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:1;
133 };
134
135 #define MPTCP_SKB_CB(__skb) ((struct mptcp_skb_cb *)&((__skb)->cb[0]))
136
before64(__u64 seq1,__u64 seq2)137 static inline bool before64(__u64 seq1, __u64 seq2)
138 {
139 return (__s64)(seq1 - seq2) < 0;
140 }
141
142 #define after64(seq2, seq1) before64(seq1, seq2)
143
144 struct mptcp_options_received {
145 u64 sndr_key;
146 u64 rcvr_key;
147 u64 data_ack;
148 u64 data_seq;
149 u32 subflow_seq;
150 u16 data_len;
151 __sum16 csum;
152 struct_group(status,
153 u16 suboptions;
154 u16 use_map:1,
155 dsn64:1,
156 data_fin:1,
157 use_ack:1,
158 ack64:1,
159 mpc_map:1,
160 reset_reason:4,
161 reset_transient:1,
162 echo:1,
163 backup:1,
164 deny_join_id0:1,
165 __unused:2;
166 );
167 u8 join_id;
168 u32 token;
169 u32 nonce;
170 u64 thmac;
171 u8 hmac[MPTCPOPT_HMAC_LEN];
172 struct mptcp_addr_info addr;
173 struct mptcp_rm_list rm_list;
174 u64 ahmac;
175 u64 fail_seq;
176 };
177
mptcp_option(u8 subopt,u8 len,u8 nib,u8 field)178 static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field)
179 {
180 return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) |
181 ((nib & 0xF) << 8) | field);
182 }
183
184 enum mptcp_pm_status {
185 MPTCP_PM_ADD_ADDR_RECEIVED,
186 MPTCP_PM_ADD_ADDR_SEND_ACK,
187 MPTCP_PM_RM_ADDR_RECEIVED,
188 MPTCP_PM_ESTABLISHED,
189 MPTCP_PM_SUBFLOW_ESTABLISHED,
190 MPTCP_PM_ALREADY_ESTABLISHED, /* persistent status, set after ESTABLISHED event */
191 MPTCP_PM_MPC_ENDPOINT_ACCOUNTED /* persistent status, set after MPC local address is
192 * accounted int id_avail_bitmap
193 */
194 };
195
196 enum mptcp_pm_type {
197 MPTCP_PM_TYPE_KERNEL = 0,
198 MPTCP_PM_TYPE_USERSPACE,
199
200 __MPTCP_PM_TYPE_NR,
201 __MPTCP_PM_TYPE_MAX = __MPTCP_PM_TYPE_NR - 1,
202 };
203
204 /* Status bits below MPTCP_PM_ALREADY_ESTABLISHED need pm worker actions */
205 #define MPTCP_PM_WORK_MASK ((1 << MPTCP_PM_ALREADY_ESTABLISHED) - 1)
206
207 enum mptcp_addr_signal_status {
208 MPTCP_ADD_ADDR_SIGNAL,
209 MPTCP_ADD_ADDR_ECHO,
210 MPTCP_RM_ADDR_SIGNAL,
211 };
212
213 /* max value of mptcp_addr_info.id */
214 #define MPTCP_PM_MAX_ADDR_ID U8_MAX
215
216 struct mptcp_pm_data {
217 struct mptcp_addr_info local;
218 struct mptcp_addr_info remote;
219 struct list_head anno_list;
220 struct list_head userspace_pm_local_addr_list;
221
222 spinlock_t lock; /*protects the whole PM data */
223
224 u8 addr_signal;
225 bool server_side;
226 bool work_pending;
227 bool accept_addr;
228 bool accept_subflow;
229 bool remote_deny_join_id0;
230 u8 add_addr_signaled;
231 u8 add_addr_accepted;
232 u8 local_addr_used;
233 u8 pm_type;
234 u8 subflows;
235 u8 status;
236 DECLARE_BITMAP(id_avail_bitmap, MPTCP_PM_MAX_ADDR_ID + 1);
237 struct mptcp_rm_list rm_list_tx;
238 struct mptcp_rm_list rm_list_rx;
239 };
240
241 struct mptcp_pm_local {
242 struct mptcp_addr_info addr;
243 u8 flags;
244 int ifindex;
245 };
246
247 struct mptcp_pm_addr_entry {
248 struct list_head list;
249 struct mptcp_addr_info addr;
250 u8 flags;
251 int ifindex;
252 struct socket *lsk;
253 };
254
255 struct mptcp_data_frag {
256 struct list_head list;
257 u64 data_seq;
258 u16 data_len;
259 u16 offset;
260 u16 overhead;
261 u16 already_sent;
262 struct page *page;
263 };
264
265 /* MPTCP connection sock */
266 struct mptcp_sock {
267 /* inet_connection_sock must be the first member */
268 struct inet_connection_sock sk;
269 u64 local_key; /* protected by the first subflow socket lock
270 * lockless access read
271 */
272 u64 remote_key; /* same as above */
273 u64 write_seq;
274 u64 bytes_sent;
275 u64 snd_nxt;
276 u64 bytes_received;
277 u64 ack_seq;
278 atomic64_t rcv_wnd_sent;
279 u64 rcv_data_fin_seq;
280 u64 bytes_retrans;
281 u64 bytes_consumed;
282 int rmem_fwd_alloc;
283 int snd_burst;
284 int old_wspace;
285 u64 recovery_snd_nxt; /* in recovery mode accept up to this seq;
286 * recovery related fields are under data_lock
287 * protection
288 */
289 u64 bytes_acked;
290 u64 snd_una;
291 u64 wnd_end;
292 u32 last_data_sent;
293 u32 last_data_recv;
294 u32 last_ack_recv;
295 unsigned long timer_ival;
296 u32 token;
297 int rmem_released;
298 unsigned long flags;
299 unsigned long cb_flags;
300 bool recovery; /* closing subflow write queue reinjected */
301 bool can_ack;
302 bool fully_established;
303 bool rcv_data_fin;
304 bool snd_data_fin_enable;
305 bool rcv_fastclose;
306 bool use_64bit_ack; /* Set when we received a 64-bit DSN */
307 bool csum_enabled;
308 bool allow_infinite_fallback;
309 u8 pending_state; /* A subflow asked to set this sk_state,
310 * protected by the msk data lock
311 */
312 u8 mpc_endpoint_id;
313 u8 recvmsg_inq:1,
314 cork:1,
315 nodelay:1,
316 fastopening:1,
317 in_accept_queue:1,
318 free_first:1,
319 rcvspace_init:1;
320 u32 notsent_lowat;
321 int keepalive_cnt;
322 int keepalive_idle;
323 int keepalive_intvl;
324 struct work_struct work;
325 struct sk_buff *ooo_last_skb;
326 struct rb_root out_of_order_queue;
327 struct sk_buff_head receive_queue;
328 struct list_head conn_list;
329 struct list_head rtx_queue;
330 struct mptcp_data_frag *first_pending;
331 struct list_head join_list;
332 struct sock *first; /* The mptcp ops can safely dereference, using suitable
333 * ONCE annotation, the subflow outside the socket
334 * lock as such sock is freed after close().
