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 u16 suboptions;
153 u32 token;
154 u32 nonce;
155 u16 use_map:1,
156 dsn64:1,
157 data_fin:1,
158 use_ack:1,
159 ack64:1,
160 mpc_map:1,
161 reset_reason:4,
162 reset_transient:1,
163 echo:1,
164 backup:1,
165 deny_join_id0:1,
166 __unused:2;
167 u8 join_id;
168 u64 thmac;
169 u8 hmac[MPTCPOPT_HMAC_LEN];
170 struct mptcp_addr_info addr;
171 struct mptcp_rm_list rm_list;
172 u64 ahmac;
173 u64 fail_seq;
174 };
175
mptcp_option(u8 subopt,u8 len,u8 nib,u8 field)176 static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field)
177 {
178 return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) |
179 ((nib & 0xF) << 8) | field);
180 }
181
182 enum mptcp_pm_status {
183 MPTCP_PM_ADD_ADDR_RECEIVED,
184 MPTCP_PM_ADD_ADDR_SEND_ACK,
185 MPTCP_PM_RM_ADDR_RECEIVED,
186 MPTCP_PM_ESTABLISHED,
187 MPTCP_PM_SUBFLOW_ESTABLISHED,
188 MPTCP_PM_ALREADY_ESTABLISHED, /* persistent status, set after ESTABLISHED event */
189 MPTCP_PM_MPC_ENDPOINT_ACCOUNTED /* persistent status, set after MPC local address is
190 * accounted int id_avail_bitmap
191 */
192 };
193
194 enum mptcp_pm_type {
195 MPTCP_PM_TYPE_KERNEL = 0,
196 MPTCP_PM_TYPE_USERSPACE,
197
198 __MPTCP_PM_TYPE_NR,
199 __MPTCP_PM_TYPE_MAX = __MPTCP_PM_TYPE_NR - 1,
200 };
201
202 /* Status bits below MPTCP_PM_ALREADY_ESTABLISHED need pm worker actions */
203 #define MPTCP_PM_WORK_MASK ((1 << MPTCP_PM_ALREADY_ESTABLISHED) - 1)
204
205 enum mptcp_addr_signal_status {
206 MPTCP_ADD_ADDR_SIGNAL,
207 MPTCP_ADD_ADDR_ECHO,
208 MPTCP_RM_ADDR_SIGNAL,
209 };
210
211 /* max value of mptcp_addr_info.id */
212 #define MPTCP_PM_MAX_ADDR_ID U8_MAX
213
214 struct mptcp_pm_data {
215 struct mptcp_addr_info local;
216 struct mptcp_addr_info remote;
217 struct list_head anno_list;
218 struct list_head userspace_pm_local_addr_list;
219
220 spinlock_t lock; /*protects the whole PM data */
221
222 u8 addr_signal;
223 bool server_side;
224 bool work_pending;
225 bool accept_addr;
226 bool accept_subflow;
227 bool remote_deny_join_id0;
228 u8 add_addr_signaled;
229 u8 add_addr_accepted;
230 u8 local_addr_used;
231 u8 pm_type;
232 u8 subflows;
233 u8 status;
234 DECLARE_BITMAP(id_avail_bitmap, MPTCP_PM_MAX_ADDR_ID + 1);
235 struct mptcp_rm_list rm_list_tx;
236 struct mptcp_rm_list rm_list_rx;
237 };
238
239 struct mptcp_pm_addr_entry {
240 struct list_head list;
241 struct mptcp_addr_info addr;
242 u8 flags;
243 int ifindex;
244 struct socket *lsk;
245 };
246
247 struct mptcp_data_frag {
248 struct list_head list;
249 u64 data_seq;
250 u16 data_len;
251 u16 offset;
252 u16 overhead;
253 u16 already_sent;
254 struct page *page;
255 };
256
257 /* MPTCP connection sock */
258 struct mptcp_sock {
259 /* inet_connection_sock must be the first member */
260 struct inet_connection_sock sk;
261 u64 local_key; /* protected by the first subflow socket lock
262 * lockless access read
263 */
264 u64 remote_key; /* same as above */
265 u64 write_seq;
266 u64 bytes_sent;
267 u64 snd_nxt;
268 u64 bytes_received;
269 u64 ack_seq;
270 atomic64_t rcv_wnd_sent;
271 u64 rcv_data_fin_seq;
272 u64 bytes_retrans;
273 u64 bytes_consumed;
274 int rmem_fwd_alloc;
275 int snd_burst;
276 int old_wspace;
277 u64 recovery_snd_nxt; /* in recovery mode accept up to this seq;
278 * recovery related fields are under data_lock
279 * protection
280 */
281 u64 bytes_acked;
282 u64 snd_una;
283 u64 wnd_end;
284 u32 last_data_sent;
285 u32 last_data_recv;
286 u32 last_ack_recv;
287 unsigned long timer_ival;
288 u32 token;
289 int rmem_released;
290 unsigned long flags;
291 unsigned long cb_flags;
292 bool recovery; /* closing subflow write queue reinjected */
293 bool can_ack;
294 bool fully_established;
295 bool rcv_data_fin;
296 bool snd_data_fin_enable;
297 bool rcv_fastclose;
298 bool use_64bit_ack; /* Set when we received a 64-bit DSN */
299 bool csum_enabled;
300 bool allow_infinite_fallback;
301 u8 pending_state; /* A subflow asked to set this sk_state,
302 * protected by the msk data lock
303 */
304 u8 mpc_endpoint_id;
305 u8 recvmsg_inq:1,
306 cork:1,
307 nodelay:1,
308 fastopening:1,
309 in_accept_queue:1,
310 free_first:1,
311 rcvspace_init:1;
312 u32 notsent_lowat;
313 int keepalive_cnt;
314 int keepalive_idle;
315 int keepalive_intvl;
316 struct work_struct work;
317 struct sk_buff *ooo_last_skb;
318 struct rb_root out_of_order_queue;
319 struct sk_buff_head receive_queue;
320 struct list_head conn_list;
321 struct list_head rtx_queue;
322 struct mptcp_data_frag *first_pending;
323 struct list_head join_list;
324 struct sock *first; /* The mptcp ops can safely dereference, using suitable
325 * ONCE annotation, the subflow outside the socket
326 * lock as such sock is freed after close().
