin allocation
1092 */
1093 ssk->sk_allocation = GFP_ATOMIC;
1094 ssk->sk_shutdown |= SEND_SHUTDOWN;
1095 tcp_shutdown(ssk, SEND_SHUTDOWN);
1096 ssk->sk_allocation = saved_allocation;
1097 }
1098 }
1099
1100 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a)
1101
mptcp_check_infinite_map(struct sk_buff * skb)1102 static inline bool mptcp_check_infinite_map(struct sk_buff *skb)
1103 {
1104 struct mptcp_ext *mpext;
1105
1106 mpext = skb ? mptcp_get_ext(skb) : NULL;
1107 if (mpext && mpext->infinite_map)
1108 return true;
1109
1110 return false;
1111 }
1112
is_active_ssk(struct mptcp_subflow_context * subflow)1113 static inline bool is_active_ssk(struct mptcp_subflow_context *subflow)
1114 {
1115 return (subflow->request_mptcp || subflow->request_join);
1116 }
1117
subflow_simultaneous_connect(struct sock * sk)1118 static inline bool subflow_simultaneous_connect(struct sock *sk)
1119 {
1120 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1121
1122 return (1 << sk->sk_state) &
1123 (TCPF_ESTABLISHED | TCPF_FIN_WAIT1 | TCPF_FIN_WAIT2 | TCPF_CLOSING) &&
1124 is_active_ssk(subflow) &&
1125 !subflow->conn_finished;
1126 }
1127
1128 #ifdef CONFIG_SYN_COOKIES
1129 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1130 struct sk_buff *skb);
1131 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1132 struct sk_buff *skb);
1133 void __init mptcp_join_cookie_init(void);
1134 #else
1135 static inline void
subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock * subflow_req,struct sk_buff * skb)1136 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1137 struct sk_buff *skb) {}
1138 static inline bool
mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock * subflow_req,struct sk_buff * skb)1139 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1140 struct sk_buff *skb)
1141 {
1142 return false;
1143 }
1144
mptcp_join_cookie_init(void)1145 static inline void mptcp_join_cookie_init(void) {}
1146 #endif
1147
1148 #endif /* __MPTCP_PROTOCOL_H */
1149