]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blob - net/mptcp/protocol.h
projects / mirror_ubuntu-jammy-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge branch 'rework/printk_safe-removal' into for-linus
[mirror_ubuntu-jammy-kernel.git] / net / mptcp / protocol.h
1 /* SPDX-License-Identifier: GPL-2.0 */
2 /* Multipath TCP
3 *
4 * Copyright (c) 2017 - 2019, Intel Corporation.
5 */
6
7 #ifndef __MPTCP_PROTOCOL_H
8 #define __MPTCP_PROTOCOL_H
9
10 #include <linux/random.h>
11 #include <net/tcp.h>
12 #include <net/inet_connection_sock.h>
13 #include <uapi/linux/mptcp.h>
14
15 #define MPTCP_SUPPORTED_VERSION 1
16
17 /* MPTCP option bits */
18 #define OPTION_MPTCP_MPC_SYN BIT(0)
19 #define OPTION_MPTCP_MPC_SYNACK BIT(1)
20 #define OPTION_MPTCP_MPC_ACK BIT(2)
21 #define OPTION_MPTCP_MPJ_SYN BIT(3)
22 #define OPTION_MPTCP_MPJ_SYNACK BIT(4)
23 #define OPTION_MPTCP_MPJ_ACK BIT(5)
24 #define OPTION_MPTCP_ADD_ADDR BIT(6)
25 #define OPTION_MPTCP_RM_ADDR BIT(7)
26 #define OPTION_MPTCP_FASTCLOSE BIT(8)
27 #define OPTION_MPTCP_PRIO BIT(9)
28 #define OPTION_MPTCP_RST BIT(10)
29
30 /* MPTCP option subtypes */
31 #define MPTCPOPT_MP_CAPABLE 0
32 #define MPTCPOPT_MP_JOIN 1
33 #define MPTCPOPT_DSS 2
34 #define MPTCPOPT_ADD_ADDR 3
35 #define MPTCPOPT_RM_ADDR 4
36 #define MPTCPOPT_MP_PRIO 5
37 #define MPTCPOPT_MP_FAIL 6
38 #define MPTCPOPT_MP_FASTCLOSE 7
39 #define MPTCPOPT_RST 8
40
41 /* MPTCP suboption lengths */
42 #define TCPOLEN_MPTCP_MPC_SYN 4
43 #define TCPOLEN_MPTCP_MPC_SYNACK 12
44 #define TCPOLEN_MPTCP_MPC_ACK 20
45 #define TCPOLEN_MPTCP_MPC_ACK_DATA 22
46 #define TCPOLEN_MPTCP_MPJ_SYN 12
47 #define TCPOLEN_MPTCP_MPJ_SYNACK 16
48 #define TCPOLEN_MPTCP_MPJ_ACK 24
49 #define TCPOLEN_MPTCP_DSS_BASE 4
50 #define TCPOLEN_MPTCP_DSS_ACK32 4
51 #define TCPOLEN_MPTCP_DSS_ACK64 8
52 #define TCPOLEN_MPTCP_DSS_MAP32 10
53 #define TCPOLEN_MPTCP_DSS_MAP64 14
54 #define TCPOLEN_MPTCP_DSS_CHECKSUM 2
55 #define TCPOLEN_MPTCP_ADD_ADDR 16
56 #define TCPOLEN_MPTCP_ADD_ADDR_PORT 18
57 #define TCPOLEN_MPTCP_ADD_ADDR_BASE 8
58 #define TCPOLEN_MPTCP_ADD_ADDR_BASE_PORT 10
59 #define TCPOLEN_MPTCP_ADD_ADDR6 28
60 #define TCPOLEN_MPTCP_ADD_ADDR6_PORT 30
61 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE 20
62 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE_PORT 22
63 #define TCPOLEN_MPTCP_PORT_LEN 2
64 #define TCPOLEN_MPTCP_PORT_ALIGN 2
65 #define TCPOLEN_MPTCP_RM_ADDR_BASE 3
66 #define TCPOLEN_MPTCP_PRIO 3
67 #define TCPOLEN_MPTCP_PRIO_ALIGN 4
68 #define TCPOLEN_MPTCP_FASTCLOSE 12
69 #define TCPOLEN_MPTCP_RST 4
70
71 #define TCPOLEN_MPTCP_MPC_ACK_DATA_CSUM (TCPOLEN_MPTCP_DSS_CHECKSUM + TCPOLEN_MPTCP_MPC_ACK_DATA)
72
73 /* MPTCP MP_JOIN flags */
74 #define MPTCPOPT_BACKUP BIT(0)
75 #define MPTCPOPT_HMAC_LEN 20
76 #define MPTCPOPT_THMAC_LEN 8
77
78 /* MPTCP MP_CAPABLE flags */
79 #define MPTCP_VERSION_MASK (0x0F)
80 #define MPTCP_CAP_CHECKSUM_REQD BIT(7)
81 #define MPTCP_CAP_EXTENSIBILITY BIT(6)
82 #define MPTCP_CAP_DENY_JOIN_ID0 BIT(5)
83 #define MPTCP_CAP_HMAC_SHA256 BIT(0)
84 #define MPTCP_CAP_FLAG_MASK (0x1F)
85
86 /* MPTCP DSS flags */
87 #define MPTCP_DSS_DATA_FIN BIT(4)
88 #define MPTCP_DSS_DSN64 BIT(3)
89 #define MPTCP_DSS_HAS_MAP BIT(2)
90 #define MPTCP_DSS_ACK64 BIT(1)
91 #define MPTCP_DSS_HAS_ACK BIT(0)
92 #define MPTCP_DSS_FLAG_MASK (0x1F)
93
94 /* MPTCP ADD_ADDR flags */
95 #define MPTCP_ADDR_ECHO BIT(0)
96
97 /* MPTCP MP_PRIO flags */