335 */
336 struct mptcp_pm_data pm;
337 struct mptcp_sched_ops *sched;
338 struct {
339 u32 space; /* bytes copied in last measurement window */
340 u32 copied; /* bytes copied in this measurement window */
341 u64 time; /* start time of measurement window */
342 u64 rtt_us; /* last maximum rtt of subflows */
343 } rcvq_space;
344 u8 scaling_ratio;
345 bool allow_subflows;
346
347 u32 subflow_id;
348 u32 setsockopt_seq;
349 char ca_name[TCP_CA_NAME_MAX];
350
351 spinlock_t fallback_lock; /* protects fallback,
352 * allow_infinite_fallback and
353 * allow_join
354 */
355 };
356
357 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock)
358 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock)
359
360 #define mptcp_for_each_subflow(__msk, __subflow) \
361 list_for_each_entry(__subflow, &((__msk)->conn_list), node)
362 #define mptcp_for_each_subflow_safe(__msk, __subflow, __tmp) \
363 list_for_each_entry_safe(__subflow, __tmp, &((__msk)->conn_list), node)
364
365 extern struct genl_family mptcp_genl_family;
366
msk_owned_by_me(const struct mptcp_sock * msk)367 static inline void msk_owned_by_me(const struct mptcp_sock *msk)
368 {
369 sock_owned_by_me((const struct sock *)msk);
370 }
371
372 #ifdef CONFIG_DEBUG_NET
373 /* MPTCP-specific: we might (indirectly) call this helper with the wrong sk */
374 #undef tcp_sk
375 #define tcp_sk(ptr) ({ \
376 typeof(ptr) _ptr = (ptr); \
377 WARN_ON(_ptr->sk_protocol != IPPROTO_TCP); \
378 container_of_const(_ptr, struct tcp_sock, inet_conn.icsk_inet.sk); \
379 })
380 #define mptcp_sk(ptr) ({ \
381 typeof(ptr) _ptr = (ptr); \
382 WARN_ON(_ptr->sk_protocol != IPPROTO_MPTCP); \
383 container_of_const(_ptr, struct mptcp_sock, sk.icsk_inet.sk); \
384 })
385
386 #else /* !CONFIG_DEBUG_NET */
387 #define mptcp_sk(ptr) container_of_const(ptr, struct mptcp_sock, sk.icsk_inet.sk)
388 #endif
389
390 /* the msk socket don't use the backlog, also account for the bulk
391 * free memory
392 */
__mptcp_rmem(const struct sock * sk)393 static inline int __mptcp_rmem(const struct sock *sk)
394 {
395 return atomic_read(&sk->sk_rmem_alloc) - READ_ONCE(mptcp_sk(sk)->rmem_released);
396 }
397
mptcp_win_from_space(const struct sock * sk,int space)398 static inline int mptcp_win_from_space(const struct sock *sk, int space)
399 {
400 return __tcp_win_from_space(mptcp_sk(sk)->scaling_ratio, space);
401 }
402
mptcp_space_from_win(const struct sock * sk,int win)403 static inline int mptcp_space_from_win(const struct sock *sk, int win)
404 {
405 return __tcp_space_from_win(mptcp_sk(sk)->scaling_ratio, win);
406 }
407
__mptcp_space(const struct sock * sk)408 static inline int __mptcp_space(const struct sock *sk)
409 {
410 return mptcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - __mptcp_rmem(sk));
411 }
412
mptcp_send_head(const struct sock * sk)413 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk)
414 {
415 const struct mptcp_sock *msk = mptcp_sk(sk);
416
417 return READ_ONCE(msk->first_pending);
418 }
419
mptcp_send_next(struct sock * sk)420 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk)
421 {
422 struct mptcp_sock *msk = mptcp_sk(sk);
423 struct mptcp_data_frag *cur;
424
425 cur = msk->first_pending;
426 return list_is_last(&cur->list, &msk->rtx_queue) ? NULL :
427 list_next_entry(cur, list);
428 }
429
mptcp_pending_tail(const struct sock * sk)430 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk)
431 {
432 const struct mptcp_sock *msk = mptcp_sk(sk);
433
434 if (!msk->first_pending)
435 return NULL;
436
437 if (WARN_ON_ONCE(list_empty(&msk->rtx_queue)))
438 return NULL;
439
440 return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list);
441 }
442
mptcp_rtx_head(struct sock * sk)443 static inline struct mptcp_data_frag *mptcp_rtx_head(struct sock *sk)
444 {
445 struct mptcp_sock *msk = mptcp_sk(sk);
446
447 if (msk->snd_una == msk->snd_nxt)
448 return NULL;
449
450 return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list);
451 }
452
453 struct csum_pseudo_header {
454 __be64 data_seq;
455 __be32 subflow_seq;
456 __be16 data_len;
457 __sum16 csum;
458 };
459
460 struct mptcp_subflow_request_sock {
461 struct tcp_request_sock sk;
462 u16 mp_capable : 1,
463 mp_join : 1,
464 backup : 1,
465 request_bkup : 1,
466 csum_reqd : 1,
467 allow_join_id0 : 1;
468 u8 local_id;
469 u8 remote_id;
470 u64 local_key;
471 u64 idsn;
472 u32 token;
473 u32 ssn_offset;
474 u64 thmac;
475 u32 local_nonce;
476 u32 remote_nonce;
477 struct mptcp_sock *msk;
478 struct hlist_nulls_node token_node;
479 };
480
481 static inline struct mptcp_subflow_request_sock *
mptcp_subflow_rsk(const struct request_sock * rsk)482 mptcp_subflow_rsk(const struct request_sock *rsk)
483 {
484 return (struct mptcp_subflow_request_sock *)rsk;
485 }
486
487 struct mptcp_delegated_action {
488 struct napi_struct napi;
489 struct list_head head;
490 };
491
492 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions);
493
494 #define MPTCP_DELEGATE_SCHEDULED 0
495 #define MPTCP_DELEGATE_SEND 1
496 #define MPTCP_DELEGATE_ACK 2
497 #define MPTCP_DELEGATE_SNDBUF 3
498
499 #define MPTCP_DELEGATE_ACTIONS_MASK (~BIT(MPTCP_DELEGATE_SCHEDULED))
500 /* MPTCP subflow context */
501 struct mptcp_subflow_context {
502 struct list_head node;/* conn_list of subflows */
