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(const struct sock * sk)389 static inline int __mptcp_space(const struct sock *sk)
390 {
391 return mptcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - __mptcp_rmem(sk));
392 }
393
mptcp_send_head(const struct sock * sk)394 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk)
395 {
396 const struct mptcp_sock *msk = mptcp_sk(sk);
397
398 return READ_ONCE(msk->first_pending);
399 }
400
mptcp_send_next(struct sock * sk)401 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk)
402 {
403 struct mptcp_sock *msk = mptcp_sk(sk);
404 struct mptcp_data_frag *cur;
405
406 cur = msk->first_pending;
407 return list_is_last(&cur->list, &msk->rtx_queue) ? NULL :
408 list_next_entry(cur, list);
409 }
410
mptcp_pending_tail(const struct sock * sk)411 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk)
412 {
413 const struct mptcp_sock *msk = mptcp_sk(sk);
414
415 if (!msk->first_pending)
416 return NULL;
417
418 if (WARN_ON_ONCE(list_empty(&msk->rtx_queue)))
419 return NULL;
420
421 return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list);
422 }
423
mptcp_rtx_head(struct sock * sk)424 static inline struct mptcp_data_frag *mptcp_rtx_head(struct sock *sk)
425 {
426 struct mptcp_sock *msk = mptcp_sk(sk);
427
428 if (msk->snd_una == msk->snd_nxt)
429 return NULL;
430
431 return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list);
432 }
433
434 struct csum_pseudo_header {
435 __be64 data_seq;
436 __be32 subflow_seq;
437 __be16 data_len;
438 __sum16 csum;
439 };
440
441 struct mptcp_subflow_request_sock {
442 struct tcp_request_sock sk;
443 u16 mp_capable : 1,
444 mp_join : 1,
445 backup : 1,
446 csum_reqd : 1,
447 allow_join_id0 : 1;
448 u8 local_id;
449 u8 remote_id;
450 u64 local_key;
451 u64 idsn;
452 u32 token;
453 u32 ssn_offset;
454 u64 thmac;
455 u32 local_nonce;
456 u32 remote_nonce;
457 struct mptcp_sock *msk;
458 struct hlist_nulls_node token_node;
459 };
460
461 static inline struct mptcp_subflow_request_sock *
mptcp_subflow_rsk(const struct request_sock * rsk)462 mptcp_subflow_rsk(const struct request_sock *rsk)
463 {
464 return (struct mptcp_subflow_request_sock *)rsk;
465 }
466
467 struct mptcp_delegated_action {
468 struct napi_struct napi;
469 struct list_head head;
470 };
471
472 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions);
473
474 #define MPTCP_DELEGATE_SCHEDULED 0
475 #define MPTCP_DELEGATE_SEND 1
476 #define MPTCP_DELEGATE_ACK 2
477 #define MPTCP_DELEGATE_SNDBUF 3
478
479 #define MPTCP_DELEGATE_ACTIONS_MASK (~BIT(MPTCP_DELEGATE_SCHEDULED))
480 /* MPTCP subflow context */
481 struct mptcp_subflow_context {
482 struct list_head node;/* conn_list of subflows */
483
484 struct_group(reset,
485
486 unsigned long avg_pacing_rate; /* protected by msk socket lock */
487 u64 local_key;
488 u64 remote_key;
489 u64 idsn;
490 u64 map_seq;
491 u32 snd_isn;
492 u32 token;
493 u32 rel_write_seq;
494 u32 map_subflow_seq;
495 u32 ssn_offset;
496 u32 map_data_len;
497 __wsum map_data_csum;
498 u32 map_csum_len;
499 u32 request_mptcp : 1, /* send MP_CAPABLE */
500 request_join : 1, /* send MP_JOIN */
501 request_bkup : 1,
502 mp_capable : 1, /* remote is MPTCP capable */
503 mp_join : 1, /* remote is JOINing */
504 fully_established : 1, /* path validated */
505 pm_notified : 1, /* PM hook called for established status */
506 conn_finished : 1,
507 map_valid : 1,
508 map_csum_reqd : 1,
509 map_data_fin : 1,
510 mpc_map : 1,
511 backup : 1,
512 send_mp_prio : 1,
513 send_mp_fail : 1,