98 #define MPTCP_PRIO_BKUP BIT(0)
99
100 /* MPTCP TCPRST flags */
101 #define MPTCP_RST_TRANSIENT BIT(0)
102
103 /* MPTCP socket flags */
104 #define MPTCP_DATA_READY 0
105 #define MPTCP_NOSPACE 1
106 #define MPTCP_WORK_RTX 2
107 #define MPTCP_WORK_EOF 3
108 #define MPTCP_FALLBACK_DONE 4
109 #define MPTCP_WORK_CLOSE_SUBFLOW 5
110 #define MPTCP_PUSH_PENDING 6
111 #define MPTCP_CLEAN_UNA 7
112 #define MPTCP_ERROR_REPORT 8
113 #define MPTCP_RETRANSMIT 9
114 #define MPTCP_WORK_SYNC_SETSOCKOPT 10
115 #define MPTCP_CONNECTED 11
116
117 static inline bool before64(__u64 seq1, __u64 seq2)
118 {
119 return (__s64)(seq1 - seq2) < 0;
120 }
121
122 #define after64(seq2, seq1) before64(seq1, seq2)
123
124 struct mptcp_options_received {
125 u64 sndr_key;
126 u64 rcvr_key;
127 u64 data_ack;
128 u64 data_seq;
129 u32 subflow_seq;
130 u16 data_len;
131 __sum16 csum;
132 u16 mp_capable : 1,
133 mp_join : 1,
134 fastclose : 1,
135 reset : 1,
136 dss : 1,
137 add_addr : 1,
138 rm_addr : 1,
139 mp_prio : 1,
140 echo : 1,
141 csum_reqd : 1,
142 backup : 1,
143 deny_join_id0 : 1;
144 u32 token;
145 u32 nonce;
146 u64 thmac;
147 u8 hmac[MPTCPOPT_HMAC_LEN];
148 u8 join_id;
149 u8 use_map:1,
150 dsn64:1,
151 data_fin:1,
152 use_ack:1,
153 ack64:1,
154 mpc_map:1,
155 __unused:2;
156 struct mptcp_addr_info addr;
157 struct mptcp_rm_list rm_list;
158 u64 ahmac;
159 u8 reset_reason:4;
160 u8 reset_transient:1;
161 };
162
163 static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field)
164 {
165 return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) |
166 ((nib & 0xF) << 8) | field);
167 }
168
169 enum mptcp_pm_status {
170 MPTCP_PM_ADD_ADDR_RECEIVED,
171 MPTCP_PM_ADD_ADDR_SEND_ACK,
172 MPTCP_PM_RM_ADDR_RECEIVED,
173 MPTCP_PM_ESTABLISHED,
174 MPTCP_PM_ALREADY_ESTABLISHED, /* persistent status, set after ESTABLISHED event */
175 MPTCP_PM_SUBFLOW_ESTABLISHED,
176 };
177
178 enum mptcp_addr_signal_status {
179 MPTCP_ADD_ADDR_SIGNAL,
180 MPTCP_ADD_ADDR_ECHO,
181 MPTCP_ADD_ADDR_IPV6,
182 MPTCP_ADD_ADDR_PORT,
183 MPTCP_RM_ADDR_SIGNAL,
184 };
185
186 struct mptcp_pm_data {
187 struct mptcp_addr_info local;
188 struct mptcp_addr_info remote;
189 struct list_head anno_list;
190
191 spinlock_t lock; /*protects the whole PM data */
192
193 u8 addr_signal;
194 bool server_side;
195 bool work_pending;
196 bool accept_addr;
197 bool accept_subflow;
198 bool remote_deny_join_id0;
199 u8 add_addr_signaled;
200 u8 add_addr_accepted;
201 u8 local_addr_used;
202 u8 subflows;
203 u8 status;
204 struct mptcp_rm_list rm_list_tx;
205 struct mptcp_rm_list rm_list_rx;
206 };
207
208 struct mptcp_data_frag {
209 struct list_head list;
210 u64 data_seq;
211 u16 data_len;
212 u16 offset;
213 u16 overhead;
214 u16 already_sent;
215 struct page *page;
216 };
217
218 /* MPTCP connection sock */
219 struct mptcp_sock {
220 /* inet_connection_sock must be the first member */
221 struct inet_connection_sock sk;
222 u64 local_key;
223 u64 remote_key;
224 u64 write_seq;
225 u64 snd_nxt;
226 u64 ack_seq;
227 u64 rcv_wnd_sent;
228 u64 rcv_data_fin_seq;
229 int wmem_reserved;
230 struct sock *last_snd;
231 int snd_burst;
232 int old_wspace;
233 u64 snd_una;
234 u64 wnd_end;
235 unsigned long timer_ival;
236 u32 token;
237 int rmem_released;
238 unsigned long flags;
239 bool can_ack;
240 bool fully_established;
241 bool rcv_data_fin;
242 bool snd_data_fin_enable;
243 bool rcv_fastclose;