503
504 struct_group(reset,
505
506 unsigned long avg_pacing_rate; /* protected by msk socket lock */
507 u64 local_key;
508 u64 remote_key;
509 u64 idsn;
510 u64 map_seq;
511 u32 snd_isn;
512 u32 token;
513 u32 rel_write_seq;
514 u32 map_subflow_seq;
515 u32 ssn_offset;
516 u32 map_data_len;
517 __wsum map_data_csum;
518 u32 map_csum_len;
519 u32 request_mptcp : 1, /* send MP_CAPABLE */
520 request_join : 1, /* send MP_JOIN */
521 request_bkup : 1,
522 mp_capable : 1, /* remote is MPTCP capable */
523 mp_join : 1, /* remote is JOINing */
524 pm_notified : 1, /* PM hook called for established status */
525 conn_finished : 1,
526 map_valid : 1,
527 map_csum_reqd : 1,
528 map_data_fin : 1,
529 mpc_map : 1,
530 backup : 1,
531 send_mp_prio : 1,
532 send_mp_fail : 1,
533 send_fastclose : 1,
534 send_infinite_map : 1,
535 remote_key_valid : 1, /* received the peer key from */
536 disposable : 1, /* ctx can be free at ulp release time */
537 stale : 1, /* unable to snd/rcv data, do not use for xmit */
538 valid_csum_seen : 1, /* at least one csum validated */
539 is_mptfo : 1, /* subflow is doing TFO */
540 close_event_done : 1, /* has done the post-closed part */
541 mpc_drop : 1, /* the MPC option has been dropped in a rtx */
542 __unused : 9;
543 bool data_avail;
544 bool scheduled;
545 bool pm_listener; /* a listener managed by the kernel PM? */
546 bool fully_established; /* path validated */
547 u32 remote_nonce;
548 u64 thmac;
549 u32 local_nonce;
550 u32 remote_token;
551 union {
552 u8 hmac[MPTCPOPT_HMAC_LEN]; /* MPJ subflow only */
553 u64 iasn; /* initial ack sequence number, MPC subflows only */
554 };
555 s16 local_id; /* if negative not initialized yet */
556 u8 remote_id;
557 u8 reset_seen:1;
558 u8 reset_transient:1;
559 u8 reset_reason:4;
560 u8 stale_count;
561
562 u32 subflow_id;
563
564 long delegated_status;
565 unsigned long fail_tout;
566
567 );
568
569 struct list_head delegated_node; /* link into delegated_action, protected by local BH */
570
571 u32 setsockopt_seq;
572 u32 stale_rcv_tstamp;
573 int cached_sndbuf; /* sndbuf size when last synced with the msk sndbuf,
574 * protected by the msk socket lock
575 */
576
577 struct sock *tcp_sock; /* tcp sk backpointer */
578 struct sock *conn; /* parent mptcp_sock */
579 const struct inet_connection_sock_af_ops *icsk_af_ops;
580 void (*tcp_state_change)(struct sock *sk);
581 void (*tcp_error_report)(struct sock *sk);
582
583 struct rcu_head rcu;
584 };
585
586 static inline struct mptcp_subflow_context *
mptcp_subflow_ctx(const struct sock * sk)587 mptcp_subflow_ctx(const struct sock *sk)
588 {
589 const struct inet_connection_sock *icsk = inet_csk(sk);
590
591 /* Use RCU on icsk_ulp_data only for sock diag code */
592 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data;
593 }
594
595 static inline struct sock *
mptcp_subflow_tcp_sock(const struct mptcp_subflow_context * subflow)596 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow)
597 {
598 return subflow->tcp_sock;
599 }
600
601 static inline void
mptcp_subflow_ctx_reset(struct mptcp_subflow_context * subflow)602 mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow)
603 {
604 memset(&subflow->reset, 0, sizeof(subflow->reset));
605 subflow->request_mptcp = 1;
606 WRITE_ONCE(subflow->local_id, -1);
607 }
608
609 /* Convert reset reasons in MPTCP to enum sk_rst_reason type */
610 static inline enum sk_rst_reason
sk_rst_convert_mptcp_reason(u32 reason)611 sk_rst_convert_mptcp_reason(u32 reason)
612 {
613 switch (reason) {
614 case MPTCP_RST_EUNSPEC:
615 return SK_RST_REASON_MPTCP_RST_EUNSPEC;
616 case MPTCP_RST_EMPTCP:
617 return SK_RST_REASON_MPTCP_RST_EMPTCP;
618 case MPTCP_RST_ERESOURCE:
619 return SK_RST_REASON_MPTCP_RST_ERESOURCE;
620 case MPTCP_RST_EPROHIBIT:
621 return SK_RST_REASON_MPTCP_RST_EPROHIBIT;
622 case MPTCP_RST_EWQ2BIG:
623 return SK_RST_REASON_MPTCP_RST_EWQ2BIG;
624 case MPTCP_RST_EBADPERF:
625 return SK_RST_REASON_MPTCP_RST_EBADPERF;
626 case MPTCP_RST_EMIDDLEBOX:
627 return SK_RST_REASON_MPTCP_RST_EMIDDLEBOX;
628 default:
629 /* It should not happen, or else errors may occur
630 * in MPTCP layer
631 */
632 return SK_RST_REASON_ERROR;
633 }
634 }
635
636 static inline void
mptcp_send_active_reset_reason(struct sock * sk)637 mptcp_send_active_reset_reason(struct sock *sk)
638 {
639 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
640 enum sk_rst_reason reason;
641
642 reason = sk_rst_convert_mptcp_reason(subflow->reset_reason);
643 tcp_send_active_reset(sk, GFP_ATOMIC, reason);
644 }
645
646 static inline u64
mptcp_subflow_get_map_offset(const struct mptcp_subflow_context * subflow)647 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow)
648 {
649 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq -
650 subflow->ssn_offset -
651 subflow->map_subflow_seq;
652 }
653
654 static inline u64
mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context * subflow)655 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow)
656 {
657 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow);
658 }
659
660 void mptcp_subflow_process_delegated(struct sock *ssk, long actions);
661
mptcp_subflow_delegate(struct mptcp_subflow_context * subflow,int action)662 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action)
