1 /* SPDX-License-Identifier: GPL-2.0 */
2 /* Multipath TCP
3 *
4 * Copyright (c) 2017 - 2019, Intel Corporation.
5 */
6
7 #ifndef __MPTCP_PROTOCOL_H
8 #define __MPTCP_PROTOCOL_H
9
10 #include <linux/random.h>
11 #include <net/tcp.h>
12 #include <net/inet_connection_sock.h>
13 #include <uapi/linux/mptcp.h>
14 #include <net/genetlink.h>
15 #include <net/rstreason.h>
16
17 #define MPTCP_SUPPORTED_VERSION 1
18
19 /* MPTCP option bits */
20 #define OPTION_MPTCP_MPC_SYN BIT(0)
21 #define OPTION_MPTCP_MPC_SYNACK BIT(1)
22 #define OPTION_MPTCP_MPC_ACK BIT(2)
23 #define OPTION_MPTCP_MPJ_SYN BIT(3)
24 #define OPTION_MPTCP_MPJ_SYNACK BIT(4)
25 #define OPTION_MPTCP_MPJ_ACK BIT(5)
26 #define OPTION_MPTCP_ADD_ADDR BIT(6)
27 #define OPTION_MPTCP_RM_ADDR BIT(7)
28 #define OPTION_MPTCP_FASTCLOSE BIT(8)
29 #define OPTION_MPTCP_PRIO BIT(9)
30 #define OPTION_MPTCP_RST BIT(10)
31 #define OPTION_MPTCP_DSS BIT(11)
32 #define OPTION_MPTCP_FAIL BIT(12)
33
34 #define OPTION_MPTCP_CSUMREQD BIT(13)
35
36 #define OPTIONS_MPTCP_MPC (OPTION_MPTCP_MPC_SYN | OPTION_MPTCP_MPC_SYNACK | \
37 OPTION_MPTCP_MPC_ACK)
38 #define OPTIONS_MPTCP_MPJ (OPTION_MPTCP_MPJ_SYN | OPTION_MPTCP_MPJ_SYNACK | \
39 OPTION_MPTCP_MPJ_ACK)
40
41 /* MPTCP option subtypes */
42 #define MPTCPOPT_MP_CAPABLE 0
43 #define MPTCPOPT_MP_JOIN 1
44 #define MPTCPOPT_DSS 2
45 #define MPTCPOPT_ADD_ADDR 3
46 #define MPTCPOPT_RM_ADDR 4
47 #define MPTCPOPT_MP_PRIO 5
48 #define MPTCPOPT_MP_FAIL 6
49 #define MPTCPOPT_MP_FASTCLOSE 7
50 #define MPTCPOPT_RST 8
51
52 /* MPTCP suboption lengths */
53 #define TCPOLEN_MPTCP_MPC_SYN 4
54 #define TCPOLEN_MPTCP_MPC_SYNACK 12
55 #define TCPOLEN_MPTCP_MPC_ACK 20
56 #define TCPOLEN_MPTCP_MPC_ACK_DATA 22
57 #define TCPOLEN_MPTCP_MPJ_SYN 12
58 #define TCPOLEN_MPTCP_MPJ_SYNACK 16
59 #define TCPOLEN_MPTCP_MPJ_ACK 24
60 #define TCPOLEN_MPTCP_DSS_BASE 4
61 #define TCPOLEN_MPTCP_DSS_ACK32 4
62 #define TCPOLEN_MPTCP_DSS_ACK64 8
63 #define TCPOLEN_MPTCP_DSS_MAP32 10
64 #define TCPOLEN_MPTCP_DSS_MAP64 14
65 #define TCPOLEN_MPTCP_DSS_CHECKSUM 2
66 #define TCPOLEN_MPTCP_ADD_ADDR 16
67 #define TCPOLEN_MPTCP_ADD_ADDR_PORT 18
68 #define TCPOLEN_MPTCP_ADD_ADDR_BASE 8
69 #define TCPOLEN_MPTCP_ADD_ADDR_BASE_PORT 10
70 #define TCPOLEN_MPTCP_ADD_ADDR6 28
71 #define TCPOLEN_MPTCP_ADD_ADDR6_PORT 30
72 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE 20
73 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE_PORT 22
74 #define TCPOLEN_MPTCP_PORT_LEN 2
75 #define TCPOLEN_MPTCP_PORT_ALIGN 2
76 #define TCPOLEN_MPTCP_RM_ADDR_BASE 3
77 #define TCPOLEN_MPTCP_PRIO 3
78 #define TCPOLEN_MPTCP_PRIO_ALIGN 4
79 #define TCPOLEN_MPTCP_FASTCLOSE 12
80 #define TCPOLEN_MPTCP_RST 4
81 #define TCPOLEN_MPTCP_FAIL 12
82
83 #define TCPOLEN_MPTCP_MPC_ACK_DATA_CSUM (TCPOLEN_MPTCP_DSS_CHECKSUM + TCPOLEN_MPTCP_MPC_ACK_DATA)
84
85 /* MPTCP MP_JOIN flags */
86 #define MPTCPOPT_BACKUP BIT(0)
87 #define MPTCPOPT_THMAC_LEN 8
88
89 /* MPTCP MP_CAPABLE flags */
90 #define MPTCP_VERSION_MASK (0x0F)
91 #define MPTCP_CAP_CHECKSUM_REQD BIT(7)
92 #define MPTCP_CAP_EXTENSIBILITY BIT(6)
93 #define MPTCP_CAP_DENY_JOIN_ID0 BIT(5)
94 #define MPTCP_CAP_HMAC_SHA256 BIT(0)
95 #define MPTCP_CAP_FLAG_MASK (0x1F)
96
97 /* MPTCP DSS flags */
98 #define MPTCP_DSS_DATA_FIN BIT(4)
99 #define MPTCP_DSS_DSN64 BIT(3)
100 #define MPTCP_DSS_HAS_MAP BIT(2)
101 #define MPTCP_DSS_ACK64 BIT(1)
102 #define MPTCP_DSS_HAS_ACK BIT(0)
103 #define MPTCP_DSS_FLAG_MASK (0x1F)
104
105 /* MPTCP ADD_ADDR flags */
106 #define MPTCP_ADDR_ECHO BIT(0)
107
108 /* MPTCP MP_PRIO flags */
109 #define MPTCP_PRIO_BKUP BIT(0)
110
111 /* MPTCP TCPRST flags */
112 #define MPTCP_RST_TRANSIENT BIT(0)
113
114 /* MPTCP socket atomic flags */
115 #define MPTCP_WORK_RTX 1
116 #define MPTCP_FALLBACK_DONE 2
117 #define MPTCP_WORK_CLOSE_SUBFLOW 3
118
119 /* MPTCP socket release cb flags */
120 #define MPTCP_PUSH_PENDING 1
121 #define MPTCP_CLEAN_UNA 2
122 #define MPTCP_ERROR_REPORT 3
123 #define MPTCP_RETRANSMIT 4
124 #define MPTCP_FLUSH_JOIN_LIST 5
125 #define MPTCP_SYNC_STATE 6
126 #define MPTCP_SYNC_SNDBUF 7
127
128 struct mptcp_skb_cb {
129 u64 map_seq;
130 u64 end_seq;
131 u32 offset;
132 u8 has_rxtstamp;
133 u8 cant_coalesce;
134 };
135
136 #define MPTCP_SKB_CB(__skb) ((struct mptcp_skb_cb *)&((__skb)->cb[0]))
137
before64(__u64 seq1,__u64 seq2)138 static inline bool before64(__u64 seq1, __u64 seq2)
139 {
140 return (__s64)(seq1 - seq2) < 0;
141 }
142
143 #define after64(seq2, seq1) before64(seq1, seq2)
144
145 struct mptcp_options_received {
146 u64 sndr_key;
147 u64 rcvr_key;
148 u64 data_ack;
149 u64 data_seq;
150 u32 subflow_seq;
151 u16 data_len;
152 __sum16 csum;