244 bool use_64bit_ack; /* Set when we received a 64-bit DSN */
245 bool csum_enabled;
246 spinlock_t join_list_lock;
247 struct work_struct work;
248 struct sk_buff *ooo_last_skb;
249 struct rb_root out_of_order_queue;
250 struct sk_buff_head receive_queue;
251 int tx_pending_data;
252 struct list_head conn_list;
253 struct list_head rtx_queue;
254 struct mptcp_data_frag *first_pending;
255 struct list_head join_list;
256 struct socket *subflow; /* outgoing connect/listener/!mp_capable */
257 struct sock *first;
258 struct mptcp_pm_data pm;
259 struct {
260 u32 space; /* bytes copied in last measurement window */
261 u32 copied; /* bytes copied in this measurement window */
262 u64 time; /* start time of measurement window */
263 u64 rtt_us; /* last maximum rtt of subflows */
264 } rcvq_space;
265
266 u32 setsockopt_seq;
267 char ca_name[TCP_CA_NAME_MAX];
268 };
269
270 #define mptcp_lock_sock(___sk, cb) do { \
271 struct sock *__sk = (___sk); /* silence macro reuse warning */ \
272 might_sleep(); \
273 spin_lock_bh(&__sk->sk_lock.slock); \
274 if (__sk->sk_lock.owned) \
275 __lock_sock(__sk); \
276 cb; \
277 __sk->sk_lock.owned = 1; \
278 spin_unlock(&__sk->sk_lock.slock); \
279 mutex_acquire(&__sk->sk_lock.dep_map, 0, 0, _RET_IP_); \
280 local_bh_enable(); \
281 } while (0)
282
283 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock)
284 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock)
285
286 #define mptcp_for_each_subflow(__msk, __subflow) \
287 list_for_each_entry(__subflow, &((__msk)->conn_list), node)
288
289 static inline void msk_owned_by_me(const struct mptcp_sock *msk)
290 {
291 sock_owned_by_me((const struct sock *)msk);
292 }
293
294 static inline struct mptcp_sock *mptcp_sk(const struct sock *sk)
295 {
296 return (struct mptcp_sock *)sk;
297 }
298
299 static inline int __mptcp_space(const struct sock *sk)
300 {
301 return tcp_space(sk) + READ_ONCE(mptcp_sk(sk)->rmem_released);
302 }
303
304 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk)
305 {
306 const struct mptcp_sock *msk = mptcp_sk(sk);
307
308 return READ_ONCE(msk->first_pending);
309 }
310
311 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk)
312 {
313 struct mptcp_sock *msk = mptcp_sk(sk);
314 struct mptcp_data_frag *cur;
315
316 cur = msk->first_pending;
317 return list_is_last(&cur->list, &msk->rtx_queue) ? NULL :
318 list_next_entry(cur, list);
319 }
320
321 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk)
322 {
323 struct mptcp_sock *msk = mptcp_sk(sk);
324
325 if (!msk->first_pending)
326 return NULL;
327
328 if (WARN_ON_ONCE(list_empty(&msk->rtx_queue)))
329 return NULL;
330
331 return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list);
332 }
333
334 static inline struct mptcp_data_frag *mptcp_rtx_head(const struct sock *sk)
335 {
336 struct mptcp_sock *msk = mptcp_sk(sk);
337
338 if (msk->snd_una == READ_ONCE(msk->snd_nxt))
339 return NULL;
340
341 return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list);
342 }
343
344 struct csum_pseudo_header {
345 __be64 data_seq;
346 __be32 subflow_seq;
347 __be16 data_len;
348 __sum16 csum;
349 };
350
351 struct mptcp_subflow_request_sock {
352 struct tcp_request_sock sk;
353 u16 mp_capable : 1,
354 mp_join : 1,
355 backup : 1,
356 csum_reqd : 1,
357 allow_join_id0 : 1;
358 u8 local_id;
359 u8 remote_id;
360 u64 local_key;
361 u64 idsn;
362 u32 token;
363 u32 ssn_offset;
364 u64 thmac;