514 send_fastclose : 1,
515 send_infinite_map : 1,
516 remote_key_valid : 1, /* received the peer key from */
517 disposable : 1, /* ctx can be free at ulp release time */
518 stale : 1, /* unable to snd/rcv data, do not use for xmit */
519 valid_csum_seen : 1, /* at least one csum validated */
520 is_mptfo : 1, /* subflow is doing TFO */
521 __unused : 10;
522 bool data_avail;
523 bool scheduled;
524 u32 remote_nonce;
525 u64 thmac;
526 u32 local_nonce;
527 u32 remote_token;
528 union {
529 u8 hmac[MPTCPOPT_HMAC_LEN]; /* MPJ subflow only */
530 u64 iasn; /* initial ack sequence number, MPC subflows only */
531 };
532 s16 local_id; /* if negative not initialized yet */
533 u8 remote_id;
534 u8 reset_seen:1;
535 u8 reset_transient:1;
536 u8 reset_reason:4;
537 u8 stale_count;
538
539 u32 subflow_id;
540
541 long delegated_status;
542 unsigned long fail_tout;
543
544 );
545
546 struct list_head delegated_node; /* link into delegated_action, protected by local BH */
547
548 u32 setsockopt_seq;
549 u32 stale_rcv_tstamp;
550 int cached_sndbuf; /* sndbuf size when last synced with the msk sndbuf,
551 * protected by the msk socket lock
552 */
553
554 struct sock *tcp_sock; /* tcp sk backpointer */
555 struct sock *conn; /* parent mptcp_sock */
556 const struct inet_connection_sock_af_ops *icsk_af_ops;
557 void (*tcp_state_change)(struct sock *sk);
558 void (*tcp_error_report)(struct sock *sk);
559
560 struct rcu_head rcu;
561 };
562
563 static inline struct mptcp_subflow_context *
mptcp_subflow_ctx(const struct sock * sk)564 mptcp_subflow_ctx(const struct sock *sk)
565 {
566 const struct inet_connection_sock *icsk = inet_csk(sk);
567
568 /* Use RCU on icsk_ulp_data only for sock diag code */
569 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data;
570 }
571
572 static inline struct sock *
mptcp_subflow_tcp_sock(const struct mptcp_subflow_context * subflow)573 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow)
574 {
575 return subflow->tcp_sock;
576 }
577
578 static inline void
mptcp_subflow_ctx_reset(struct mptcp_subflow_context * subflow)579 mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow)
580 {
581 memset(&subflow->reset, 0, sizeof(subflow->reset));
582 subflow->request_mptcp = 1;
583 WRITE_ONCE(subflow->local_id, -1);
584 }
585
586 /* Convert reset reasons in MPTCP to enum sk_rst_reason type */
587 static inline enum sk_rst_reason
sk_rst_convert_mptcp_reason(u32 reason)588 sk_rst_convert_mptcp_reason(u32 reason)
589 {
590 switch (reason) {
591 case MPTCP_RST_EUNSPEC:
592 return SK_RST_REASON_MPTCP_RST_EUNSPEC;
593 case MPTCP_RST_EMPTCP:
594 return SK_RST_REASON_MPTCP_RST_EMPTCP;
595 case MPTCP_RST_ERESOURCE:
596 return SK_RST_REASON_MPTCP_RST_ERESOURCE;
597 case MPTCP_RST_EPROHIBIT:
598 return SK_RST_REASON_MPTCP_RST_EPROHIBIT;
599 case MPTCP_RST_EWQ2BIG:
600 return SK_RST_REASON_MPTCP_RST_EWQ2BIG;
601 case MPTCP_RST_EBADPERF:
602 return SK_RST_REASON_MPTCP_RST_EBADPERF;
603 case MPTCP_RST_EMIDDLEBOX:
604 return SK_RST_REASON_MPTCP_RST_EMIDDLEBOX;
605 default:
606 /* It should not happen, or else errors may occur
607 * in MPTCP layer
608 */
609 return SK_RST_REASON_ERROR;
610 }
611 }
612
613 static inline void
mptcp_send_active_reset_reason(struct sock * sk)614 mptcp_send_active_reset_reason(struct sock *sk)
615 {
616 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
617 enum sk_rst_reason reason;
618
619 reason = sk_rst_convert_mptcp_reason(subflow->reset_reason);
620 tcp_send_active_reset(sk, GFP_ATOMIC, reason);
621 }
622
623 static inline u64