327 */
328 struct mptcp_pm_data pm;
329 struct mptcp_sched_ops *sched;
330 struct {
331 u32 space; /* bytes copied in last measurement window */
332 u32 copied; /* bytes copied in this measurement window */
333 u64 time; /* start time of measurement window */
334 u64 rtt_us; /* last maximum rtt of subflows */
335 } rcvq_space;
336 u8 scaling_ratio;
337
338 u32 subflow_id;
339 u32 setsockopt_seq;
340 char ca_name[TCP_CA_NAME_MAX];
341 };
342
343 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock)
344 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock)
345
346 #define mptcp_for_each_subflow(__msk, __subflow) \
347 list_for_each_entry(__subflow, &((__msk)->conn_list), node)
348 #define mptcp_for_each_subflow_safe(__msk, __subflow, __tmp) \
349 list_for_each_entry_safe(__subflow, __tmp, &((__msk)->conn_list), node)
350
351 extern struct genl_family mptcp_genl_family;
352
msk_owned_by_me(const struct mptcp_sock * msk)353 static inline void msk_owned_by_me(const struct mptcp_sock *msk)
354 {
355 sock_owned_by_me((const struct sock *)msk);
356 }
357
358 #ifdef CONFIG_DEBUG_NET
359 /* MPTCP-specific: we might (indirectly) call this helper with the wrong sk */
360 #undef tcp_sk
361 #define tcp_sk(ptr) ({ \
362 typeof(ptr) _ptr = (ptr); \
363 WARN_ON(_ptr->sk_protocol != IPPROTO_TCP); \
364 container_of_const(_ptr, struct tcp_sock, inet_conn.icsk_inet.sk); \
365 })
366 #define mptcp_sk(ptr) ({ \
367 typeof(ptr) _ptr = (ptr); \
368 WARN_ON(_ptr->sk_protocol != IPPROTO_MPTCP); \
369 container_of_const(_ptr, struct mptcp_sock, sk.icsk_inet.sk); \
370 })
371
372 #else /* !CONFIG_DEBUG_NET */
373 #define mptcp_sk(ptr) container_of_const(ptr, struct mptcp_sock, sk.icsk_inet.sk)
374 #endif
375
376 /* the msk socket don't use the backlog, also account for the bulk
377 * free memory
378 */
__mptcp_rmem(const struct sock * sk)379 static inline int __mptcp_rmem(const struct sock *sk)
380 {
381 return atomic_read(&sk->sk_rmem_alloc) - READ_ONCE(mptcp_sk(sk)->rmem_released);
382 }
383
mptcp_win_from_space(const struct sock * sk,int space)384 static inline int mptcp_win_from_space(const struct sock *sk, int space)
385 {
386 return __tcp_win_from_space(mptcp_sk(sk)->scaling_ratio, space);
387 }
388
mptcp_space_from_win(const struct sock * sk,int win)389 static inline int mptcp_space_from_win(const struct sock *sk, int win)
390 {
391 return __tcp_space_from_win(mptcp_sk(sk)->scaling_ratio, win);
392 }
393
__mptcp_space(const struct sock * sk)394 static inline int __mptcp_space(const struct sock *sk)
395 {
396 return mptcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - __mptcp_rmem(sk));
397 }
398
mptcp_send_head(const struct sock * sk)399 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk)
400 {
401 const struct mptcp_sock *msk = mptcp_sk(sk);
402
403 return READ_ONCE(msk->first_pending);
404 }
405
mptcp_send_next(struct sock * sk)406 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk)
407 {
408 struct mptcp_sock *msk = mptcp_sk(sk);
409 struct mptcp_data_frag *cur;
410
411 cur = msk->first_pending;
412 return list_is_last(&cur->list, &msk->rtx_queue) ? NULL :
413 list_next_entry(cur, list);
414 }
415
mptcp_pending_tail(const struct sock * sk)416 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk)
417 {
418 const struct mptcp_sock *msk = mptcp_sk(sk);
419
420 if (!msk->first_pending)
421 return NULL;
422
423 if (WARN_ON_ONCE(list_empty(&msk->rtx_queue)))
424 return NULL;
425
426 return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list);
427 }
428
mptcp_rtx_head(struct sock * sk)429 static inline struct mptcp_data_frag *mptcp_rtx_head(struct sock *sk)
430 {
431 struct mptcp_sock *msk = mptcp_sk(sk);
432
433 if (msk->snd_una == msk->snd_nxt)
434 return NULL;
435
436 return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list);
437 }
438
439 struct csum_pseudo_header {
440 __be64 data_seq;
441 __be32 subflow_seq;
442 __be16 data_len;
443 __sum16 csum;
444 };
445
446 struct mptcp_subflow_request_sock {
447 struct tcp_request_sock sk;
448 u16 mp_capable : 1,
449 mp_join : 1,
450 backup : 1,
451 request_bkup : 1,
452 csum_reqd : 1,
453 allow_join_id0 : 1;
454 u8 local_id;
455 u8 remote_id;
456 u64 local_key;
457 u64 idsn;
458 u32 token;
459 u32 ssn_offset;
460 u64 thmac;
461 u32 local_nonce;
462 u32 remote_nonce;
463 struct mptcp_sock *msk;
464 struct hlist_nulls_node token_node;
465 };
466
467 static inline struct mptcp_subflow_request_sock *
mptcp_subflow_rsk(const struct request_sock * rsk)468 mptcp_subflow_rsk(const struct request_sock *rsk)
469 {
470 return (struct mptcp_subflow_request_sock *)rsk;
471 }
472
473 struct mptcp_delegated_action {
474 struct napi_struct napi;
475 struct list_head head;
476 };
477
478 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions);
479
480 #define MPTCP_DELEGATE_SCHEDULED 0
481 #define MPTCP_DELEGATE_SEND 1
482 #define MPTCP_DELEGATE_ACK 2
483 #define MPTCP_DELEGATE_SNDBUF 3
484
485 #define MPTCP_DELEGATE_ACTIONS_MASK (~BIT(MPTCP_DELEGATE_SCHEDULED))
486 /* MPTCP subflow context */
487 struct mptcp_subflow_context {
488 struct list_head node;/* conn_list of subflows */
489
490 struct_group(reset,