365 u32 local_nonce;
366 u32 remote_nonce;
367 struct mptcp_sock *msk;
368 struct hlist_nulls_node token_node;
369 };
370
371 static inline struct mptcp_subflow_request_sock *
372 mptcp_subflow_rsk(const struct request_sock *rsk)
373 {
374 return (struct mptcp_subflow_request_sock *)rsk;
375 }
376
377 enum mptcp_data_avail {
378 MPTCP_SUBFLOW_NODATA,
379 MPTCP_SUBFLOW_DATA_AVAIL,
380 };
381
382 struct mptcp_delegated_action {
383 struct napi_struct napi;
384 struct list_head head;
385 };
386
387 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions);
388
389 #define MPTCP_DELEGATE_SEND 0
390
391 /* MPTCP subflow context */
392 struct mptcp_subflow_context {
393 struct list_head node;/* conn_list of subflows */
394 u64 local_key;
395 u64 remote_key;
396 u64 idsn;
397 u64 map_seq;
398 u32 snd_isn;
399 u32 token;
400 u32 rel_write_seq;
401 u32 map_subflow_seq;
402 u32 ssn_offset;
403 u32 map_data_len;
404 __wsum map_data_csum;
405 u32 map_csum_len;
406 u32 request_mptcp : 1, /* send MP_CAPABLE */
407 request_join : 1, /* send MP_JOIN */
408 request_bkup : 1,
409 mp_capable : 1, /* remote is MPTCP capable */
410 mp_join : 1, /* remote is JOINing */
411 fully_established : 1, /* path validated */
412 pm_notified : 1, /* PM hook called for established status */
413 conn_finished : 1,
414 map_valid : 1,
415 map_csum_reqd : 1,
416 map_data_fin : 1,
417 mpc_map : 1,
418 backup : 1,
419 send_mp_prio : 1,
420 rx_eof : 1,
421 can_ack : 1, /* only after processing the remote a key */
422 disposable : 1; /* ctx can be free at ulp release time */
423 enum mptcp_data_avail data_avail;
424 u32 remote_nonce;
425 u64 thmac;
426 u32 local_nonce;
427 u32 remote_token;
428 u8 hmac[MPTCPOPT_HMAC_LEN];
429 u8 local_id;
430 u8 remote_id;
431 u8 reset_seen:1;
432 u8 reset_transient:1;
433 u8 reset_reason:4;
434
435 long delegated_status;
436 struct list_head delegated_node; /* link into delegated_action, protected by local BH */
437
438 u32 setsockopt_seq;
439
440 struct sock *tcp_sock; /* tcp sk backpointer */
441 struct sock *conn; /* parent mptcp_sock */
442 const struct inet_connection_sock_af_ops *icsk_af_ops;
443 void (*tcp_data_ready)(struct sock *sk);
444 void (*tcp_state_change)(struct sock *sk);
445 void (*tcp_write_space)(struct sock *sk);
446 void (*tcp_error_report)(struct sock *sk);
447
448 struct rcu_head rcu;
449 };
450
451 static inline struct mptcp_subflow_context *
452 mptcp_subflow_ctx(const struct sock *sk)
453 {
454 struct inet_connection_sock *icsk = inet_csk(sk);
455
456 /* Use RCU on icsk_ulp_data only for sock diag code */
457 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data;
458 }
459
460 static inline struct sock *
461 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow)
462 {
463 return subflow->tcp_sock;
464 }
465
466 static inline u64
467 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow)
468 {
469 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq -
470 subflow->ssn_offset -
471 subflow->map_subflow_seq;
472 }
473
474 static inline u64
475 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow)
476 {
477 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow);
478 }
479
480 static inline void mptcp_add_pending_subflow(struct mptcp_sock *msk,
481 struct mptcp_subflow_context *subflow)
482 {
483 sock_hold(mptcp_subflow_tcp_sock(subflow));
484 spin_lock_bh(&msk->join_list_lock);
485 list_add_tail(&subflow->node, &msk->join_list);