mptcp_subflow_get_map_offset(const struct mptcp_subflow_context * subflow)624 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow)
625 {
626 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq -
627 subflow->ssn_offset -
628 subflow->map_subflow_seq;
629 }
630
631 static inline u64
mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context * subflow)632 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow)
633 {
634 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow);
635 }
636
637 void mptcp_subflow_process_delegated(struct sock *ssk, long actions);
638
mptcp_subflow_delegate(struct mptcp_subflow_context * subflow,int action)639 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action)
640 {
641 long old, set_bits = BIT(MPTCP_DELEGATE_SCHEDULED) | BIT(action);
642 struct mptcp_delegated_action *delegated;
643 bool schedule;
644
645 /* the caller held the subflow bh socket lock */
646 lockdep_assert_in_softirq();
647
648 /* The implied barrier pairs with tcp_release_cb_override()
649 * mptcp_napi_poll(), and ensures the below list check sees list
650 * updates done prior to delegated status bits changes
651 */
652 old = set_mask_bits(&subflow->delegated_status, 0, set_bits);
653 if (!(old & BIT(MPTCP_DELEGATE_SCHEDULED))) {
654 if (WARN_ON_ONCE(!list_empty(&subflow->delegated_node)))
655 return;
656
657 delegated = this_cpu_ptr(&mptcp_delegated_actions);
658 schedule = list_empty(&delegated->head);
659 list_add_tail(&subflow->delegated_node, &delegated->head);
660 sock_hold(mptcp_subflow_tcp_sock(subflow));
661 if (schedule)
662 napi_schedule(&delegated->napi);
663 }
664 }
665
666 static inline struct mptcp_subflow_context *
mptcp_subflow_delegated_next(struct mptcp_delegated_action * delegated)667 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated)
668 {
669 struct mptcp_subflow_context *ret;
670
671 if (list_empty(&delegated->head))
672 return NULL;
673
674 ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node);
675 list_del_init(&ret->delegated_node);
676 return ret;
677 }
678
679 int mptcp_is_enabled(const struct net *net);
680 unsigned int mptcp_get_add_addr_timeout(const struct net *net);
681 int mptcp_is_checksum_enabled(const struct net *net);
682 int mptcp_allow_join_id0(const struct net *net);
683 unsigned int mptcp_stale_loss_cnt(const struct net *net);
684 unsigned int mptcp_close_timeout(const struct sock *sk);
685 int mptcp_get_pm_type(const struct net *net);
686 const char *mptcp_get_scheduler(const struct net *net);
687 void mptcp_get_available_schedulers(char *buf, size_t maxlen);
688 void __mptcp_subflow_fully_established(struct mptcp_sock *msk,
689 struct mptcp_subflow_context *subflow,
690 const struct mptcp_options_received *mp_opt);
691 bool __mptcp_retransmit_pending_data(struct sock *sk);
692 void mptcp_check_and_set_pending(struct sock *sk);
693 void __mptcp_push_pending(struct sock *sk, unsigned int flags);
694 bool mptcp_subflow_data_available(struct sock *sk);
695 void __init mptcp_subflow_init(void);
696 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how);
697 void mptcp_close_ssk(struct sock *sk, struct sock *ssk,
698 struct mptcp_subflow_context *subflow);
699 void __mptcp_subflow_send_ack(struct sock *ssk);
700 void mptcp_subflow_reset(struct sock *ssk);
701 void mptcp_subflow_queue_clean(struct sock *sk, struct sock *ssk);
702 void mptcp_sock_graft(struct sock *sk, struct socket *parent);
703 struct sock *__mptcp_nmpc_sk(struct mptcp_sock *msk);
704 bool __mptcp_close(struct sock *sk, long timeout);