491
492 unsigned long avg_pacing_rate; /* protected by msk socket lock */
493 u64 local_key;
494 u64 remote_key;
495 u64 idsn;
496 u64 map_seq;
497 u32 snd_isn;
498 u32 token;
499 u32 rel_write_seq;
500 u32 map_subflow_seq;
501 u32 ssn_offset;
502 u32 map_data_len;
503 __wsum map_data_csum;
504 u32 map_csum_len;
505 u32 request_mptcp : 1, /* send MP_CAPABLE */
506 request_join : 1, /* send MP_JOIN */
507 request_bkup : 1,
508 mp_capable : 1, /* remote is MPTCP capable */
509 mp_join : 1, /* remote is JOINing */
510 fully_established : 1, /* path validated */
511 pm_notified : 1, /* PM hook called for established status */
512 conn_finished : 1,
513 map_valid : 1,
514 map_csum_reqd : 1,
515 map_data_fin : 1,
516 mpc_map : 1,
517 backup : 1,
518 send_mp_prio : 1,
519 send_mp_fail : 1,
520 send_fastclose : 1,
521 send_infinite_map : 1,
522 remote_key_valid : 1, /* received the peer key from */
523 disposable : 1, /* ctx can be free at ulp release time */
524 stale : 1, /* unable to snd/rcv data, do not use for xmit */
525 valid_csum_seen : 1, /* at least one csum validated */
526 is_mptfo : 1, /* subflow is doing TFO */
527 __unused : 10;
528 bool data_avail;
529 bool scheduled;
530 u32 remote_nonce;
531 u64 thmac;
532 u32 local_nonce;
533 u32 remote_token;
534 union {
535 u8 hmac[MPTCPOPT_HMAC_LEN]; /* MPJ subflow only */
536 u64 iasn; /* initial ack sequence number, MPC subflows only */
537 };
538 s16 local_id; /* if negative not initialized yet */
539 u8 remote_id;
540 u8 reset_seen:1;
541 u8 reset_transient:1;
542 u8 reset_reason:4;
543 u8 stale_count;
544
545 u32 subflow_id;
546
547 long delegated_status;
548 unsigned long fail_tout;
549
550 );
551
552 struct list_head delegated_node; /* link into delegated_action, protected by local BH */
553
554 u32 setsockopt_seq;
555 u32 stale_rcv_tstamp;
556 int cached_sndbuf; /* sndbuf size when last synced with the msk sndbuf,
557 * protected by the msk socket lock
558 */
559
560 struct sock *tcp_sock; /* tcp sk backpointer */
561 struct sock *conn; /* parent mptcp_sock */
562 const struct inet_connection_sock_af_ops *icsk_af_ops;
563 void (*tcp_state_change)(struct sock *sk);
564 void (*tcp_error_report)(struct sock *sk);
565
566 struct rcu_head rcu;
567 };
568
569 static inline struct mptcp_subflow_context *
mptcp_subflow_ctx(const struct sock * sk)570 mptcp_subflow_ctx(const struct sock *sk)
571 {
572 const struct inet_connection_sock *icsk = inet_csk(sk);
573
574 /* Use RCU on icsk_ulp_data only for sock diag code */
575 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data;
576 }
577
578 static inline struct sock *
mptcp_subflow_tcp_sock(const struct mptcp_subflow_context * subflow)579 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow)
580 {
581 return subflow->tcp_sock;
582 }
583
584 static inline void
mptcp_subflow_ctx_reset(struct mptcp_subflow_context * subflow)585 mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow)
586 {
587 memset(&subflow->reset, 0, sizeof(subflow->reset));
588 subflow->request_mptcp = 1;
589 WRITE_ONCE(subflow->local_id, -1);
590 }
591
592 /* Convert reset reasons in MPTCP to enum sk_rst_reason type */
593 static inline enum sk_rst_reason
sk_rst_convert_mptcp_reason(u32 reason)594 sk_rst_convert_mptcp_reason(u32 reason)
595 {
596 switch (reason) {
597 case MPTCP_RST_EUNSPEC:
598 return SK_RST_REASON_MPTCP_RST_EUNSPEC;
599 case MPTCP_RST_EMPTCP:
600 return SK_RST_REASON_MPTCP_RST_EMPTCP;
601 case MPTCP_RST_ERESOURCE:
602 return SK_RST_REASON_MPTCP_RST_ERESOURCE;
603 case MPTCP_RST_EPROHIBIT:
604 return SK_RST_REASON_MPTCP_RST_EPROHIBIT;
605 case MPTCP_RST_EWQ2BIG:
606 return SK_RST_REASON_MPTCP_RST_EWQ2BIG;
607 case MPTCP_RST_EBADPERF:
608 return SK_RST_REASON_MPTCP_RST_EBADPERF;
609 case MPTCP_RST_EMIDDLEBOX:
610 return SK_RST_REASON_MPTCP_RST_EMIDDLEBOX;
611 default:
612 /* It should not happen, or else errors may occur
613 * in MPTCP layer
614 */
615 return SK_RST_REASON_ERROR;
616 }
617 }
618
619 static inline void
mptcp_send_active_reset_reason(struct sock * sk)620 mptcp_send_active_reset_reason(struct sock *sk)
621 {
622 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
623 enum sk_rst_reason reason;
624
625 reason = sk_rst_convert_mptcp_reason(subflow->reset_reason);
626 tcp_send_active_reset(sk, GFP_ATOMIC, reason);
627 }
628
629 static inline u64
mptcp_subflow_get_map_offset(const struct mptcp_subflow_context * subflow)630 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow)
631 {
632 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq -
633 subflow->ssn_offset -
634 subflow->map_subflow_seq;
635 }
636
637 static inline u64
mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context * subflow)638 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow)
639 {
640 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow);
641 }
642
643 void mptcp_subflow_process_delegated(struct sock *ssk, long actions);
644
mptcp_subflow_delegate(struct mptcp_subflow_context * subflow,int action)645 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action)
646 {
647 long old, set_bits = BIT(MPTCP_DELEGATE_SCHEDULED) | BIT(action);