486 spin_unlock_bh(&msk->join_list_lock);
487 }
488
489 void mptcp_subflow_process_delegated(struct sock *ssk);
490
491 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow)
492 {
493 struct mptcp_delegated_action *delegated;
494 bool schedule;
495
496 /* The implied barrier pairs with mptcp_subflow_delegated_done(), and
497 * ensures the below list check sees list updates done prior to status
498 * bit changes
499 */
500 if (!test_and_set_bit(MPTCP_DELEGATE_SEND, &subflow->delegated_status)) {
501 /* still on delegated list from previous scheduling */
502 if (!list_empty(&subflow->delegated_node))
503 return;
504
505 /* the caller held the subflow bh socket lock */
506 lockdep_assert_in_softirq();
507
508 delegated = this_cpu_ptr(&mptcp_delegated_actions);
509 schedule = list_empty(&delegated->head);
510 list_add_tail(&subflow->delegated_node, &delegated->head);
511 sock_hold(mptcp_subflow_tcp_sock(subflow));
512 if (schedule)
513 napi_schedule(&delegated->napi);
514 }
515 }
516
517 static inline struct mptcp_subflow_context *
518 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated)
519 {
520 struct mptcp_subflow_context *ret;
521
522 if (list_empty(&delegated->head))
523 return NULL;
524
525 ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node);
526 list_del_init(&ret->delegated_node);
527 return ret;
528 }
529
530 static inline bool mptcp_subflow_has_delegated_action(const struct mptcp_subflow_context *subflow)
531 {
532 return test_bit(MPTCP_DELEGATE_SEND, &subflow->delegated_status);
533 }
534
535 static inline void mptcp_subflow_delegated_done(struct mptcp_subflow_context *subflow)
536 {
537 /* pairs with mptcp_subflow_delegate, ensures delegate_node is updated before
538 * touching the status bit
539 */
540 smp_wmb();
541 clear_bit(MPTCP_DELEGATE_SEND, &subflow->delegated_status);
542 }
543
544 int mptcp_is_enabled(struct net *net);
545 unsigned int mptcp_get_add_addr_timeout(struct net *net);
546 int mptcp_is_checksum_enabled(struct net *net);
547 int mptcp_allow_join_id0(struct net *net);
548 void mptcp_subflow_fully_established(struct mptcp_subflow_context *subflow,
549 struct mptcp_options_received *mp_opt);
550 bool mptcp_subflow_data_available(struct sock *sk);
551 void __init mptcp_subflow_init(void);
552 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how);
553 void mptcp_close_ssk(struct sock *sk, struct sock *ssk,
554 struct mptcp_subflow_context *subflow);
555 void mptcp_subflow_reset(struct sock *ssk);
556 void mptcp_sock_graft(struct sock *sk, struct socket *parent);
557 struct socket *__mptcp_nmpc_socket(const struct mptcp_sock *msk);
558
559 /* called with sk socket lock held */
560 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc,
561 const struct mptcp_addr_info *remote,
562 u8 flags, int ifindex);
563 int mptcp_subflow_create_socket(struct sock *sk, struct socket **new_sock);
564 void mptcp_info2sockaddr(const struct mptcp_addr_info *info,
565 struct sockaddr_storage *addr,
566 unsigned short family);
567
568 static inline bool mptcp_subflow_active(struct mptcp_subflow_context *subflow)
569 {
570 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
571
572 /* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */
573 if (subflow->request_join && !subflow->fully_established)
574 return false;
575
576 /* only send if our side has not closed yet */