663 {
664 long old, set_bits = BIT(MPTCP_DELEGATE_SCHEDULED) | BIT(action);
665 struct mptcp_delegated_action *delegated;
666 bool schedule;
667
668 /* the caller held the subflow bh socket lock */
669 lockdep_assert_in_softirq();
670
671 /* The implied barrier pairs with tcp_release_cb_override()
672 * mptcp_napi_poll(), and ensures the below list check sees list
673 * updates done prior to delegated status bits changes
674 */
675 old = set_mask_bits(&subflow->delegated_status, 0, set_bits);
676 if (!(old & BIT(MPTCP_DELEGATE_SCHEDULED))) {
677 if (WARN_ON_ONCE(!list_empty(&subflow->delegated_node)))
678 return;
679
680 delegated = this_cpu_ptr(&mptcp_delegated_actions);
681 schedule = list_empty(&delegated->head);
682 list_add_tail(&subflow->delegated_node, &delegated->head);
683 sock_hold(mptcp_subflow_tcp_sock(subflow));
684 if (schedule)
685 napi_schedule(&delegated->napi);
686 }
687 }
688
689 static inline struct mptcp_subflow_context *
mptcp_subflow_delegated_next(struct mptcp_delegated_action * delegated)690 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated)
691 {
692 struct mptcp_subflow_context *ret;
693
694 if (list_empty(&delegated->head))
695 return NULL;
696
697 ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node);
698 list_del_init(&ret->delegated_node);
699 return ret;
700 }
701
702 int mptcp_is_enabled(const struct net *net);
703 unsigned int mptcp_get_add_addr_timeout(const struct net *net);
704 int mptcp_is_checksum_enabled(const struct net *net);
705 int mptcp_allow_join_id0(const struct net *net);
706 unsigned int mptcp_stale_loss_cnt(const struct net *net);
707 unsigned int mptcp_close_timeout(const struct sock *sk);
708 int mptcp_get_pm_type(const struct net *net);
709 const char *mptcp_get_scheduler(const struct net *net);
710
711 void mptcp_active_disable(struct sock *sk);
712 bool mptcp_active_should_disable(struct sock *ssk);
713 void mptcp_active_enable(struct sock *sk);
714
715 void mptcp_get_available_schedulers(char *buf, size_t maxlen);
716 void __mptcp_subflow_fully_established(struct mptcp_sock *msk,
717 struct mptcp_subflow_context *subflow,
718 const struct mptcp_options_received *mp_opt);
719 bool __mptcp_retransmit_pending_data(struct sock *sk);
720 void mptcp_check_and_set_pending(struct sock *sk);
721 void __mptcp_push_pending(struct sock *sk, unsigned int flags);
722 bool mptcp_subflow_data_available(struct sock *sk);
723 void __init mptcp_subflow_init(void);
724 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how);
725 void mptcp_close_ssk(struct sock *sk, struct sock *ssk,
726 struct mptcp_subflow_context *subflow);
727 void __mptcp_subflow_send_ack(struct sock *ssk);
728 void mptcp_subflow_reset(struct sock *ssk);
729 void mptcp_subflow_queue_clean(struct sock *sk, struct sock *ssk);
730 void mptcp_sock_graft(struct sock *sk, struct socket *parent);
731 struct sock *__mptcp_nmpc_sk(struct mptcp_sock *msk);
732 bool __mptcp_close(struct sock *sk, long timeout);
733 void mptcp_cancel_work(struct sock *sk);
734 void __mptcp_unaccepted_force_close(struct sock *sk);
735 void mptcp_set_owner_r(struct sk_buff *skb, struct sock *sk);
736 void mptcp_set_state(struct sock *sk, int state);
737
738 bool mptcp_addresses_equal(const struct mptcp_addr_info *a,
739 const struct mptcp_addr_info *b, bool use_port);
740 void mptcp_local_address(const struct sock_common *skc, struct mptcp_addr_info *addr);
741
742 /* called with sk socket lock held */
743 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_pm_local *local,
744 const struct mptcp_addr_info *remote);
745 int mptcp_subflow_create_socket(struct sock *sk, unsigned short family,
746 struct socket **new_sock);
747 void mptcp_info2sockaddr(const struct mptcp_addr_info *info,
748 struct sockaddr_storage *addr,
749 unsigned short family);
750 struct mptcp_sched_ops *mptcp_sched_find(const char *name);
751 int mptcp_register_scheduler(struct mptcp_sched_ops *sched);
752 void mptcp_unregister_scheduler(struct mptcp_sched_ops *sched);
753 void mptcp_sched_init(void);
754 int mptcp_init_sched(struct mptcp_sock *msk,
755 struct mptcp_sched_ops *sched);
756 void mptcp_release_sched(struct mptcp_sock *msk);
757 void mptcp_subflow_set_scheduled(struct mptcp_subflow_context *subflow,
758 bool scheduled);
759 struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk);
760 struct sock *mptcp_subflow_get_retrans(struct mptcp_sock *msk);
761 int mptcp_sched_get_send(struct mptcp_sock *msk);
762 int mptcp_sched_get_retrans(struct mptcp_sock *msk);
763
mptcp_data_avail(const struct mptcp_sock * msk)764 static inline u64 mptcp_data_avail(const struct mptcp_sock *msk)
765 {
766 return READ_ONCE(msk->bytes_received) - READ_ONCE(msk->bytes_consumed);
767 }
768
mptcp_epollin_ready(const struct sock * sk)769 static inline bool mptcp_epollin_ready(const struct sock *sk)
770 {
771 u64 data_avail = mptcp_data_avail(mptcp_sk(sk));
772
773 if (!data_avail)
774 return false;
775
776 /* mptcp doesn't have to deal with small skbs in the receive queue,
777 * as it can always coalesce them
778 */
779 return (data_avail >= sk->sk_rcvlowat) ||
780 (mem_cgroup_sockets_enabled && sk->sk_memcg &&
781 mem_cgroup_under_socket_pressure(sk->sk_memcg)) ||