153 struct_group(status,
154 u16 suboptions;
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 );
168 u8 join_id;
169 u32 token;
170 u32 nonce;
171 u64 thmac;
172 u8 hmac[MPTCPOPT_HMAC_LEN];
173 struct mptcp_addr_info addr;
174 struct mptcp_rm_list rm_list;
175 u64 ahmac;
176 u64 fail_seq;
177 };
178
mptcp_option(u8 subopt,u8 len,u8 nib,u8 field)179 static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field)
180 {
181 return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) |
182 ((nib & 0xF) << 8) | field);
183 }
184
185 enum mptcp_pm_status {
186 MPTCP_PM_ADD_ADDR_RECEIVED,
187 MPTCP_PM_ADD_ADDR_SEND_ACK,
188 MPTCP_PM_RM_ADDR_RECEIVED,
189 MPTCP_PM_ESTABLISHED,
190 MPTCP_PM_SUBFLOW_ESTABLISHED,
191 MPTCP_PM_ALREADY_ESTABLISHED, /* persistent status, set after ESTABLISHED event */
192 MPTCP_PM_MPC_ENDPOINT_ACCOUNTED /* persistent status, set after MPC local address is
193 * accounted int id_avail_bitmap
194 */
195 };
196
197 enum mptcp_pm_type {
198 MPTCP_PM_TYPE_KERNEL = 0,
199 MPTCP_PM_TYPE_USERSPACE,
200
201 __MPTCP_PM_TYPE_NR,
202 __MPTCP_PM_TYPE_MAX = __MPTCP_PM_TYPE_NR - 1,
203 };
204
205 /* Status bits below MPTCP_PM_ALREADY_ESTABLISHED need pm worker actions */
206 #define MPTCP_PM_WORK_MASK ((1 << MPTCP_PM_ALREADY_ESTABLISHED) - 1)
207
208 enum mptcp_addr_signal_status {
209 MPTCP_ADD_ADDR_SIGNAL,
210 MPTCP_ADD_ADDR_ECHO,
211 MPTCP_RM_ADDR_SIGNAL,
212 };
213
214 /* max value of mptcp_addr_info.id */
215 #define MPTCP_PM_MAX_ADDR_ID U8_MAX
216
217 struct mptcp_pm_data {
218 struct mptcp_addr_info local;
219 struct mptcp_addr_info remote;
220 struct list_head anno_list;
221 struct list_head userspace_pm_local_addr_list;
222
223 spinlock_t lock; /*protects the whole PM data */
224
225 struct_group(reset,
226
227 u8 addr_signal;
228 bool server_side;
229 bool work_pending;
230 bool accept_addr;
231 bool accept_subflow;
232 bool remote_deny_join_id0;
233 u8 add_addr_signaled;
234 u8 add_addr_accepted;
235 u8 local_addr_used;
236 u8 pm_type;
237 u8 extra_subflows;
238 u8 status;
239
240 );
241
242 DECLARE_BITMAP(id_avail_bitmap, MPTCP_PM_MAX_ADDR_ID + 1);
243 struct mptcp_rm_list rm_list_tx;
244 struct mptcp_rm_list rm_list_rx;
245 };
246
247 struct mptcp_pm_local {
248 struct mptcp_addr_info addr;
249 u8 flags;
250 int ifindex;
251 };
252
253 struct mptcp_pm_addr_entry {
254 struct list_head list;
255 struct mptcp_addr_info addr;
256 u8 flags;
257 int ifindex;
258 struct socket *lsk;
259 };
260
261 struct mptcp_data_frag {
262 struct list_head list;
263 u64 data_seq;
264 u16 data_len;
265 u16 offset;
266 u16 overhead;
267 u16 already_sent;
268 struct page *page;
269 };
270
271 /* MPTCP connection sock */
272 struct mptcp_sock {
273 /* inet_connection_sock must be the first member */
274 struct inet_connection_sock sk;
275 u64 local_key; /* protected by the first subflow socket lock
276 * lockless access read
277 */
278 u64 remote_key; /* same as above */
279 u64 write_seq;
280 u64 bytes_sent;
281 u64 snd_nxt;
282 u64 bytes_received;
283 u64 ack_seq;
284 atomic64_t rcv_wnd_sent;
285 u64 rcv_data_fin_seq;
286 u64 bytes_retrans;
287 u64 bytes_consumed;
288 int snd_burst;
289 int old_wspace;
290 u64 recovery_snd_nxt; /* in recovery mode accept up to this seq;
291 * recovery related fields are under data_lock
292 * protection
293 */
294 u64 bytes_acked;
295 u64 snd_una;
296 u64 wnd_end;
297 u32 last_data_sent;
298 u32 last_data_recv;
299 u32 last_ack_recv;
300 unsigned long timer_ival;
301 u32 token;
302 unsigned long flags;
303 unsigned long cb_flags;
304 bool recovery; /* closing subflow write queue reinjected */
305 bool can_ack;
306 bool fully_established;
307 bool rcv_data_fin;
308 bool snd_data_fin_enable;
309 bool rcv_fastclose;
310 bool use_64bit_ack; /* Set when we received a 64-bit DSN */
311 bool csum_enabled;
312 bool allow_infinite_fallback;
313 u8 pending_state; /* A subflow asked to set this sk_state,
314 * protected by the msk data lock
315 */
316 u8 mpc_endpoint_id;
317 u8 recvmsg_inq:1,
318 cork:1,
319 nodelay:1,
320 fastopening:1,
321 in_accept_queue:1,
322 free_first:1,
323 rcvspace_init:1;
324 u32 notsent_lowat;
325 int keepalive_cnt;
326 int keepalive_idle;
327 int keepalive_intvl;
328 int maxseg;
329 struct work_struct work;
330 struct sk_buff *ooo_last_skb;
331 struct rb_root out_of_order_queue;
332 struct list_head conn_list;
333 struct list_head rtx_queue;
334 struct mptcp_data_frag *first_pending;
335 struct list_head join_list;
336 struct sock *first; /* The mptcp ops can safely dereference, using suitable
337 * ONCE annotation, the subflow outside the socket
338 * lock as such sock is freed after close().