577 return ((1 << ssk->sk_state) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT));
578 }
579
580 static inline void mptcp_subflow_tcp_fallback(struct sock *sk,
581 struct mptcp_subflow_context *ctx)
582 {
583 sk->sk_data_ready = ctx->tcp_data_ready;
584 sk->sk_state_change = ctx->tcp_state_change;
585 sk->sk_write_space = ctx->tcp_write_space;
586 sk->sk_error_report = ctx->tcp_error_report;
587
588 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops;
589 }
590
591 void __init mptcp_proto_init(void);
592 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
593 int __init mptcp_proto_v6_init(void);
594 #endif
595
596 struct sock *mptcp_sk_clone(const struct sock *sk,
597 const struct mptcp_options_received *mp_opt,
598 struct request_sock *req);
599 void mptcp_get_options(const struct sock *sk,
600 const struct sk_buff *skb,
601 struct mptcp_options_received *mp_opt);
602
603 void mptcp_finish_connect(struct sock *sk);
604 void __mptcp_set_connected(struct sock *sk);
605 static inline bool mptcp_is_fully_established(struct sock *sk)
606 {
607 return inet_sk_state_load(sk) == TCP_ESTABLISHED &&
608 READ_ONCE(mptcp_sk(sk)->fully_established);
609 }
610 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk);
611 void mptcp_data_ready(struct sock *sk, struct sock *ssk);
612 bool mptcp_finish_join(struct sock *sk);
613 bool mptcp_schedule_work(struct sock *sk);
614 int mptcp_setsockopt(struct sock *sk, int level, int optname,
615 sockptr_t optval, unsigned int optlen);
616 int mptcp_getsockopt(struct sock *sk, int level, int optname,
617 char __user *optval, int __user *option);
618
619 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq);
620 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit)
621 {
622 if (use_64bit)
623 return cur_seq;
624
625 return __mptcp_expand_seq(old_seq, cur_seq);
626 }
627 void __mptcp_check_push(struct sock *sk, struct sock *ssk);
628 void __mptcp_data_acked(struct sock *sk);
629 void __mptcp_error_report(struct sock *sk);
630 void mptcp_subflow_eof(struct sock *sk);
631 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit);
632 void __mptcp_flush_join_list(struct mptcp_sock *msk);
633 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk)
634 {
635 return READ_ONCE(msk->snd_data_fin_enable) &&
636 READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt);
637 }
638
639 static inline bool mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
640 {
641 if ((sk->sk_userlocks & SOCK_SNDBUF_LOCK) || ssk->sk_sndbuf <= READ_ONCE(sk->sk_sndbuf))
642 return false;
643
644 WRITE_ONCE(sk->sk_sndbuf, ssk->sk_sndbuf);
645 return true;
646 }
647
648 static inline void mptcp_write_space(struct sock *sk)
649 {
650 if (sk_stream_is_writeable(sk)) {
651 /* pairs with memory barrier in mptcp_poll */
652 smp_mb();
653 if (test_and_clear_bit(MPTCP_NOSPACE, &mptcp_sk(sk)->flags))
654 sk_stream_write_space(sk);
655 }
656 }
657
658 void mptcp_destroy_common(struct mptcp_sock *msk);
659
660 #define MPTCP_TOKEN_MAX_RETRIES 4
661
662 void __init mptcp_token_init(void);
663 static inline void mptcp_token_init_request(struct request_sock *req)
664 {
665 mptcp_subflow_rsk(req)->token_node.pprev = NULL;
666 }
667
668 int mptcp_token_new_request(struct request_sock *req);
669 void mptcp_token_destroy_request(struct request_sock *req);
670 int mptcp_token_new_connect(struct sock *sk);
671 void mptcp_token_accept(struct mptcp_subflow_request_sock *r,