705 void mptcp_cancel_work(struct sock *sk);
706 void __mptcp_unaccepted_force_close(struct sock *sk);
707 void mptcp_set_owner_r(struct sk_buff *skb, struct sock *sk);
708 void mptcp_set_state(struct sock *sk, int state);
709
710 bool mptcp_addresses_equal(const struct mptcp_addr_info *a,
711 const struct mptcp_addr_info *b, bool use_port);
712 void mptcp_local_address(const struct sock_common *skc, struct mptcp_addr_info *addr);
713
714 /* called with sk socket lock held */
715 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc,
716 const struct mptcp_addr_info *remote);
717 int mptcp_subflow_create_socket(struct sock *sk, unsigned short family,
718 struct socket **new_sock);
719 void mptcp_info2sockaddr(const struct mptcp_addr_info *info,
720 struct sockaddr_storage *addr,
721 unsigned short family);
722 struct mptcp_sched_ops *mptcp_sched_find(const char *name);
723 int mptcp_register_scheduler(struct mptcp_sched_ops *sched);
724 void mptcp_unregister_scheduler(struct mptcp_sched_ops *sched);
725 void mptcp_sched_init(void);
726 int mptcp_init_sched(struct mptcp_sock *msk,
727 struct mptcp_sched_ops *sched);
728 void mptcp_release_sched(struct mptcp_sock *msk);
729 void mptcp_subflow_set_scheduled(struct mptcp_subflow_context *subflow,
730 bool scheduled);
731 struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk);
732 struct sock *mptcp_subflow_get_retrans(struct mptcp_sock *msk);
733 int mptcp_sched_get_send(struct mptcp_sock *msk);
734 int mptcp_sched_get_retrans(struct mptcp_sock *msk);
735
mptcp_data_avail(const struct mptcp_sock * msk)736 static inline u64 mptcp_data_avail(const struct mptcp_sock *msk)
737 {
738 return READ_ONCE(msk->bytes_received) - READ_ONCE(msk->bytes_consumed);
739 }
740
mptcp_epollin_ready(const struct sock * sk)741 static inline bool mptcp_epollin_ready(const struct sock *sk)
742 {
743 /* mptcp doesn't have to deal with small skbs in the receive queue,
744 * at it can always coalesce them
745 */
746 return (mptcp_data_avail(mptcp_sk(sk)) >= sk->sk_rcvlowat) ||
747 (mem_cgroup_sockets_enabled && sk->sk_memcg &&
748 mem_cgroup_under_socket_pressure(sk->sk_memcg)) ||
749 READ_ONCE(tcp_memory_pressure);
750 }
751
752 int mptcp_set_rcvlowat(struct sock *sk, int val);
753
__tcp_can_send(const struct sock * ssk)754 static inline bool __tcp_can_send(const struct sock *ssk)
755 {
756 /* only send if our side has not closed yet */
757 return ((1 << inet_sk_state_load(ssk)) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT));
758 }
759
__mptcp_subflow_active(struct mptcp_subflow_context * subflow)760 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow)
761 {
762 /* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */
763 if (subflow->request_join && !subflow->fully_established)
764 return false;
765
766 return __tcp_can_send(mptcp_subflow_tcp_sock(subflow));
767 }
768
769 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow);
770
771 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow);
772
773 void mptcp_subflow_drop_ctx(struct sock *ssk);
774
mptcp_subflow_tcp_fallback(struct sock * sk,struct mptcp_subflow_context * ctx)775 static inline void mptcp_subflow_tcp_fallback(struct sock *sk,
776 struct mptcp_subflow_context *ctx)
777 {
778 sk->sk_data_ready = sock_def_readable;
779 sk->sk_state_change = ctx->tcp_state_change;
780 sk->sk_write_space = sk_stream_write_space;
781 sk->sk_error_report = ctx->tcp_error_report;
782
783 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops;
784 }
785
786 void __init mptcp_proto_init(void);