648 struct mptcp_delegated_action *delegated;
649 bool schedule;
650
651 /* the caller held the subflow bh socket lock */
652 lockdep_assert_in_softirq();
653
654 /* The implied barrier pairs with tcp_release_cb_override()
655 * mptcp_napi_poll(), and ensures the below list check sees list
656 * updates done prior to delegated status bits changes
657 */
658 old = set_mask_bits(&subflow->delegated_status, 0, set_bits);
659 if (!(old & BIT(MPTCP_DELEGATE_SCHEDULED))) {
660 if (WARN_ON_ONCE(!list_empty(&subflow->delegated_node)))
661 return;
662
663 delegated = this_cpu_ptr(&mptcp_delegated_actions);
664 schedule = list_empty(&delegated->head);
665 list_add_tail(&subflow->delegated_node, &delegated->head);
666 sock_hold(mptcp_subflow_tcp_sock(subflow));
667 if (schedule)
668 napi_schedule(&delegated->napi);
669 }
670 }
671
672 static inline struct mptcp_subflow_context *
mptcp_subflow_delegated_next(struct mptcp_delegated_action * delegated)673 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated)
674 {
675 struct mptcp_subflow_context *ret;
676
677 if (list_empty(&delegated->head))
678 return NULL;
679
680 ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node);
681 list_del_init(&ret->delegated_node);
682 return ret;
683 }
684
685 int mptcp_is_enabled(const struct net *net);
686 unsigned int mptcp_get_add_addr_timeout(const struct net *net);
687 int mptcp_is_checksum_enabled(const struct net *net);
688 int mptcp_allow_join_id0(const struct net *net);
689 unsigned int mptcp_stale_loss_cnt(const struct net *net);
690 unsigned int mptcp_close_timeout(const struct sock *sk);
691 int mptcp_get_pm_type(const struct net *net);
692 const char *mptcp_get_scheduler(const struct net *net);
693 void mptcp_get_available_schedulers(char *buf, size_t maxlen);
694 void __mptcp_subflow_fully_established(struct mptcp_sock *msk,
695 struct mptcp_subflow_context *subflow,
696 const struct mptcp_options_received *mp_opt);
697 bool __mptcp_retransmit_pending_data(struct sock *sk);
698 void mptcp_check_and_set_pending(struct sock *sk);
699 void __mptcp_push_pending(struct sock *sk, unsigned int flags);
700 bool mptcp_subflow_data_available(struct sock *sk);
701 void __init mptcp_subflow_init(void);
702 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how);
703 void mptcp_close_ssk(struct sock *sk, struct sock *ssk,
704 struct mptcp_subflow_context *subflow);
705 void __mptcp_subflow_send_ack(struct sock *ssk);
706 void mptcp_subflow_reset(struct sock *ssk);
707 void mptcp_subflow_queue_clean(struct sock *sk, struct sock *ssk);
708 void mptcp_sock_graft(struct sock *sk, struct socket *parent);
709 struct sock *__mptcp_nmpc_sk(struct mptcp_sock *msk);
710 bool __mptcp_close(struct sock *sk, long timeout);
711 void mptcp_cancel_work(struct sock *sk);
712 void __mptcp_unaccepted_force_close(struct sock *sk);
713 void mptcp_set_owner_r(struct sk_buff *skb, struct sock *sk);
714 void mptcp_set_state(struct sock *sk, int state);
715
716 bool mptcp_addresses_equal(const struct mptcp_addr_info *a,
717 const struct mptcp_addr_info *b, bool use_port);
718 void mptcp_local_address(const struct sock_common *skc, struct mptcp_addr_info *addr);
719
720 /* called with sk socket lock held */
721 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc,
722 const struct mptcp_addr_info *remote);
723 int mptcp_subflow_create_socket(struct sock *sk, unsigned short family,
724 struct socket **new_sock);
725 void mptcp_info2sockaddr(const struct mptcp_addr_info *info,
726 struct sockaddr_storage *addr,
727 unsigned short family);
728 struct mptcp_sched_ops *mptcp_sched_find(const char *name);
729 int mptcp_register_scheduler(struct mptcp_sched_ops *sched);
730 void mptcp_unregister_scheduler(struct mptcp_sched_ops *sched);
731 void mptcp_sched_init(void);
732 int mptcp_init_sched(struct mptcp_sock *msk,
733 struct mptcp_sched_ops *sched);
734 void mptcp_release_sched(struct mptcp_sock *msk);
735 void mptcp_subflow_set_scheduled(struct mptcp_subflow_context *subflow,
736 bool scheduled);
737 struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk);
738 struct sock *mptcp_subflow_get_retrans(struct mptcp_sock *msk);
739 int mptcp_sched_get_send(struct mptcp_sock *msk);
740 int mptcp_sched_get_retrans(struct mptcp_sock *msk);
741
mptcp_data_avail(const struct mptcp_sock * msk)742 static inline u64 mptcp_data_avail(const struct mptcp_sock *msk)
743 {
744 return READ_ONCE(msk->bytes_received) - READ_ONCE(msk->bytes_consumed);
745 }
746
mptcp_epollin_ready(const struct sock * sk)747 static inline bool mptcp_epollin_ready(const struct sock *sk)
748 {
749 /* mptcp doesn't have to deal with small skbs in the receive queue,
750 * at it can always coalesce them
751 */
752 return (mptcp_data_avail(mptcp_sk(sk)) >= sk->sk_rcvlowat) ||
753 (mem_cgroup_sockets_enabled && sk->sk_memcg &&
754 mem_cgroup_under_socket_pressure(sk->sk_memcg)) ||
755 READ_ONCE(tcp_memory_pressure);
756 }
757
758 int mptcp_set_rcvlowat(struct sock *sk, int val);
759
__tcp_can_send(const struct sock * ssk)760 static inline bool __tcp_can_send(const struct sock *ssk)
761 {
762 /* only send if our side has not closed yet */