672 struct mptcp_sock *msk);
673 bool mptcp_token_exists(u32 token);
674 struct mptcp_sock *mptcp_token_get_sock(u32 token);
675 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot,
676 long *s_num);
677 void mptcp_token_destroy(struct mptcp_sock *msk);
678
679 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn);
680
681 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac);
682
683 void __init mptcp_pm_init(void);
684 void mptcp_pm_data_init(struct mptcp_sock *msk);
685 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side);
686 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk, gfp_t gfp);
687 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk);
688 void mptcp_pm_connection_closed(struct mptcp_sock *msk);
689 void mptcp_pm_subflow_established(struct mptcp_sock *msk);
690 void mptcp_pm_subflow_closed(struct mptcp_sock *msk, u8 id);
691 void mptcp_pm_add_addr_received(struct mptcp_sock *msk,
692 const struct mptcp_addr_info *addr);
693 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk,
694 struct mptcp_addr_info *addr);
695 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk);
696 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk);
697 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk,
698 const struct mptcp_rm_list *rm_list);
699 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup);
700 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk,
701 struct mptcp_addr_info *addr,
702 u8 bkup);
703 void mptcp_pm_free_anno_list(struct mptcp_sock *msk);
704 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk);
705 struct mptcp_pm_add_entry *
706 mptcp_pm_del_add_timer(struct mptcp_sock *msk,
707 struct mptcp_addr_info *addr, bool check_id);
708 struct mptcp_pm_add_entry *
709 mptcp_lookup_anno_list_by_saddr(struct mptcp_sock *msk,
710 struct mptcp_addr_info *addr);
711
712 int mptcp_pm_announce_addr(struct mptcp_sock *msk,
713 const struct mptcp_addr_info *addr,
714 bool echo);
715 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
716 int mptcp_pm_remove_subflow(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
717
718 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk,
719 const struct sock *ssk, gfp_t gfp);
720 void mptcp_event_addr_announced(const struct mptcp_sock *msk, const struct mptcp_addr_info *info);
721 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id);
722
723 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk)
724 {
725 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL);
726 }
727
728 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk)
729 {
730 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO);
731 }
732
733 static inline bool mptcp_pm_should_add_signal_ipv6(struct mptcp_sock *msk)
734 {
735 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_IPV6);
736 }
737
738 static inline bool mptcp_pm_should_add_signal_port(struct mptcp_sock *msk)
739 {
740 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_PORT);
741 }
742
743 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk)
744 {
745 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL);
746 }
747
748 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port)
749 {
750 u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE;
751
752 if (family == AF_INET6)
753 len = TCPOLEN_MPTCP_ADD_ADDR6_BASE;
754 if (!echo)
755 len += MPTCPOPT_THMAC_LEN;
756 /* account for 2 trailing 'nop' options */
757 if (port)