782 READ_ONCE(tcp_memory_pressure);
783 }
784
785 int mptcp_set_rcvlowat(struct sock *sk, int val);
786
__tcp_can_send(const struct sock * ssk)787 static inline bool __tcp_can_send(const struct sock *ssk)
788 {
789 /* only send if our side has not closed yet */
790 return ((1 << inet_sk_state_load(ssk)) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT));
791 }
792
__mptcp_subflow_active(struct mptcp_subflow_context * subflow)793 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow)
794 {
795 /* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */
796 if (subflow->request_join && !READ_ONCE(subflow->fully_established))
797 return false;
798
799 return __tcp_can_send(mptcp_subflow_tcp_sock(subflow));
800 }
801
802 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow);
803
804 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow);
805
806 void mptcp_subflow_drop_ctx(struct sock *ssk);
807
mptcp_subflow_tcp_fallback(struct sock * sk,struct mptcp_subflow_context * ctx)808 static inline void mptcp_subflow_tcp_fallback(struct sock *sk,
809 struct mptcp_subflow_context *ctx)
810 {
811 sk->sk_data_ready = sock_def_readable;
812 sk->sk_state_change = ctx->tcp_state_change;
813 sk->sk_write_space = sk_stream_write_space;
814 sk->sk_error_report = ctx->tcp_error_report;
815
816 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops;
817 }
818
819 void __init mptcp_proto_init(void);
820 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
821 int __init mptcp_proto_v6_init(void);
822 #endif
823
824 struct sock *mptcp_sk_clone_init(const struct sock *sk,
825 const struct mptcp_options_received *mp_opt,
826 struct sock *ssk,
827 struct request_sock *req);
828 void mptcp_get_options(const struct sk_buff *skb,
829 struct mptcp_options_received *mp_opt);
830
831 void mptcp_finish_connect(struct sock *sk);
832 void __mptcp_sync_state(struct sock *sk, int state);
833 void mptcp_reset_tout_timer(struct mptcp_sock *msk, unsigned long fail_tout);
834
mptcp_stop_tout_timer(struct sock * sk)835 static inline void mptcp_stop_tout_timer(struct sock *sk)
836 {
837 if (!inet_csk(sk)->icsk_mtup.probe_timestamp)
838 return;
839
840 sk_stop_timer(sk, &sk->sk_timer);
841 inet_csk(sk)->icsk_mtup.probe_timestamp = 0;
842 }
843
mptcp_set_close_tout(struct sock * sk,unsigned long tout)844 static inline void mptcp_set_close_tout(struct sock *sk, unsigned long tout)
845 {
846 /* avoid 0 timestamp, as that means no close timeout */
847 inet_csk(sk)->icsk_mtup.probe_timestamp = tout ? : 1;
848 }
849
mptcp_start_tout_timer(struct sock * sk)850 static inline void mptcp_start_tout_timer(struct sock *sk)
851 {
852 mptcp_set_close_tout(sk, tcp_jiffies32);
853 mptcp_reset_tout_timer(mptcp_sk(sk), 0);
854 }
855
mptcp_is_fully_established(struct sock * sk)856 static inline bool mptcp_is_fully_established(struct sock *sk)
857 {
858 return inet_sk_state_load(sk) == TCP_ESTABLISHED &&
859 READ_ONCE(mptcp_sk(sk)->fully_established);
860 }
861
862 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk);
863 void mptcp_data_ready(struct sock *sk, struct sock *ssk);
864 bool mptcp_finish_join(struct sock *sk);
865 bool mptcp_schedule_work(struct sock *sk);
866 int mptcp_setsockopt(struct sock *sk, int level, int optname,
867 sockptr_t optval, unsigned int optlen);
868 int mptcp_getsockopt(struct sock *sk, int level, int optname,
869 char __user *optval, int __user *option);
870
871 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq);
mptcp_expand_seq(u64 old_seq,u64 cur_seq,bool use_64bit)872 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit)
873 {
874 if (use_64bit)
875 return cur_seq;
876
877 return __mptcp_expand_seq(old_seq, cur_seq);
878 }
879 void __mptcp_check_push(struct sock *sk, struct sock *ssk);
880 void __mptcp_data_acked(struct sock *sk);
881 void __mptcp_error_report(struct sock *sk);
882 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)883 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk)
884 {
885 return READ_ONCE(msk->snd_data_fin_enable) &&
886 READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt);
887 }
888
mptcp_notsent_lowat(const struct sock * sk)889 static inline u32 mptcp_notsent_lowat(const struct sock *sk)
890 {
891 struct net *net = sock_net(sk);
892 u32 val;
893
894 val = READ_ONCE(mptcp_sk(sk)->notsent_lowat);
895 return val ?: READ_ONCE(net->ipv4.sysctl_tcp_notsent_lowat);
896 }
897
mptcp_stream_memory_free(const struct sock * sk,int wake)898 static inline bool mptcp_stream_memory_free(const struct sock *sk, int wake)
899 {
900 const struct mptcp_sock *msk = mptcp_sk(sk);
901 u32 notsent_bytes;
902
903 notsent_bytes = READ_ONCE(msk->write_seq) - READ_ONCE(msk->snd_nxt);
904 return (notsent_bytes << wake) < mptcp_notsent_lowat(sk);
905 }
906
__mptcp_stream_is_writeable(const struct sock * sk,int wake)907 static inline bool __mptcp_stream_is_writeable(const struct sock *sk, int wake)
908 {
909 return mptcp_stream_memory_free(sk, wake) &&
910 __sk_stream_is_writeable(sk, wake);
911 }
912
mptcp_write_space(struct sock * sk)913 static inline void mptcp_write_space(struct sock *sk)
914 {
915 /* pairs with memory barrier in mptcp_poll */
916 smp_mb();
917 if (mptcp_stream_memory_free(sk, 1))
918 sk_stream_write_space(sk);
919 }
920