763 return ((1 << inet_sk_state_load(ssk)) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT));
764 }
765
__mptcp_subflow_active(struct mptcp_subflow_context * subflow)766 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow)
767 {
768 /* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */
769 if (subflow->request_join && !subflow->fully_established)
770 return false;
771
772 return __tcp_can_send(mptcp_subflow_tcp_sock(subflow));
773 }
774
775 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow);
776
777 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow);
778
779 void mptcp_subflow_drop_ctx(struct sock *ssk);
780
mptcp_subflow_tcp_fallback(struct sock * sk,struct mptcp_subflow_context * ctx)781 static inline void mptcp_subflow_tcp_fallback(struct sock *sk,
782 struct mptcp_subflow_context *ctx)
783 {
784 sk->sk_data_ready = sock_def_readable;
785 sk->sk_state_change = ctx->tcp_state_change;
786 sk->sk_write_space = sk_stream_write_space;
787 sk->sk_error_report = ctx->tcp_error_report;
788
789 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops;
790 }
791
792 void __init mptcp_proto_init(void);
793 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
794 int __init mptcp_proto_v6_init(void);
795 #endif
796
797 struct sock *mptcp_sk_clone_init(const struct sock *sk,
798 const struct mptcp_options_received *mp_opt,
799 struct sock *ssk,
800 struct request_sock *req);
801 void mptcp_get_options(const struct sk_buff *skb,
802 struct mptcp_options_received *mp_opt);
803
804 void mptcp_finish_connect(struct sock *sk);
805 void __mptcp_sync_state(struct sock *sk, int state);
806 void mptcp_reset_tout_timer(struct mptcp_sock *msk, unsigned long fail_tout);
807
mptcp_stop_tout_timer(struct sock * sk)808 static inline void mptcp_stop_tout_timer(struct sock *sk)
809 {
810 if (!inet_csk(sk)->icsk_mtup.probe_timestamp)
811 return;
812
813 sk_stop_timer(sk, &sk->sk_timer);
814 inet_csk(sk)->icsk_mtup.probe_timestamp = 0;
815 }
816
mptcp_set_close_tout(struct sock * sk,unsigned long tout)817 static inline void mptcp_set_close_tout(struct sock *sk, unsigned long tout)
818 {
819 /* avoid 0 timestamp, as that means no close timeout */
820 inet_csk(sk)->icsk_mtup.probe_timestamp = tout ? : 1;
821 }
822
mptcp_start_tout_timer(struct sock * sk)823 static inline void mptcp_start_tout_timer(struct sock *sk)
824 {
825 mptcp_set_close_tout(sk, tcp_jiffies32);
826 mptcp_reset_tout_timer(mptcp_sk(sk), 0);
827 }
828
mptcp_is_fully_established(struct sock * sk)829 static inline bool mptcp_is_fully_established(struct sock *sk)
830 {
831 return inet_sk_state_load(sk) == TCP_ESTABLISHED &&
832 READ_ONCE(mptcp_sk(sk)->fully_established);
833 }
834
835 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk);
836 void mptcp_data_ready(struct sock *sk, struct sock *ssk);
837 bool mptcp_finish_join(struct sock *sk);
838 bool mptcp_schedule_work(struct sock *sk);
839 int mptcp_setsockopt(struct sock *sk, int level, int optname,
840 sockptr_t optval, unsigned int optlen);
841 int mptcp_getsockopt(struct sock *sk, int level, int optname,
842 char __user *optval, int __user *option);
843
844 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq);
mptcp_expand_seq(u64 old_seq,u64 cur_seq,bool use_64bit)845 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit)
846 {
847 if (use_64bit)
848 return cur_seq;
849
850 return __mptcp_expand_seq(old_seq, cur_seq);
851 }
852 void __mptcp_check_push(struct sock *sk, struct sock *ssk);
853 void __mptcp_data_acked(struct sock *sk);
854 void __mptcp_error_report(struct sock *sk);
855 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)856 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk)
857 {
858 return READ_ONCE(msk->snd_data_fin_enable) &&
859 READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt);
860 }
861
mptcp_notsent_lowat(const struct sock * sk)862 static inline u32 mptcp_notsent_lowat(const struct sock *sk)
863 {
864 struct net *net = sock_net(sk);
865 u32 val;
866
867 val = READ_ONCE(mptcp_sk(sk)->notsent_lowat);
868 return val ?: READ_ONCE(net->ipv4.sysctl_tcp_notsent_lowat);
869 }
870
mptcp_stream_memory_free(const struct sock * sk,int wake)871 static inline bool mptcp_stream_memory_free(const struct sock *sk, int wake)
872 {
873 const struct mptcp_sock *msk = mptcp_sk(sk);
874 u32 notsent_bytes;
875
876 notsent_bytes = READ_ONCE(msk->write_seq) - READ_ONCE(msk->snd_nxt);
877 return (notsent_bytes << wake) < mptcp_notsent_lowat(sk);
878 }
879
__mptcp_stream_is_writeable(const struct sock * sk,int wake)880 static inline bool __mptcp_stream_is_writeable(const struct sock *sk, int wake)
881 {
882 return mptcp_stream_memory_free(sk, wake) &&
883 __sk_stream_is_writeable(sk, wake);
884 }
885
mptcp_write_space(struct sock * sk)886 static inline void mptcp_write_space(struct sock *sk)
887 {
888 /* pairs with memory barrier in mptcp_poll */
889 smp_mb();
890 if (mptcp_stream_memory_free(sk, 1))
891 sk_stream_write_space(sk);
892 }
893
__mptcp_sync_sndbuf(struct sock * sk)894 static inline void __mptcp_sync_sndbuf(struct sock *sk)
895 {
896 struct mptcp_subflow_context *subflow;
897 int ssk_sndbuf, new_sndbuf;