758 len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN;
759
760 return len;
761 }
762
763 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list)
764 {
765 if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX)
766 return -EINVAL;
767
768 return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1;
769 }
770
771 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, unsigned int remaining,
772 struct mptcp_addr_info *saddr, bool *echo, bool *port);
773 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining,
774 struct mptcp_rm_list *rm_list);
775 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
776
777 void __init mptcp_pm_nl_init(void);
778 void mptcp_pm_nl_data_init(struct mptcp_sock *msk);
779 void mptcp_pm_nl_work(struct mptcp_sock *msk);
780 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk,
781 const struct mptcp_rm_list *rm_list);
782 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
783 unsigned int mptcp_pm_get_add_addr_signal_max(struct mptcp_sock *msk);
784 unsigned int mptcp_pm_get_add_addr_accept_max(struct mptcp_sock *msk);
785 unsigned int mptcp_pm_get_subflows_max(struct mptcp_sock *msk);
786 unsigned int mptcp_pm_get_local_addr_max(struct mptcp_sock *msk);
787
788 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk);
789 void mptcp_sockopt_sync_all(struct mptcp_sock *msk);
790
791 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb)
792 {
793 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP);
794 }
795
796 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops);
797
798 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk)
799 {
800 return test_bit(MPTCP_FALLBACK_DONE, &msk->flags);
801 }
802
803 static inline bool mptcp_check_fallback(const struct sock *sk)
804 {
805 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
806 struct mptcp_sock *msk = mptcp_sk(subflow->conn);
807
808 return __mptcp_check_fallback(msk);
809 }
810
811 static inline void __mptcp_do_fallback(struct mptcp_sock *msk)
812 {
813 if (test_bit(MPTCP_FALLBACK_DONE, &msk->flags)) {
814 pr_debug("TCP fallback already done (msk=%p)", msk);
815 return;
816 }
817 set_bit(MPTCP_FALLBACK_DONE, &msk->flags);
818 }
819
820 static inline void mptcp_do_fallback(struct sock *sk)
821 {
822 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
823 struct mptcp_sock *msk = mptcp_sk(subflow->conn);
824
825 __mptcp_do_fallback(msk);
826 }
827
828 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a)
829
830 static inline bool subflow_simultaneous_connect(struct sock *sk)
831 {
832 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
833 struct sock *parent = subflow->conn;
834
835 return sk->sk_state == TCP_ESTABLISHED &&
836 !mptcp_sk(parent)->pm.server_side &&
837 !subflow->conn_finished;
838 }
839
840 #ifdef CONFIG_SYN_COOKIES
841 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
842 struct sk_buff *skb);
843 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
844 struct sk_buff *skb);
845 void __init mptcp_join_cookie_init(void);
846 #else
847 static inline void
848 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
849 struct sk_buff *skb) {}
850 static inline bool
851 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
852 struct sk_buff *skb)
853 {
854 return false;
855 }
856
857 static inline void mptcp_join_cookie_init(void) {}
858 #endif
859
860 #endif /* __MPTCP_PROTOCOL_H */
Ubuntu Jammy Linux kernel mirror