__mptcp_sync_sndbuf(struct sock * sk)921 static inline void __mptcp_sync_sndbuf(struct sock *sk)
922 {
923 struct mptcp_subflow_context *subflow;
924 int ssk_sndbuf, new_sndbuf;
925
926 if (sk->sk_userlocks & SOCK_SNDBUF_LOCK)
927 return;
928
929 new_sndbuf = READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_wmem[0]);
930 mptcp_for_each_subflow(mptcp_sk(sk), subflow) {
931 ssk_sndbuf = READ_ONCE(mptcp_subflow_tcp_sock(subflow)->sk_sndbuf);
932
933 subflow->cached_sndbuf = ssk_sndbuf;
934 new_sndbuf += ssk_sndbuf;
935 }
936
937 /* the msk max wmem limit is <nr_subflows> * tcp wmem[2] */
938 WRITE_ONCE(sk->sk_sndbuf, new_sndbuf);
939 mptcp_write_space(sk);
940 }
941
942 /* The called held both the msk socket and the subflow socket locks,
943 * possibly under BH
944 */
__mptcp_propagate_sndbuf(struct sock * sk,struct sock * ssk)945 static inline void __mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
946 {
947 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
948
949 if (READ_ONCE(ssk->sk_sndbuf) != subflow->cached_sndbuf)
950 __mptcp_sync_sndbuf(sk);
951 }
952
953 /* the caller held only the subflow socket lock, either in process or
954 * BH context. Additionally this can be called under the msk data lock,
955 * so we can't acquire such lock here: let the delegate action acquires
956 * the needed locks in suitable order.
957 */
mptcp_propagate_sndbuf(struct sock * sk,struct sock * ssk)958 static inline void mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
959 {
960 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
961
962 if (likely(READ_ONCE(ssk->sk_sndbuf) == subflow->cached_sndbuf))
963 return;
964
965 local_bh_disable();
966 mptcp_subflow_delegate(subflow, MPTCP_DELEGATE_SNDBUF);
967 local_bh_enable();
968 }
969
970 void mptcp_destroy_common(struct mptcp_sock *msk, unsigned int flags);
971
972 #define MPTCP_TOKEN_MAX_RETRIES 4
973
974 void __init mptcp_token_init(void);
mptcp_token_init_request(struct request_sock * req)975 static inline void mptcp_token_init_request(struct request_sock *req)
976 {
977 mptcp_subflow_rsk(req)->token_node.pprev = NULL;
978 }
979
980 int mptcp_token_new_request(struct request_sock *req);
981 void mptcp_token_destroy_request(struct request_sock *req);
982 int mptcp_token_new_connect(struct sock *ssk);
983 void mptcp_token_accept(struct mptcp_subflow_request_sock *r,
984 struct mptcp_sock *msk);
985 bool mptcp_token_exists(u32 token);
986 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token);
987 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot,
988 long *s_num);
989 void mptcp_token_destroy(struct mptcp_sock *msk);
990
991 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn);
992
993 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac);
994 __sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum);
995
996 void __init mptcp_pm_init(void);
997 void mptcp_pm_data_init(struct mptcp_sock *msk);
998 void mptcp_pm_data_reset(struct mptcp_sock *msk);
999 int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info,
1000 struct mptcp_addr_info *addr);
1001 int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info,
1002 bool require_family,
1003 struct mptcp_pm_addr_entry *entry);
1004 bool mptcp_pm_addr_families_match(const struct sock *sk,
1005 const struct mptcp_addr_info *loc,
1006 const struct mptcp_addr_info *rem);
1007 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
1008 void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
1009 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side);
1010 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk);
1011 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk);
1012 void mptcp_pm_connection_closed(struct mptcp_sock *msk);
1013 void mptcp_pm_subflow_established(struct mptcp_sock *msk);
1014 bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk);
1015 void mptcp_pm_subflow_check_next(struct mptcp_sock *msk,
1016 const struct mptcp_subflow_context *subflow);
1017 void mptcp_pm_add_addr_received(const struct sock *ssk,
1018 const struct mptcp_addr_info *addr);
1019 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk,
1020 const struct mptcp_addr_info *addr);
1021 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk);
1022 bool mptcp_pm_nl_is_init_remote_addr(struct mptcp_sock *msk,
1023 const struct mptcp_addr_info *remote);
1024 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk);
1025 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk,
1026 const struct mptcp_rm_list *rm_list);
1027 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup);
1028 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq);
1029 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk,
1030 struct mptcp_addr_info *addr,
1031 struct mptcp_addr_info *rem,
1032 u8 bkup);
1033 bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk,
1034 const struct mptcp_addr_info *addr);
1035 void mptcp_pm_free_anno_list(struct mptcp_sock *msk);
1036 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk);
1037 struct mptcp_pm_add_entry *
1038 mptcp_pm_del_add_timer(struct mptcp_sock *msk,
1039 const struct mptcp_addr_info *addr, bool check_id);
1040 struct mptcp_pm_add_entry *
1041 mptcp_lookup_anno_list_by_saddr(const struct mptcp_sock *msk,