898
899 if (sk->sk_userlocks & SOCK_SNDBUF_LOCK)
900 return;
901
902 new_sndbuf = READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_wmem[0]);
903 mptcp_for_each_subflow(mptcp_sk(sk), subflow) {
904 ssk_sndbuf = READ_ONCE(mptcp_subflow_tcp_sock(subflow)->sk_sndbuf);
905
906 subflow->cached_sndbuf = ssk_sndbuf;
907 new_sndbuf += ssk_sndbuf;
908 }
909
910 /* the msk max wmem limit is <nr_subflows> * tcp wmem[2] */
911 WRITE_ONCE(sk->sk_sndbuf, new_sndbuf);
912 mptcp_write_space(sk);
913 }
914
915 /* The called held both the msk socket and the subflow socket locks,
916 * possibly under BH
917 */
__mptcp_propagate_sndbuf(struct sock * sk,struct sock * ssk)918 static inline void __mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
919 {
920 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
921
922 if (READ_ONCE(ssk->sk_sndbuf) != subflow->cached_sndbuf)
923 __mptcp_sync_sndbuf(sk);
924 }
925
926 /* the caller held only the subflow socket lock, either in process or
927 * BH context. Additionally this can be called under the msk data lock,
928 * so we can't acquire such lock here: let the delegate action acquires
929 * the needed locks in suitable order.
930 */
mptcp_propagate_sndbuf(struct sock * sk,struct sock * ssk)931 static inline void mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
932 {
933 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
934
935 if (likely(READ_ONCE(ssk->sk_sndbuf) == subflow->cached_sndbuf))
936 return;
937
938 local_bh_disable();
939 mptcp_subflow_delegate(subflow, MPTCP_DELEGATE_SNDBUF);
940 local_bh_enable();
941 }
942
943 void mptcp_destroy_common(struct mptcp_sock *msk, unsigned int flags);
944
945 #define MPTCP_TOKEN_MAX_RETRIES 4
946
947 void __init mptcp_token_init(void);
mptcp_token_init_request(struct request_sock * req)948 static inline void mptcp_token_init_request(struct request_sock *req)
949 {
950 mptcp_subflow_rsk(req)->token_node.pprev = NULL;
951 }
952
953 int mptcp_token_new_request(struct request_sock *req);
954 void mptcp_token_destroy_request(struct request_sock *req);
955 int mptcp_token_new_connect(struct sock *ssk);
956 void mptcp_token_accept(struct mptcp_subflow_request_sock *r,
957 struct mptcp_sock *msk);
958 bool mptcp_token_exists(u32 token);
959 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token);
960 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot,
961 long *s_num);
962 void mptcp_token_destroy(struct mptcp_sock *msk);
963
964 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn);
965
966 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac);
967 __sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum);
968
969 void __init mptcp_pm_init(void);
970 void mptcp_pm_data_init(struct mptcp_sock *msk);
971 void mptcp_pm_data_reset(struct mptcp_sock *msk);
972 int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info,
973 struct mptcp_addr_info *addr);
974 int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info,
975 bool require_family,
976 struct mptcp_pm_addr_entry *entry);
977 bool mptcp_pm_addr_families_match(const struct sock *sk,
978 const struct mptcp_addr_info *loc,
979 const struct mptcp_addr_info *rem);
980 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
981 void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
982 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side);
983 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk);
984 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk);
985 void mptcp_pm_connection_closed(struct mptcp_sock *msk);
986 void mptcp_pm_subflow_established(struct mptcp_sock *msk);
987 bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk);
988 void mptcp_pm_subflow_check_next(struct mptcp_sock *msk,
989 const struct mptcp_subflow_context *subflow);
990 void mptcp_pm_add_addr_received(const struct sock *ssk,
991 const struct mptcp_addr_info *addr);
992 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk,
993 const struct mptcp_addr_info *addr);
994 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk);
995 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk);
996 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk,
997 const struct mptcp_rm_list *rm_list);
998 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup);
999 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq);
1000 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk,
1001 struct mptcp_addr_info *addr,
1002 struct mptcp_addr_info *rem,
1003 u8 bkup);
1004 bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk,
1005 const struct mptcp_addr_info *addr);
1006 void mptcp_pm_free_anno_list(struct mptcp_sock *msk);
1007 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk);
1008 struct mptcp_pm_add_entry *
1009 mptcp_pm_del_add_timer(struct mptcp_sock *msk,
1010 const struct mptcp_addr_info *addr, bool check_id);
1011 struct mptcp_pm_add_entry *
1012 mptcp_lookup_anno_list_by_saddr(const struct mptcp_sock *msk,
1013 const struct mptcp_addr_info *addr);
1014 int mptcp_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk,
1015 unsigned int id,
1016 u8 *flags, int *ifindex);