339 */
340 struct mptcp_pm_data pm;
341 struct mptcp_sched_ops *sched;
342 struct {
343 int space; /* bytes copied in last measurement window */
344 int copied; /* bytes copied in this measurement window */
345 u64 time; /* start time of measurement window */
346 u64 rtt_us; /* last maximum rtt of subflows */
347 } rcvq_space;
348 u8 scaling_ratio;
349 bool allow_subflows;
350
351 u32 subflow_id;
352 u32 setsockopt_seq;
353 char ca_name[TCP_CA_NAME_MAX];
354
355 spinlock_t fallback_lock; /* protects fallback,
356 * allow_infinite_fallback and
357 * allow_join
358 */
359
360 struct list_head backlog_list; /* protected by the data lock */
361 u32 backlog_len;
362 u32 backlog_unaccounted;
363 };
364
365 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock)
366 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock)
367
368 #define mptcp_for_each_subflow(__msk, __subflow) \
369 list_for_each_entry(__subflow, &((__msk)->conn_list), node)
370 #define mptcp_for_each_subflow_safe(__msk, __subflow, __tmp) \
371 list_for_each_entry_safe(__subflow, __tmp, &((__msk)->conn_list), node)
372 #define mptcp_next_subflow(__msk, __subflow) \
373 list_next_entry_circular(__subflow, &((__msk)->conn_list), node)
374
375 extern struct genl_family mptcp_genl_family;
376
msk_owned_by_me(const struct mptcp_sock * msk)377 static inline void msk_owned_by_me(const struct mptcp_sock *msk)
378 {
379 sock_owned_by_me((const struct sock *)msk);
380 }
381
382 #ifdef CONFIG_DEBUG_NET
383 /* MPTCP-specific: we might (indirectly) call this helper with the wrong sk */
384 #undef tcp_sk
385 #define tcp_sk(ptr) ({ \
386 typeof(ptr) _ptr = (ptr); \
387 WARN_ON(_ptr->sk_protocol != IPPROTO_TCP); \
388 container_of_const(_ptr, struct tcp_sock, inet_conn.icsk_inet.sk); \
389 })
390 #define mptcp_sk(ptr) ({ \
391 typeof(ptr) _ptr = (ptr); \
392 WARN_ON(_ptr->sk_protocol != IPPROTO_MPTCP); \
393 container_of_const(_ptr, struct mptcp_sock, sk.icsk_inet.sk); \
394 })
395
396 #else /* !CONFIG_DEBUG_NET */
397 #define mptcp_sk(ptr) container_of_const(ptr, struct mptcp_sock, sk.icsk_inet.sk)
398 #endif
399
mptcp_win_from_space(const struct sock * sk,int space)400 static inline int mptcp_win_from_space(const struct sock *sk, int space)
401 {
402 return __tcp_win_from_space(mptcp_sk(sk)->scaling_ratio, space);
403 }
404
mptcp_space_from_win(const struct sock * sk,int win)405 static inline int mptcp_space_from_win(const struct sock *sk, int win)
406 {
407 return __tcp_space_from_win(mptcp_sk(sk)->scaling_ratio, win);
408 }
409
__mptcp_space(const struct sock * sk)410 static inline int __mptcp_space(const struct sock *sk)
411 {
412 return mptcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) -
413 READ_ONCE(mptcp_sk(sk)->backlog_len) -
414 sk_rmem_alloc_get(sk));
415 }
416
mptcp_send_head(const struct sock * sk)417 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk)
418 {
419 const struct mptcp_sock *msk = mptcp_sk(sk);
420
421 return msk->first_pending;
422 }
423
mptcp_send_next(struct sock * sk)424 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk)
425 {
426 struct mptcp_sock *msk = mptcp_sk(sk);
427 struct mptcp_data_frag *cur;
428
429 cur = msk->first_pending;
430 return list_is_last(&cur->list, &msk->rtx_queue) ? NULL :
431 list_next_entry(cur, list);
432 }
433
mptcp_pending_tail(const struct sock * sk)434 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk)
435 {
436 const struct mptcp_sock *msk = mptcp_sk(sk);
437
438 if (!msk->first_pending)
439 return NULL;
440
441 if (WARN_ON_ONCE(list_empty(&msk->rtx_queue)))
442 return NULL;
443
444 return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list);
445 }
446
mptcp_rtx_head(struct sock * sk)447 static inline struct mptcp_data_frag *mptcp_rtx_head(struct sock *sk)
448 {
449 struct mptcp_sock *msk = mptcp_sk(sk);
450
451 if (msk->snd_una == msk->snd_nxt)
452 return NULL;
453
454 return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list);
455 }
456
457 struct csum_pseudo_header {
458 __be64 data_seq;
459 __be32 subflow_seq;
460 __be16 data_len;
461 __sum16 csum;
462 };
463
464 struct mptcp_subflow_request_sock {
465 struct tcp_request_sock sk;
466 u16 mp_capable : 1,
467 mp_join : 1,
468 backup : 1,
469 request_bkup : 1,
470 csum_reqd : 1,
471 allow_join_id0 : 1;
472 u8 local_id;
473 u8 remote_id;
474 u64 local_key;
475 u64 idsn;
476 u32 token;
477 u32 ssn_offset;
478 u64 thmac;
479 u32 local_nonce;
480 u32 remote_nonce;
481 struct mptcp_sock *msk;
482 struct hlist_nulls_node token_node;
483 };
484
485 static inline struct mptcp_subflow_request_sock *
mptcp_subflow_rsk(const struct request_sock * rsk)486 mptcp_subflow_rsk(const struct request_sock *rsk)
487 {
488 return (struct mptcp_subflow_request_sock *)rsk;
489 }
490
491 struct mptcp_delegated_action {
492 struct napi_struct napi;
493 local_lock_t bh_lock;
494 struct list_head head;
495 };
496
497 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions);
498
499 #define MPTCP_DELEGATE_SCHEDULED 0
500 #define MPTCP_DELEGATE_SEND 1
501 #define MPTCP_DELEGATE_ACK 2
502 #define MPTCP_DELEGATE_SNDBUF 3
503
504 #define MPTCP_DELEGATE_ACTIONS_MASK (~BIT(MPTCP_DELEGATE_SCHEDULED))
505 /* MPTCP subflow context */
506 struct mptcp_subflow_context {
507 struct list_head node;/* conn_list of subflows */
508
509 struct_group(reset,
510
511 unsigned long avg_pacing_rate; /* protected by msk socket lock */
512 u64 local_key;
513 u64 remote_key;
514 u64 idsn;
515 u64 map_seq;
516 u64 rcv_wnd_sent;
517 u32 snd_isn;
518 u32 token;
519 u32 rel_write_seq;
520 u32 map_subflow_seq;
521 u32 ssn_offset;
522 u32 map_data_len;
523 __wsum map_data_csum;
524 u32 map_csum_len;
525 u32 request_mptcp : 1, /* send MP_CAPABLE */
526 request_join : 1, /* send MP_JOIN */
527 request_bkup : 1,
528 mp_capable : 1, /* remote is MPTCP capable */
529 mp_join : 1, /* remote is JOINing */
530 pm_notified : 1, /* PM hook called for established status */
531 conn_finished : 1,
532 map_valid : 1,
533 map_csum_reqd : 1,
534 map_data_fin : 1,
535 mpc_map : 1,
536 backup : 1,
537 send_mp_prio : 1,
538 send_mp_fail : 1,
539 send_fastclose : 1,