1042 const struct mptcp_addr_info *addr);
1043 int mptcp_pm_set_flags(struct sk_buff *skb, struct genl_info *info);
1044 int mptcp_pm_nl_set_flags(struct sk_buff *skb, struct genl_info *info);
1045 int mptcp_userspace_pm_set_flags(struct sk_buff *skb, struct genl_info *info);
1046 int mptcp_pm_announce_addr(struct mptcp_sock *msk,
1047 const struct mptcp_addr_info *addr,
1048 bool echo);
1049 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
1050 void mptcp_pm_remove_addrs(struct mptcp_sock *msk, struct list_head *rm_list);
1051
1052 void mptcp_free_local_addr_list(struct mptcp_sock *msk);
1053
1054 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk,
1055 const struct sock *ssk, gfp_t gfp);
1056 void mptcp_event_addr_announced(const struct sock *ssk, const struct mptcp_addr_info *info);
1057 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id);
1058 void mptcp_event_pm_listener(const struct sock *ssk,
1059 enum mptcp_event_type event);
1060 bool mptcp_userspace_pm_active(const struct mptcp_sock *msk);
1061
1062 void __mptcp_fastopen_gen_msk_ackseq(struct mptcp_sock *msk, struct mptcp_subflow_context *subflow,
1063 const struct mptcp_options_received *mp_opt);
1064 void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subflow,
1065 struct request_sock *req);
1066 int mptcp_nl_fill_addr(struct sk_buff *skb,
1067 struct mptcp_pm_addr_entry *entry);
1068
mptcp_pm_should_add_signal(struct mptcp_sock * msk)1069 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk)
1070 {
1071 return READ_ONCE(msk->pm.addr_signal) &
1072 (BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO));
1073 }
1074
mptcp_pm_should_add_signal_addr(struct mptcp_sock * msk)1075 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk)
1076 {
1077 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL);
1078 }
1079
mptcp_pm_should_add_signal_echo(struct mptcp_sock * msk)1080 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk)
1081 {
1082 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO);
1083 }
1084
mptcp_pm_should_rm_signal(struct mptcp_sock * msk)1085 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk)
1086 {
1087 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL);
1088 }
1089
mptcp_pm_is_userspace(const struct mptcp_sock * msk)1090 static inline bool mptcp_pm_is_userspace(const struct mptcp_sock *msk)
1091 {
1092 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_USERSPACE;
1093 }
1094
mptcp_pm_is_kernel(const struct mptcp_sock * msk)1095 static inline bool mptcp_pm_is_kernel(const struct mptcp_sock *msk)
1096 {
1097 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_KERNEL;
1098 }
1099
mptcp_add_addr_len(int family,bool echo,bool port)1100 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port)
1101 {
1102 u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE;
1103
1104 if (family == AF_INET6)
1105 len = TCPOLEN_MPTCP_ADD_ADDR6_BASE;
1106 if (!echo)
1107 len += MPTCPOPT_THMAC_LEN;
1108 /* account for 2 trailing 'nop' options */
1109 if (port)
1110 len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN;
1111
1112 return len;
1113 }
1114
mptcp_rm_addr_len(const struct mptcp_rm_list * rm_list)1115 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list)
1116 {
1117 if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX)
1118 return -EINVAL;
1119
1120 return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1;
1121 }
1122
1123 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, const struct sk_buff *skb,
1124 unsigned int opt_size, unsigned int remaining,
1125 struct mptcp_addr_info *addr, bool *echo,
1126 bool *drop_other_suboptions);
1127 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining,
1128 struct mptcp_rm_list *rm_list);
1129 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
1130 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1131 int mptcp_userspace_pm_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1132 bool mptcp_pm_is_backup(struct mptcp_sock *msk, struct sock_common *skc);
1133 bool mptcp_pm_nl_is_backup(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1134 bool mptcp_userspace_pm_is_backup(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1135 int mptcp_pm_dump_addr(struct sk_buff *msg, struct netlink_callback *cb);
1136 int mptcp_pm_nl_dump_addr(struct sk_buff *msg,
1137 struct netlink_callback *cb);
1138 int mptcp_userspace_pm_dump_addr(struct sk_buff *msg,
1139 struct netlink_callback *cb);
1140 int mptcp_pm_get_addr(struct sk_buff *skb, struct genl_info *info);
1141 int mptcp_pm_nl_get_addr(struct sk_buff *skb, struct genl_info *info);
1142 int mptcp_userspace_pm_get_addr(struct sk_buff *skb,
1143 struct genl_info *info);
1144
subflow_get_local_id(const struct mptcp_subflow_context * subflow)1145 static inline u8 subflow_get_local_id(const struct mptcp_subflow_context *subflow)
1146 {
1147 int local_id = READ_ONCE(subflow->local_id);
1148
1149 if (local_id < 0)
1150 return 0;
1151 return local_id;
1152 }
1153
1154 void __init mptcp_pm_nl_init(void);
1155 void mptcp_pm_nl_work(struct mptcp_sock *msk);
1156 unsigned int mptcp_pm_get_add_addr_signal_max(const struct mptcp_sock *msk);
1157 unsigned int mptcp_pm_get_add_addr_accept_max(const struct mptcp_sock *msk);