1017 int mptcp_pm_nl_get_flags_and_ifindex_by_id(struct mptcp_sock *msk, unsigned int id,
1018 u8 *flags, int *ifindex);
1019 int mptcp_userspace_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk,
1020 unsigned int id,
1021 u8 *flags, int *ifindex);
1022 int mptcp_pm_set_flags(struct sk_buff *skb, struct genl_info *info);
1023 int mptcp_pm_nl_set_flags(struct sk_buff *skb, struct genl_info *info);
1024 int mptcp_userspace_pm_set_flags(struct sk_buff *skb, struct genl_info *info);
1025 int mptcp_pm_announce_addr(struct mptcp_sock *msk,
1026 const struct mptcp_addr_info *addr,
1027 bool echo);
1028 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
1029 int mptcp_pm_remove_subflow(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
1030 void mptcp_pm_remove_addrs(struct mptcp_sock *msk, struct list_head *rm_list);
1031
1032 void mptcp_free_local_addr_list(struct mptcp_sock *msk);
1033
1034 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk,
1035 const struct sock *ssk, gfp_t gfp);
1036 void mptcp_event_addr_announced(const struct sock *ssk, const struct mptcp_addr_info *info);
1037 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id);
1038 void mptcp_event_pm_listener(const struct sock *ssk,
1039 enum mptcp_event_type event);
1040 bool mptcp_userspace_pm_active(const struct mptcp_sock *msk);
1041
1042 void __mptcp_fastopen_gen_msk_ackseq(struct mptcp_sock *msk, struct mptcp_subflow_context *subflow,
1043 const struct mptcp_options_received *mp_opt);
1044 void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subflow,
1045 struct request_sock *req);
1046 int mptcp_nl_fill_addr(struct sk_buff *skb,
1047 struct mptcp_pm_addr_entry *entry);
1048
mptcp_pm_should_add_signal(struct mptcp_sock * msk)1049 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk)
1050 {
1051 return READ_ONCE(msk->pm.addr_signal) &
1052 (BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO));
1053 }
1054
mptcp_pm_should_add_signal_addr(struct mptcp_sock * msk)1055 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk)
1056 {
1057 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL);
1058 }
1059
mptcp_pm_should_add_signal_echo(struct mptcp_sock * msk)1060 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk)
1061 {
1062 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO);
1063 }
1064
mptcp_pm_should_rm_signal(struct mptcp_sock * msk)1065 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk)
1066 {
1067 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL);
1068 }
1069
mptcp_pm_is_userspace(const struct mptcp_sock * msk)1070 static inline bool mptcp_pm_is_userspace(const struct mptcp_sock *msk)
1071 {
1072 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_USERSPACE;
1073 }
1074
mptcp_pm_is_kernel(const struct mptcp_sock * msk)1075 static inline bool mptcp_pm_is_kernel(const struct mptcp_sock *msk)
1076 {
1077 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_KERNEL;
1078 }
1079
mptcp_add_addr_len(int family,bool echo,bool port)1080 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port)
1081 {
1082 u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE;
1083
1084 if (family == AF_INET6)
1085 len = TCPOLEN_MPTCP_ADD_ADDR6_BASE;
1086 if (!echo)
1087 len += MPTCPOPT_THMAC_LEN;
1088 /* account for 2 trailing 'nop' options */
1089 if (port)
1090 len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN;
1091
1092 return len;
1093 }
1094
mptcp_rm_addr_len(const struct mptcp_rm_list * rm_list)1095 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list)
1096 {
1097 if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX)
1098 return -EINVAL;
1099
1100 return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1;
1101 }
1102
1103 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, const struct sk_buff *skb,
1104 unsigned int opt_size, unsigned int remaining,
1105 struct mptcp_addr_info *addr, bool *echo,
1106 bool *drop_other_suboptions);
1107 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining,
1108 struct mptcp_rm_list *rm_list);
1109 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
1110 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1111 int mptcp_userspace_pm_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1112 bool mptcp_pm_is_backup(struct mptcp_sock *msk, struct sock_common *skc);
1113 bool mptcp_pm_nl_is_backup(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1114 bool mptcp_userspace_pm_is_backup(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1115 int mptcp_pm_dump_addr(struct sk_buff *msg, struct netlink_callback *cb);
1116 int mptcp_pm_nl_dump_addr(struct sk_buff *msg,
1117 struct netlink_callback *cb);
1118 int mptcp_userspace_pm_dump_addr(struct sk_buff *msg,
1119 struct netlink_callback *cb);
1120 int mptcp_pm_get_addr(struct sk_buff *skb, struct genl_info *info);
1121 int mptcp_pm_nl_get_addr(struct sk_buff *skb, struct genl_info *info);
1122 int mptcp_userspace_pm_get_addr(struct sk_buff *skb,
1123 struct genl_info *info);
1124
subflow_get_local_id(const struct mptcp_subflow_context * subflow)1125 static inline u8 subflow_get_local_id(const struct mptcp_subflow_context *subflow)
1126 {
1127 int local_id = READ_ONCE(subflow->local_id);