540 send_infinite_map : 1,
541 remote_key_valid : 1, /* received the peer key from */
542 disposable : 1, /* ctx can be free at ulp release time */
543 closing : 1, /* must not pass rx data to msk anymore */
544 stale : 1, /* unable to snd/rcv data, do not use for xmit */
545 valid_csum_seen : 1, /* at least one csum validated */
546 is_mptfo : 1, /* subflow is doing TFO */
547 close_event_done : 1, /* has done the post-closed part */
548 mpc_drop : 1, /* the MPC option has been dropped in a rtx */
549 __unused : 8;
550 bool data_avail;
551 bool scheduled;
552 bool pm_listener; /* a listener managed by the kernel PM? */
553 bool fully_established; /* path validated */
554 u32 lent_mem_frag;
555 u32 remote_nonce;
556 u64 thmac;
557 u32 local_nonce;
558 u32 remote_token;
559 union {
560 u8 hmac[MPTCPOPT_HMAC_LEN]; /* MPJ subflow only */
561 u64 iasn; /* initial ack sequence number, MPC subflows only */
562 };
563 s16 local_id; /* if negative not initialized yet */
564 u8 remote_id;
565 u8 reset_seen:1;
566 u8 reset_transient:1;
567 u8 reset_reason:4;
568 u8 stale_count;
569
570 u32 subflow_id;
571
572 long delegated_status;
573 unsigned long fail_tout;
574
575 );
576
577 struct list_head delegated_node; /* link into delegated_action, protected by local BH */
578
579 u32 setsockopt_seq;
580 u32 stale_rcv_tstamp;
581 int cached_sndbuf; /* sndbuf size when last synced with the msk sndbuf,
582 * protected by the msk socket lock
583 */
584
585 struct sock *tcp_sock; /* tcp sk backpointer */
586 struct sock *conn; /* parent mptcp_sock */
587 const struct inet_connection_sock_af_ops *icsk_af_ops;
588 void (*tcp_state_change)(struct sock *sk);
589 void (*tcp_error_report)(struct sock *sk);
590
591 struct rcu_head rcu;
592 };
593
594 static inline struct mptcp_subflow_context *
mptcp_subflow_ctx(const struct sock * sk)595 mptcp_subflow_ctx(const struct sock *sk)
596 {
597 const struct inet_connection_sock *icsk = inet_csk(sk);
598
599 /* Use RCU on icsk_ulp_data only for sock diag code */
600 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data;
601 }
602
603 static inline struct sock *
mptcp_subflow_tcp_sock(const struct mptcp_subflow_context * subflow)604 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow)
605 {
606 return subflow->tcp_sock;
607 }
608
609 static inline void
mptcp_subflow_ctx_reset(struct mptcp_subflow_context * subflow)610 mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow)
611 {
612 memset(&subflow->reset, 0, sizeof(subflow->reset));
613 subflow->request_mptcp = 1;
614 WRITE_ONCE(subflow->local_id, -1);
615 }
616
617 /* Convert reset reasons in MPTCP to enum sk_rst_reason type */
618 static inline enum sk_rst_reason
sk_rst_convert_mptcp_reason(u32 reason)619 sk_rst_convert_mptcp_reason(u32 reason)
620 {
621 switch (reason) {
622 case MPTCP_RST_EUNSPEC:
623 return SK_RST_REASON_MPTCP_RST_EUNSPEC;
624 case MPTCP_RST_EMPTCP:
625 return SK_RST_REASON_MPTCP_RST_EMPTCP;
626 case MPTCP_RST_ERESOURCE:
627 return SK_RST_REASON_MPTCP_RST_ERESOURCE;
628 case MPTCP_RST_EPROHIBIT:
629 return SK_RST_REASON_MPTCP_RST_EPROHIBIT;
630 case MPTCP_RST_EWQ2BIG:
631 return SK_RST_REASON_MPTCP_RST_EWQ2BIG;
632 case MPTCP_RST_EBADPERF:
633 return SK_RST_REASON_MPTCP_RST_EBADPERF;
634 case MPTCP_RST_EMIDDLEBOX:
635 return SK_RST_REASON_MPTCP_RST_EMIDDLEBOX;
636 default:
637 /* It should not happen, or else errors may occur
638 * in MPTCP layer
639 */
640 return SK_RST_REASON_ERROR;
641 }
642 }
643
644 static inline void
mptcp_send_active_reset_reason(struct sock * sk)645 mptcp_send_active_reset_reason(struct sock *sk)
646 {
647 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
648 enum sk_rst_reason reason;
649
650 reason = sk_rst_convert_mptcp_reason(subflow->reset_reason);
651 tcp_send_active_reset(sk, GFP_ATOMIC, reason);
652 }
653
654 /* Made the fwd mem carried by the given skb available to the msk,
655 * To be paired with a previous mptcp_subflow_lend_fwdmem() before freeing
656 * the skb or setting the skb ownership.
657 */
mptcp_borrow_fwdmem(struct sock * sk,struct sk_buff * skb)658 static inline void mptcp_borrow_fwdmem(struct sock *sk, struct sk_buff *skb)
659 {
660 struct sock *ssk = skb->sk;
661
662 /* The subflow just lend the skb fwd memory; if the subflow meanwhile
663 * closed, mptcp_close_ssk() already released the ssk rcv memory.
664 */
665 DEBUG_NET_WARN_ON_ONCE(skb->destructor);
666 sk_forward_alloc_add(sk, skb->truesize);
667 if (!ssk)
668 return;
669
670 atomic_sub(skb->truesize, &ssk->sk_rmem_alloc);
671 skb->sk = NULL;
672 }
673
674 static inline void
__mptcp_subflow_lend_fwdmem(struct mptcp_subflow_context * subflow,int size)675 __mptcp_subflow_lend_fwdmem(struct mptcp_subflow_context *subflow, int size)
676 {
677 int frag = (subflow->lent_mem_frag + size) & (PAGE_SIZE - 1);
678
679 subflow->lent_mem_frag = frag;
680 }
681
682 static inline void
mptcp_subflow_lend_fwdmem(struct mptcp_subflow_context * subflow,struct sk_buff * skb)683 mptcp_subflow_lend_fwdmem(struct mptcp_subflow_context *subflow,
684 struct sk_buff *skb)
685 {
686 __mptcp_subflow_lend_fwdmem(subflow, skb->truesize);
687 skb->destructor = NULL;
688 }
689
690 static inline u64
mptcp_subflow_get_map_offset(const struct mptcp_subflow_context * subflow)691 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow)
692 {
693 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq -
694 subflow->ssn_offset -
695 subflow->map_subflow_seq;
696 }
697
698 static inline u64
mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context * subflow)699 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow)
700 {
701 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow);
702 }
703
704 void mptcp_subflow_process_delegated(struct sock *ssk, long actions);
705
mptcp_subflow_delegate(struct mptcp_subflow_context * subflow,int action)706 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action)
707 {
708 long old, set_bits = BIT(MPTCP_DELEGATE_SCHEDULED) | BIT(action);
709 struct mptcp_delegated_action *delegated;
710 bool schedule;
711
712 /* the caller held the subflow bh socket lock */