1158 unsigned int mptcp_pm_get_subflows_max(const struct mptcp_sock *msk);
1159 unsigned int mptcp_pm_get_local_addr_max(const struct mptcp_sock *msk);
1160
1161 /* called under PM lock */
__mptcp_pm_close_subflow(struct mptcp_sock * msk)1162 static inline void __mptcp_pm_close_subflow(struct mptcp_sock *msk)
1163 {
1164 if (--msk->pm.subflows < mptcp_pm_get_subflows_max(msk))
1165 WRITE_ONCE(msk->pm.accept_subflow, true);
1166 }
1167
mptcp_pm_close_subflow(struct mptcp_sock * msk)1168 static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk)
1169 {
1170 spin_lock_bh(&msk->pm.lock);
1171 __mptcp_pm_close_subflow(msk);
1172 spin_unlock_bh(&msk->pm.lock);
1173 }
1174
1175 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk);
1176
mptcp_get_ext(const struct sk_buff * skb)1177 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb)
1178 {
1179 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP);
1180 }
1181
1182 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops);
1183
__mptcp_check_fallback(const struct mptcp_sock * msk)1184 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk)
1185 {
1186 return test_bit(MPTCP_FALLBACK_DONE, &msk->flags);
1187 }
1188
mptcp_check_fallback(const struct sock * sk)1189 static inline bool mptcp_check_fallback(const struct sock *sk)
1190 {
1191 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1192 struct mptcp_sock *msk = mptcp_sk(subflow->conn);
1193
1194 return __mptcp_check_fallback(msk);
1195 }
1196
__mptcp_try_fallback(struct mptcp_sock * msk)1197 static inline bool __mptcp_try_fallback(struct mptcp_sock *msk)
1198 {
1199 if (__mptcp_check_fallback(msk)) {
1200 pr_debug("TCP fallback already done (msk=%p)\n", msk);
1201 return true;
1202 }
1203 spin_lock_bh(&msk->fallback_lock);
1204 if (!msk->allow_infinite_fallback) {
1205 spin_unlock_bh(&msk->fallback_lock);
1206 return false;
1207 }
1208
1209 msk->allow_subflows = false;
1210 set_bit(MPTCP_FALLBACK_DONE, &msk->flags);
1211 spin_unlock_bh(&msk->fallback_lock);
1212 return true;
1213 }
1214
__mptcp_has_initial_subflow(const struct mptcp_sock * msk)1215 static inline bool __mptcp_has_initial_subflow(const struct mptcp_sock *msk)
1216 {
1217 struct sock *ssk = READ_ONCE(msk->first);
1218
1219 return ssk && ((1 << inet_sk_state_load(ssk)) &
1220 (TCPF_ESTABLISHED | TCPF_SYN_SENT |
1221 TCPF_SYN_RECV | TCPF_LISTEN));
1222 }
1223
mptcp_try_fallback(struct sock * ssk)1224 static inline bool mptcp_try_fallback(struct sock *ssk)
1225 {
1226 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
1227 struct sock *sk = subflow->conn;
1228 struct mptcp_sock *msk;
1229
1230 msk = mptcp_sk(sk);
1231 if (!__mptcp_try_fallback(msk))
1232 return false;
1233 if (READ_ONCE(msk->snd_data_fin_enable) && !(ssk->sk_shutdown & SEND_SHUTDOWN)) {
1234 gfp_t saved_allocation = ssk->sk_allocation;
1235
1236 /* we are in a atomic (BH) scope, override ssk default for data
1237 * fin allocation
1238 */
1239 ssk->sk_allocation = GFP_ATOMIC;
1240 ssk->sk_shutdown |= SEND_SHUTDOWN;
1241 tcp_shutdown(ssk, SEND_SHUTDOWN);
1242 ssk->sk_allocation = saved_allocation;
1243 }
1244 return true;
1245 }
1246
1247 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)\n", __func__, a)
1248
mptcp_subflow_early_fallback(struct mptcp_sock * msk,struct mptcp_subflow_context * subflow)1249 static inline void mptcp_subflow_early_fallback(struct mptcp_sock *msk,
1250 struct mptcp_subflow_context *subflow)
1251 {
1252 pr_fallback(msk);
1253 subflow->request_mptcp = 0;
1254 WARN_ON_ONCE(!__mptcp_try_fallback(msk));
1255 }
1256
mptcp_check_infinite_map(struct sk_buff * skb)1257 static inline bool mptcp_check_infinite_map(struct sk_buff *skb)
1258 {
1259 struct mptcp_ext *mpext;
1260
1261 mpext = skb ? mptcp_get_ext(skb) : NULL;
1262 if (mpext && mpext->infinite_map)
1263 return true;
1264
1265 return false;
1266 }
1267
is_active_ssk(struct mptcp_subflow_context * subflow)1268 static inline bool is_active_ssk(struct mptcp_subflow_context *subflow)
1269 {
1270 return (subflow->request_mptcp || subflow->request_join);
1271 }
1272
subflow_simultaneous_connect(struct sock * sk)1273 static inline bool subflow_simultaneous_connect(struct sock *sk)
1274 {
1275 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1276
1277 return (1 << sk->sk_state) &
1278 (TCPF_ESTABLISHED | TCPF_FIN_WAIT1 | TCPF_FIN_WAIT2 | TCPF_CLOSING) &&
1279 is_active_ssk(subflow) &&
1280 !subflow->conn_finished;
1281 }
1282
1283 #ifdef CONFIG_SYN_COOKIES
1284 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1285 struct sk_buff *skb);
1286 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1287 struct sk_buff *skb);
1288 void __init mptcp_join_cookie_init(void);
1289 #else
1290 static inline void
subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock * subflow_req,struct sk_buff * skb)1291 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1292 struct sk_buff *skb) {}
1293 static inline bool
mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock * subflow_req,struct sk_buff * skb)1294 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1295 struct sk_buff *skb)
1296 {
1297 return false;
1298 }
1299
mptcp_join_cookie_init(void)1300 static inline void mptcp_join_cookie_init(void) {}
1301 #endif
1302
1303 #endif /* __MPTCP_PROTOCOL_H */
1304