1128
1129 if (local_id < 0)
1130 return 0;
1131 return local_id;
1132 }
1133
1134 void __init mptcp_pm_nl_init(void);
1135 void mptcp_pm_nl_work(struct mptcp_sock *msk);
1136 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk,
1137 const struct mptcp_rm_list *rm_list);
1138 unsigned int mptcp_pm_get_add_addr_signal_max(const struct mptcp_sock *msk);
1139 unsigned int mptcp_pm_get_add_addr_accept_max(const struct mptcp_sock *msk);
1140 unsigned int mptcp_pm_get_subflows_max(const struct mptcp_sock *msk);
1141 unsigned int mptcp_pm_get_local_addr_max(const struct mptcp_sock *msk);
1142
1143 /* called under PM lock */
__mptcp_pm_close_subflow(struct mptcp_sock * msk)1144 static inline void __mptcp_pm_close_subflow(struct mptcp_sock *msk)
1145 {
1146 if (--msk->pm.subflows < mptcp_pm_get_subflows_max(msk))
1147 WRITE_ONCE(msk->pm.accept_subflow, true);
1148 }
1149
mptcp_pm_close_subflow(struct mptcp_sock * msk)1150 static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk)
1151 {
1152 spin_lock_bh(&msk->pm.lock);
1153 __mptcp_pm_close_subflow(msk);
1154 spin_unlock_bh(&msk->pm.lock);
1155 }
1156
1157 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk);
1158 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk);
1159
mptcp_get_ext(const struct sk_buff * skb)1160 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb)
1161 {
1162 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP);
1163 }
1164
1165 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops);
1166
__mptcp_check_fallback(const struct mptcp_sock * msk)1167 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk)
1168 {
1169 return test_bit(MPTCP_FALLBACK_DONE, &msk->flags);
1170 }
1171
mptcp_check_fallback(const struct sock * sk)1172 static inline bool mptcp_check_fallback(const struct sock *sk)
1173 {
1174 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1175 struct mptcp_sock *msk = mptcp_sk(subflow->conn);
1176
1177 return __mptcp_check_fallback(msk);
1178 }
1179
__mptcp_do_fallback(struct mptcp_sock * msk)1180 static inline void __mptcp_do_fallback(struct mptcp_sock *msk)
1181 {
1182 if (__mptcp_check_fallback(msk)) {
1183 pr_debug("TCP fallback already done (msk=%p)", msk);
1184 return;
1185 }
1186 set_bit(MPTCP_FALLBACK_DONE, &msk->flags);
1187 }
1188
__mptcp_has_initial_subflow(const struct mptcp_sock * msk)1189 static inline bool __mptcp_has_initial_subflow(const struct mptcp_sock *msk)
1190 {
1191 struct sock *ssk = READ_ONCE(msk->first);
1192
1193 return ssk && ((1 << inet_sk_state_load(ssk)) &
1194 (TCPF_ESTABLISHED | TCPF_SYN_SENT |
1195 TCPF_SYN_RECV | TCPF_LISTEN));
1196 }
1197
mptcp_do_fallback(struct sock * ssk)1198 static inline void mptcp_do_fallback(struct sock *ssk)
1199 {
1200 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
1201 struct sock *sk = subflow->conn;
1202 struct mptcp_sock *msk;
1203
1204 msk = mptcp_sk(sk);
1205 __mptcp_do_fallback(msk);
1206 if (READ_ONCE(msk->snd_data_fin_enable) && !(ssk->sk_shutdown & SEND_SHUTDOWN)) {
1207 gfp_t saved_allocation = ssk->sk_allocation;
1208
1209 /* we are in a atomic (BH) scope, override ssk default for data
1210 * fin allocation
1211 */
1212 ssk->sk_allocation = GFP_ATOMIC;
1213 ssk->sk_shutdown |= SEND_SHUTDOWN;
1214 tcp_shutdown(ssk, SEND_SHUTDOWN);
1215 ssk->sk_allocation = saved_allocation;
1216 }
1217 }
1218
1219 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a)
1220
mptcp_check_infinite_map(struct sk_buff * skb)1221 static inline bool mptcp_check_infinite_map(struct sk_buff *skb)
1222 {
1223 struct mptcp_ext *mpext;
1224
1225 mpext = skb ? mptcp_get_ext(skb) : NULL;
1226 if (mpext && mpext->infinite_map)
1227 return true;
1228
1229 return false;
1230 }
1231
is_active_ssk(struct mptcp_subflow_context * subflow)1232 static inline bool is_active_ssk(struct mptcp_subflow_context *subflow)
1233 {
1234 return (subflow->request_mptcp || subflow->request_join);
1235 }
1236
subflow_simultaneous_connect(struct sock * sk)1237 static inline bool subflow_simultaneous_connect(struct sock *sk)
1238 {
1239 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1240
1241 return (1 << sk->sk_state) &
1242 (TCPF_ESTABLISHED | TCPF_FIN_WAIT1 | TCPF_FIN_WAIT2 | TCPF_CLOSING) &&
1243 is_active_ssk(subflow) &&
1244 !subflow->conn_finished;
1245 }
1246
1247 #ifdef CONFIG_SYN_COOKIES
1248 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1249 struct sk_buff *skb);
1250 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1251 struct sk_buff *skb);
1252 void __init mptcp_join_cookie_init(void);
1253 #else
1254 static inline void
subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock * subflow_req,struct sk_buff * skb)1255 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1256 struct sk_buff *skb) {}
1257 static inline bool
mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock * subflow_req,struct sk_buff * skb)1258 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1259 struct sk_buff *skb)
1260 {
1261 return false;
1262 }
1263
mptcp_join_cookie_init(void)1264 static inline void mptcp_join_cookie_init(void) {}
1265 #endif
1266
1267 #endif /* __MPTCP_PROTOCOL_H */
1268