713 lockdep_assert_in_softirq();
714
715 /* The implied barrier pairs with tcp_release_cb_override()
716 * mptcp_napi_poll(), and ensures the below list check sees list
717 * updates done prior to delegated status bits changes
718 */
719 old = set_mask_bits(&subflow->delegated_status, 0, set_bits);
720 if (!(old & BIT(MPTCP_DELEGATE_SCHEDULED))) {
721 if (WARN_ON_ONCE(!list_empty(&subflow->delegated_node)))
722 return;
723
724 local_lock_nested_bh(&mptcp_delegated_actions.bh_lock);
725 delegated = this_cpu_ptr(&mptcp_delegated_actions);
726 schedule = list_empty(&delegated->head);
727 list_add_tail(&subflow->delegated_node, &delegated->head);
728 local_unlock_nested_bh(&mptcp_delegated_actions.bh_lock);
729 sock_hold(mptcp_subflow_tcp_sock(subflow));
730 if (schedule)
731 napi_schedule(&delegated->napi);
732 }
733 }
734
735 static inline struct mptcp_subflow_context *
mptcp_subflow_delegated_next(struct mptcp_delegated_action * delegated)736 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated)
737 {
738 struct mptcp_subflow_context *ret;
739
740 local_lock_nested_bh(&mptcp_delegated_actions.bh_lock);
741 if (list_empty(&delegated->head)) {
742 local_unlock_nested_bh(&mptcp_delegated_actions.bh_lock);
743 return NULL;
744 }
745
746 ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node);
747 list_del_init(&ret->delegated_node);
748 local_unlock_nested_bh(&mptcp_delegated_actions.bh_lock);
749 return ret;
750 }
751
752 void __mptcp_inherit_memcg(struct sock *sk, struct sock *ssk, gfp_t gfp);
753 void __mptcp_inherit_cgrp_data(struct sock *sk, struct sock *ssk);
754
755 int mptcp_is_enabled(const struct net *net);
756 unsigned int mptcp_get_add_addr_timeout(const struct net *net);
757 int mptcp_is_checksum_enabled(const struct net *net);
758 int mptcp_allow_join_id0(const struct net *net);
759 unsigned int mptcp_stale_loss_cnt(const struct net *net);
760 unsigned int mptcp_close_timeout(const struct sock *sk);
761 int mptcp_get_pm_type(const struct net *net);
762 const char *mptcp_get_path_manager(const struct net *net);
763 const char *mptcp_get_scheduler(const struct net *net);
764
765 void mptcp_active_disable(struct sock *sk);
766 bool mptcp_active_should_disable(struct sock *ssk);
767 void mptcp_active_enable(struct sock *sk);
768
769 void mptcp_get_available_schedulers(char *buf, size_t maxlen);
770 void __mptcp_subflow_fully_established(struct mptcp_sock *msk,
771 struct mptcp_subflow_context *subflow,
772 const struct mptcp_options_received *mp_opt);
773 bool __mptcp_retransmit_pending_data(struct sock *sk);
774 void mptcp_check_and_set_pending(struct sock *sk);
775 void __mptcp_push_pending(struct sock *sk, unsigned int flags);
776 bool mptcp_subflow_data_available(struct sock *sk);
777 void __init mptcp_subflow_init(void);
778 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how);
779 void mptcp_close_ssk(struct sock *sk, struct sock *ssk,
780 struct mptcp_subflow_context *subflow);
781 void __mptcp_subflow_send_ack(struct sock *ssk);
782 void mptcp_subflow_reset(struct sock *ssk);
783 void mptcp_subflow_queue_clean(struct sock *sk, struct sock *ssk);
784 void mptcp_sock_graft(struct sock *sk, struct socket *parent);
785 struct sock *__mptcp_nmpc_sk(struct mptcp_sock *msk);
786 bool __mptcp_close(struct sock *sk, long timeout);
787 void mptcp_cancel_work(struct sock *sk);
788 void __mptcp_unaccepted_force_close(struct sock *sk);
789 void mptcp_set_state(struct sock *sk, int state);
790
791 bool mptcp_addresses_equal(const struct mptcp_addr_info *a,
792 const struct mptcp_addr_info *b, bool use_port);
793 void mptcp_local_address(const struct sock_common *skc,
794 struct mptcp_addr_info *addr);
795 void mptcp_remote_address(const struct sock_common *skc,
796 struct mptcp_addr_info *addr);
797
798 /* called with sk socket lock held */
799 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_pm_local *local,
800 const struct mptcp_addr_info *remote);
801 int mptcp_subflow_create_socket(struct sock *sk, unsigned short family,
802 struct socket **new_sock);
803 void mptcp_info2sockaddr(const struct mptcp_addr_info *info,
804 struct sockaddr_storage *addr,
805 unsigned short family);
806 struct mptcp_sched_ops *mptcp_sched_find(const char *name);
807 int mptcp_validate_scheduler(struct mptcp_sched_ops *sched);
808 int mptcp_register_scheduler(struct mptcp_sched_ops *sched);
809 void mptcp_unregister_scheduler(struct mptcp_sched_ops *sched);
810 void mptcp_sched_init(void);
811 int mptcp_init_sched(struct mptcp_sock *msk,
812 struct mptcp_sched_ops *sched);
813 void mptcp_release_sched(struct mptcp_sock *msk);
814 void mptcp_subflow_set_scheduled(struct mptcp_subflow_context *subflow,
815 bool scheduled);
816 struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk);
817 struct sock *mptcp_subflow_get_retrans(struct mptcp_sock *msk);
818 int mptcp_sched_get_send(struct mptcp_sock *msk);
819 int mptcp_sched_get_retrans(struct mptcp_sock *msk);
820
mptcp_data_avail(const struct mptcp_sock * msk)821 static inline u64 mptcp_data_avail(const struct mptcp_sock *msk)
822 {
823 return READ_ONCE(msk->bytes_received) - READ_ONCE(msk->bytes_consumed);
824 }
825
mptcp_epollin_ready(const struct sock * sk)826 static inline bool mptcp_epollin_ready(const struct sock *sk)
827 {
828 u64 data_avail = mptcp_data_avail(mptcp_sk(sk));
829
830 if (!data_avail)
831 return false;
832
833 /* mptcp doesn't have to deal with small skbs in the receive queue,
834 * as it can always coalesce them
835 */
836 return (data_avail >= sk->sk_rcvlowat) ||
837 tcp_under_memory_pressure(sk);
838 }
839
840 int mptcp_set_rcvlowat(struct sock *sk, int val);
841
__tcp_can_send(const struct sock * ssk)842 static inline bool __tcp_can_send(const struct sock *ssk)
843 {
844 /* only send if our side has not closed yet */
845 return ((1 << inet_sk_state_load(ssk)) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT));
846 }
847
__mptcp_subflow_active(struct mptcp_subflow_context * subflow)848 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow)
849 {
850 /* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */
851 if (subflow->request_join && !READ_ONCE(subflow->fully_established))
852 return false;
853
854 return __tcp_can_send(mptcp_subflow_tcp_sock(subflow));
855 }
856
857 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow);
858
859 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow);
860
861 void mptcp_subflow_drop_ctx(struct sock *ssk);
862
mptcp_subflow_tcp_fallback(struct sock * sk,struct mptcp_subflow_context * ctx)863 static inline void mptcp_subflow_tcp_fallback(struct sock *sk,
864 struct mptcp_subflow_context *ctx)
865 {
866 sk->sk_data_ready = sock_def_readable;
867 sk->sk_state_change = ctx->tcp_state_change;
868 sk->sk_write_space = sk_stream_write_space;
869 sk->sk_error_report = ctx->tcp_error_report;
870
871 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops;
872 }
873
874 void __init mptcp_proto_init(void);
875 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
876 int __init mptcp_proto_v6_init(void);
877 #endif
878
879 struct sock *mptcp_sk_clone_init(const struct sock *sk,
880 const struct mptcp_options_received *mp_opt,
881 struct sock *ssk,
882 struct request_sock *req);
883 void mptcp_get_options(const struct sk_buff *skb,
884 struct mptcp_options_received *mp_opt);
885
886 void mptcp_finish_connect(struct sock *sk);
887 void __mptcp_sync_state(struct sock *sk, int state);
888 void mptcp_reset_tout_timer(struct mptcp_sock *msk, unsigned long fail_tout);
889
mptcp_stop_tout_timer(struct sock * sk)890 static inline void mptcp_stop_tout_timer(struct sock *sk)
891 {
892 if (!inet_csk(sk)->icsk_mtup.probe_timestamp)
893 return;
894
895 sk_stop_timer(sk, &inet_csk(sk)->mptcp_tout_timer);
896 inet_csk(sk)->icsk_mtup.probe_timestamp = 0;
897 }
898
mptcp_set_close_tout(struct sock * sk,unsigned long tout)899 static inline void mptcp_set_close_tout(struct sock *sk, unsigned long tout)
900 {
901 /* avoid 0 timestamp, as that means no close timeout */
902 inet_csk(sk)->icsk_mtup.probe_timestamp = tout ? : 1;
903 }
904
mptcp_start_tout_timer(struct sock * sk)905 static inline void mptcp_start_tout_timer(struct sock *sk)
906 {
907 mptcp_set_close_tout(sk, tcp_jiffies32);
908 mptcp_reset_tout_timer(mptcp_sk(sk), 0);
909 }
910
mptcp_is_fully_established(struct sock * sk)911 static inline bool mptcp_is_fully_established(struct sock *sk)
912 {
913 return inet_sk_state_load(sk) == TCP_ESTABLISHED &&
914 READ_ONCE(mptcp_sk(sk)->fully_established);
915 }
916
917 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk);
918 void mptcp_data_ready(struct sock *sk, struct sock *ssk);
919 bool mptcp_finish_join(struct sock *sk);
920 bool mptcp_schedule_work(struct sock *sk);
921 int mptcp_setsockopt(struct sock *sk, int level, int optname,
922 sockptr_t optval, unsigned int optlen);
923 int mptcp_getsockopt(struct sock *sk, int level, int optname,
924 char __user *optval, int __user *option);
925
926 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq);
mptcp_expand_seq(u64 old_seq,u64 cur_seq,bool use_64bit)927 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit)
928 {
929 if (use_64bit)
930 return cur_seq;
931
932 return __mptcp_expand_seq(old_seq, cur_seq);
933 }
934 void __mptcp_check_push(struct sock *sk, struct sock *ssk);
935 void __mptcp_data_acked(struct sock *sk);
936 void __mptcp_error_report(struct sock *sk);
937 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)938 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk)
939 {
940 return READ_ONCE(msk->snd_data_fin_enable) &&
941 READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt);
942 }
943
mptcp_notsent_lowat(const struct sock * sk)944 static inline u32 mptcp_notsent_lowat(const struct sock *sk)
945 {
946 struct net *net = sock_net(sk);
947 u32 val;
948
949 val = READ_ONCE(mptcp_sk(sk)->notsent_lowat);
950 return val ?: READ_ONCE(net->ipv4.sysctl_tcp_notsent_lowat);
951 }
952
mptcp_stream_memory_free(const struct sock * sk,int wake)953 static inline bool mptcp_stream_memory_free(const struct sock *sk, int wake)
954 {
955 const struct mptcp_sock *msk = mptcp_sk(sk);
956 u32 notsent_bytes;
957
958 notsent_bytes = READ_ONCE(msk->write_seq) - READ_ONCE(msk->snd_nxt);
959 return (notsent_bytes << wake) < mptcp_notsent_lowat(sk);
960 }
961
__mptcp_stream_is_writeable(const struct sock * sk,int wake)962 static inline bool __mptcp_stream_is_writeable(const struct sock *sk, int wake)
963 {
964 return mptcp_stream_memory_free(sk, wake) &&
965 __sk_stream_is_writeable(sk, wake);
966 }
967
mptcp_write_space(struct sock * sk)968 static inline void mptcp_write_space(struct sock *sk)
969 {
970 /* pairs with memory barrier in mptcp_poll */
971 smp_mb();
972 if (mptcp_stream_memory_free(sk, 1))
973 sk_stream_write_space(sk);
974 }
975
__mptcp_sync_sndbuf(struct sock * sk)976 static inline void __mptcp_sync_sndbuf(struct sock *sk)
977 {
978 struct mptcp_subflow_context *subflow;
979 int ssk_sndbuf, new_sndbuf;
980
981 if (sk->sk_userlocks & SOCK_SNDBUF_LOCK)
982 return;
983
984 new_sndbuf = READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_wmem[0]);
985 mptcp_for_each_subflow(mptcp_sk(sk), subflow) {
986 ssk_sndbuf = READ_ONCE(mptcp_subflow_tcp_sock(subflow)->sk_sndbuf);
987
988 subflow->cached_sndbuf = ssk_sndbuf;
989 new_sndbuf += ssk_sndbuf;
990 }
991
992 /* the msk max wmem limit is <nr_subflows> * tcp wmem[2] */
993 WRITE_ONCE(sk->sk_sndbuf, new_sndbuf);
994 mptcp_write_space(sk);
995 }
996
997 /* The called held both the msk socket and the subflow socket locks,
998 * possibly under BH
999 */
__mptcp_propagate_sndbuf(struct sock * sk,struct sock * ssk)1000 static inline void __mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
1001 {
1002 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
1003
1004 if (READ_ONCE(ssk->sk_sndbuf) != subflow->cached_sndbuf)
1005 __mptcp_sync_sndbuf(sk);
1006 }
1007
1008 /* the caller held only the subflow socket lock, either in process or
1009 * BH context. Additionally this can be called under the msk data lock,
1010 * so we can't acquire such lock here: let the delegate action acquires
1011 * the needed locks in suitable order.
1012 */
mptcp_propagate_sndbuf(struct sock * sk,struct sock * ssk)1013 static inline void mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
1014 {
1015 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
1016
1017 if (likely(READ_ONCE(ssk->sk_sndbuf) == subflow->cached_sndbuf))
1018 return;
1019
1020 local_bh_disable();
1021 mptcp_subflow_delegate(subflow, MPTCP_DELEGATE_SNDBUF);
1022 local_bh_enable();
1023 }
1024
1025 #define MPTCP_TOKEN_MAX_RETRIES 4
1026
1027 void __init mptcp_token_init(void);
mptcp_token_init_request(struct request_sock * req)1028 static inline void mptcp_token_init_request(struct request_sock *req)
1029 {
1030 mptcp_subflow_rsk(req)->token_node.pprev = NULL;
1031 }
1032
1033 int mptcp_token_new_request(struct request_sock *req);
1034 void mptcp_token_destroy_request(struct request_sock *req);
1035 int mptcp_token_new_connect(struct sock *ssk);
1036 void mptcp_token_accept(struct mptcp_subflow_request_sock *r,
1037 struct mptcp_sock *msk);
1038 bool mptcp_token_exists(u32 token);
1039 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token);
1040 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot,
1041 long *s_num);
1042 void mptcp_token_destroy(struct mptcp_sock *msk);
1043
1044 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn);
1045
1046 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac);
1047 __sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum);
1048
1049 void __init mptcp_pm_init(void);
1050 void mptcp_pm_data_init(struct mptcp_sock *msk);
1051 void mptcp_pm_data_reset(struct mptcp_sock *msk);
1052 void mptcp_pm_destroy(struct mptcp_sock *msk);
1053 int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info,
1054 struct mptcp_addr_info *addr);
1055 int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info,
1056 bool require_family,
1057 struct mptcp_pm_addr_entry *entry);
1058 bool mptcp_pm_addr_families_match(const struct sock *sk,
1059 const struct mptcp_addr_info *loc,
1060 const struct mptcp_addr_info *rem);
1061 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
1062 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side);
1063 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk);
1064 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk);
1065 void mptcp_pm_connection_closed(struct mptcp_sock *msk);
1066 void mptcp_pm_subflow_established(struct mptcp_sock *msk);
1067 bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk);
1068 void mptcp_pm_subflow_check_next(struct mptcp_sock *msk,
1069 const struct mptcp_subflow_context *subflow);
1070 void mptcp_pm_add_addr_received(const struct sock *ssk,
1071 const struct mptcp_addr_info *addr);
1072 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk,
1073 const struct mptcp_addr_info *addr);
1074 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk);
1075 void mptcp_pm_send_ack(struct mptcp_sock *msk,
1076 struct mptcp_subflow_context *subflow,
1077 bool prio, bool backup);
1078 void mptcp_pm_addr_send_ack(struct mptcp_sock *msk);
1079 void mptcp_pm_nl_rm_addr(struct mptcp_sock *msk, u8 rm_id);
1080 void mptcp_pm_rm_subflow(struct mptcp_sock *msk,
1081 const struct mptcp_rm_list *rm_list);
1082 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk,
1083 const struct mptcp_rm_list *rm_list);
1084 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup);
1085 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq);
1086 int mptcp_pm_mp_prio_send_ack(struct mptcp_sock *msk,
1087 struct mptcp_addr_info *addr,
1088 struct mptcp_addr_info *rem,
1089 u8 bkup);
1090 bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk,
1091 const struct mptcp_addr_info *addr);
1092 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk);
1093 struct mptcp_pm_add_entry *
1094 mptcp_pm_del_add_timer(struct mptcp_sock *msk,
1095 const struct mptcp_addr_info *addr, bool check_id);
1096 bool mptcp_lookup_subflow_by_saddr(const struct list_head *list,
1097 const struct mptcp_addr_info *saddr);
1098 bool mptcp_remove_anno_list_by_saddr(struct mptcp_sock *msk,
1099 const struct mptcp_addr_info *addr);
1100 int mptcp_pm_nl_set_flags(struct mptcp_pm_addr_entry *local,
1101 struct genl_info *info);
1102 int mptcp_userspace_pm_set_flags(struct mptcp_pm_addr_entry *local,
1103 struct genl_info *info);
1104 int mptcp_pm_announce_addr(struct mptcp_sock *msk,
1105 const struct mptcp_addr_info *addr,
1106 bool echo);
1107 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
1108 void mptcp_pm_remove_addr_entry(struct mptcp_sock *msk,
1109 struct mptcp_pm_addr_entry *entry);
1110
1111 /* the default path manager, used in mptcp_pm_unregister */
1112 extern struct mptcp_pm_ops mptcp_pm_kernel;
1113
1114 struct mptcp_pm_ops *mptcp_pm_find(const char *name);
1115 int mptcp_pm_register(struct mptcp_pm_ops *pm_ops);
1116 void mptcp_pm_unregister(struct mptcp_pm_ops *pm_ops);
1117 int mptcp_pm_validate(struct mptcp_pm_ops *pm_ops);
1118 void mptcp_pm_get_available(char *buf, size_t maxlen);
1119
1120 void mptcp_userspace_pm_free_local_addr_list(struct mptcp_sock *msk);
1121
1122 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk,
1123 const struct sock *ssk, gfp_t gfp);
1124 void mptcp_event_addr_announced(const struct sock *ssk, const struct mptcp_addr_info *info);
1125 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id);
1126 void mptcp_event_pm_listener(const struct sock *ssk,
1127 enum mptcp_event_type event);
1128 bool mptcp_userspace_pm_active(const struct mptcp_sock *msk);
1129
1130 void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subflow,
1131 struct request_sock *req);
1132 int mptcp_pm_genl_fill_addr(struct sk_buff *msg,
1133 struct netlink_callback *cb,
1134 struct mptcp_pm_addr_entry *entry);
1135
mptcp_pm_should_add_signal(struct mptcp_sock * msk)1136 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk)
1137 {
1138 return READ_ONCE(msk->pm.addr_signal) &
1139 (BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO));
1140 }
1141
mptcp_pm_should_add_signal_addr(struct mptcp_sock * msk)1142 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk)
1143 {
1144 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL);
1145 }
1146
mptcp_pm_should_add_signal_echo(struct mptcp_sock * msk)1147 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk)
1148 {
1149 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO);
1150 }
1151
mptcp_pm_should_rm_signal(struct mptcp_sock * msk)1152 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk)
1153 {
1154 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL);
1155 }
1156
mptcp_pm_is_userspace(const struct mptcp_sock * msk)1157 static inline bool mptcp_pm_is_userspace(const struct mptcp_sock *msk)
1158 {
1159 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_USERSPACE;
1160 }
1161
mptcp_pm_is_kernel(const struct mptcp_sock * msk)1162 static inline bool mptcp_pm_is_kernel(const struct mptcp_sock *msk)
1163 {
1164 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_KERNEL;
1165 }
1166
mptcp_add_addr_len(int family,bool echo,bool port)1167 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port)
1168 {
1169 u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE;
1170
1171 if (family == AF_INET6)
1172 len = TCPOLEN_MPTCP_ADD_ADDR6_BASE;
1173 if (!echo)
1174 len += MPTCPOPT_THMAC_LEN;
1175 /* account for 2 trailing 'nop' options */
1176 if (port)
1177 len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN;
1178
1179 return len;
1180 }
1181
mptcp_rm_addr_len(const struct mptcp_rm_list * rm_list)1182 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list)
1183 {
1184 if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX)
1185 return -EINVAL;
1186
1187 return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1;
1188 }
1189
1190 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, const struct sk_buff *skb,
1191 unsigned int opt_size, unsigned int remaining,
1192 struct mptcp_addr_info *addr, bool *echo,
1193 bool *drop_other_suboptions);
1194 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining,
1195 struct mptcp_rm_list *rm_list);
1196 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
1197 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk,
1198 struct mptcp_pm_addr_entry *skc);
1199 int mptcp_userspace_pm_get_local_id(struct mptcp_sock *msk,
1200 struct mptcp_pm_addr_entry *skc);
1201 bool mptcp_pm_is_backup(struct mptcp_sock *msk, struct sock_common *skc);
1202 bool mptcp_pm_nl_is_backup(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1203 bool mptcp_userspace_pm_is_backup(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1204 int mptcp_pm_nl_dump_addr(struct sk_buff *msg,
1205 struct netlink_callback *cb);
1206 int mptcp_userspace_pm_dump_addr(struct sk_buff *msg,
1207 struct netlink_callback *cb);
1208 int mptcp_pm_nl_get_addr(u8 id, struct mptcp_pm_addr_entry *addr,
1209 struct genl_info *info);
1210 int mptcp_userspace_pm_get_addr(u8 id, struct mptcp_pm_addr_entry *addr,
1211 struct genl_info *info);
1212
subflow_get_local_id(const struct mptcp_subflow_context * subflow)1213 static inline u8 subflow_get_local_id(const struct mptcp_subflow_context *subflow)
1214 {
1215 int local_id = READ_ONCE(subflow->local_id);
1216
1217 if (local_id < 0)
1218 return 0;
1219 return local_id;
1220 }
1221
1222 void __init mptcp_pm_kernel_register(void);
1223 void __init mptcp_pm_userspace_register(void);
1224 void __init mptcp_pm_nl_init(void);
1225 void mptcp_pm_worker(struct mptcp_sock *msk);
1226 void __mptcp_pm_kernel_worker(struct mptcp_sock *msk);
1227 u8 mptcp_pm_get_endp_signal_max(const struct mptcp_sock *msk);
1228 u8 mptcp_pm_get_endp_subflow_max(const struct mptcp_sock *msk);
1229 u8 mptcp_pm_get_endp_laminar_max(const struct mptcp_sock *msk);
1230 u8 mptcp_pm_get_endp_fullmesh_max(const struct mptcp_sock *msk);
1231 u8 mptcp_pm_get_limit_add_addr_accepted(const struct mptcp_sock *msk);
1232 u8 mptcp_pm_get_limit_extra_subflows(const struct mptcp_sock *msk);
1233
1234 /* called under PM lock */
__mptcp_pm_close_subflow(struct mptcp_sock * msk)1235 static inline void __mptcp_pm_close_subflow(struct mptcp_sock *msk)
1236 {
1237 if (--msk->pm.extra_subflows < mptcp_pm_get_limit_extra_subflows(msk))
1238 WRITE_ONCE(msk->pm.accept_subflow, true);
1239 }
1240
mptcp_pm_close_subflow(struct mptcp_sock * msk)1241 static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk)
1242 {
1243 spin_lock_bh(&msk->pm.lock);
1244 __mptcp_pm_close_subflow(msk);
1245 spin_unlock_bh(&msk->pm.lock);
1246 }
1247
mptcp_pm_add_addr_c_flag_case(struct mptcp_sock * msk)1248 static inline bool mptcp_pm_add_addr_c_flag_case(struct mptcp_sock *msk)
1249 {
1250 return READ_ONCE(msk->pm.remote_deny_join_id0) &&
1251 msk->pm.local_addr_used == 0 &&
1252 mptcp_pm_get_limit_add_addr_accepted(msk) == 0 &&
1253 msk->pm.extra_subflows < mptcp_pm_get_limit_extra_subflows(msk);
1254 }
1255
1256 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk);
1257
mptcp_get_ext(const struct sk_buff * skb)1258 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb)
1259 {
1260 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP);
1261 }
1262
1263 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops);
1264
__mptcp_check_fallback(const struct mptcp_sock * msk)1265 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk)
1266 {
1267 return test_bit(MPTCP_FALLBACK_DONE, &msk->flags);
1268 }
1269
mptcp_check_fallback(const struct sock * sk)1270 static inline bool mptcp_check_fallback(const struct sock *sk)
1271 {
1272 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1273 struct mptcp_sock *msk = mptcp_sk(subflow->conn);
1274
1275 return __mptcp_check_fallback(msk);
1276 }
1277
__mptcp_has_initial_subflow(const struct mptcp_sock * msk)1278 static inline bool __mptcp_has_initial_subflow(const struct mptcp_sock *msk)
1279 {
1280 struct sock *ssk = READ_ONCE(msk->first);
1281
1282 return ssk && ((1 << inet_sk_state_load(ssk)) &
1283 (TCPF_ESTABLISHED | TCPF_SYN_SENT |
1284 TCPF_SYN_RECV | TCPF_LISTEN));
1285 }
1286
1287 bool __mptcp_try_fallback(struct mptcp_sock *msk, int fb_mib);
1288
mptcp_try_fallback(struct sock * ssk,int fb_mib)1289 static inline bool mptcp_try_fallback(struct sock *ssk, int fb_mib)
1290 {
1291 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
1292 struct sock *sk = subflow->conn;
1293 struct mptcp_sock *msk;
1294
1295 msk = mptcp_sk(sk);
1296 if (!__mptcp_try_fallback(msk, fb_mib))
1297 return false;
1298 if (READ_ONCE(msk->snd_data_fin_enable) && !(ssk->sk_shutdown & SEND_SHUTDOWN)) {
1299 gfp_t saved_allocation = ssk->sk_allocation;
1300
1301 /* we are in a atomic (BH) scope, override ssk default for data
1302 * fin allocation
1303 */
1304 ssk->sk_allocation = GFP_ATOMIC;
1305 ssk->sk_shutdown |= SEND_SHUTDOWN;
1306 tcp_shutdown(ssk, SEND_SHUTDOWN);
1307 ssk->sk_allocation = saved_allocation;
1308 }
1309 return true;
1310 }
1311
mptcp_early_fallback(struct mptcp_sock * msk,struct mptcp_subflow_context * subflow,int fb_mib)1312 static inline void mptcp_early_fallback(struct mptcp_sock *msk,
1313 struct mptcp_subflow_context *subflow,
1314 int fb_mib)
1315 {
1316 subflow->request_mptcp = 0;
1317 WARN_ON_ONCE(!__mptcp_try_fallback(msk, fb_mib));
1318 }
1319
mptcp_check_infinite_map(struct sk_buff * skb)1320 static inline bool mptcp_check_infinite_map(struct sk_buff *skb)
1321 {
1322 struct mptcp_ext *mpext;
1323
1324 mpext = skb ? mptcp_get_ext(skb) : NULL;
1325 if (mpext && mpext->infinite_map)
1326 return true;
1327
1328 return false;
1329 }
1330
is_active_ssk(struct mptcp_subflow_context * subflow)1331 static inline bool is_active_ssk(struct mptcp_subflow_context *subflow)
1332 {
1333 return (subflow->request_mptcp || subflow->request_join);
1334 }
1335
subflow_simultaneous_connect(struct sock * sk)1336 static inline bool subflow_simultaneous_connect(struct sock *sk)
1337 {
1338 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1339
1340 return (1 << sk->sk_state) &
1341 (TCPF_ESTABLISHED | TCPF_FIN_WAIT1 | TCPF_FIN_WAIT2 | TCPF_CLOSING) &&
1342 is_active_ssk(subflow) &&
1343 !subflow->conn_finished;
1344 }
1345
1346 #ifdef CONFIG_SYN_COOKIES
1347 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1348 struct sk_buff *skb);
1349 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1350 struct sk_buff *skb);
1351 void __init mptcp_join_cookie_init(void);
1352 #else
1353 static inline void
subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock * subflow_req,struct sk_buff * skb)1354 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1355 struct sk_buff *skb) {}
1356 static inline bool
mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock * subflow_req,struct sk_buff * skb)1357 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1358 struct sk_buff *skb)
1359 {
1360 return false;
1361 }
1362
mptcp_join_cookie_init(void)1363 static inline void mptcp_join_cookie_init(void) {}
1364 #endif
1365
1366 #endif /* __MPTCP_PROTOCOL_H */
1367