787 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
788 int __init mptcp_proto_v6_init(void);
789 #endif
790
791 struct sock *mptcp_sk_clone_init(const struct sock *sk,
792 const struct mptcp_options_received *mp_opt,
793 struct sock *ssk,
794 struct request_sock *req);
795 void mptcp_get_options(const struct sk_buff *skb,
796 struct mptcp_options_received *mp_opt);
797
798 void mptcp_finish_connect(struct sock *sk);
799 void __mptcp_sync_state(struct sock *sk, int state);
800 void mptcp_reset_tout_timer(struct mptcp_sock *msk, unsigned long fail_tout);
801
mptcp_stop_tout_timer(struct sock * sk)802 static inline void mptcp_stop_tout_timer(struct sock *sk)
803 {
804 if (!inet_csk(sk)->icsk_mtup.probe_timestamp)
805 return;
806
807 sk_stop_timer(sk, &sk->sk_timer);
808 inet_csk(sk)->icsk_mtup.probe_timestamp = 0;
809 }
810
mptcp_set_close_tout(struct sock * sk,unsigned long tout)811 static inline void mptcp_set_close_tout(struct sock *sk, unsigned long tout)
812 {
813 /* avoid 0 timestamp, as that means no close timeout */
814 inet_csk(sk)->icsk_mtup.probe_timestamp = tout ? : 1;
815 }
816
mptcp_start_tout_timer(struct sock * sk)817 static inline void mptcp_start_tout_timer(struct sock *sk)
818 {
819 mptcp_set_close_tout(sk, tcp_jiffies32);
820 mptcp_reset_tout_timer(mptcp_sk(sk), 0);
821 }
822
mptcp_is_fully_established(struct sock * sk)823 static inline bool mptcp_is_fully_established(struct sock *sk)
824 {
825 return inet_sk_state_load(sk) == TCP_ESTABLISHED &&
826 READ_ONCE(mptcp_sk(sk)->fully_established);
827 }
828
829 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk);
830 void mptcp_data_ready(struct sock *sk, struct sock *ssk);
831 bool mptcp_finish_join(struct sock *sk);
832 bool mptcp_schedule_work(struct sock *sk);
833 int mptcp_setsockopt(struct sock *sk, int level, int optname,
834 sockptr_t optval, unsigned int optlen);
835 int mptcp_getsockopt(struct sock *sk, int level, int optname,
836 char __user *optval, int __user *option);
837
838 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq);
mptcp_expand_seq(u64 old_seq,u64 cur_seq,bool use_64bit)839 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit)
840 {
841 if (use_64bit)
842 return cur_seq;
843
844 return __mptcp_expand_seq(old_seq, cur_seq);
845 }
846 void __mptcp_check_push(struct sock *sk, struct sock *ssk);
847 void __mptcp_data_acked(struct sock *sk);
848 void __mptcp_error_report(struct sock *sk);
849 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)850 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk)
851 {
852 return READ_ONCE(msk->snd_data_fin_enable) &&
853 READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt);
854 }
855
mptcp_notsent_lowat(const struct sock * sk)856 static inline u32 mptcp_notsent_lowat(const struct sock *sk)
857 {
858 struct net *net = sock_net(sk);
859 u32 val;
860
861 val = READ_ONCE(mptcp_sk(sk)->notsent_lowat);
862 return val ?: READ_ONCE(net->ipv4.sysctl_tcp_notsent_lowat);
863 }
864
mptcp_stream_memory_free(const struct sock * sk,int wake)865 static inline bool mptcp_stream_memory_free(const struct sock *sk, int wake)
866 {
867 const struct mptcp_sock *msk = mptcp_sk(sk);
868 u32 notsent_bytes;
869
870 notsent_bytes = READ_ONCE(msk->write_seq) - READ_ONCE(msk->snd_nxt);
871 return (notsent_bytes << wake) < mptcp_notsent_lowat(sk);
872 }
873
__mptcp_stream_is_writeable(const struct sock * sk,int wake)874 static inline bool __mptcp_stream_is_writeable(const struct sock *sk, int wake)
875 {
876 return mptcp_stream_memory_free(sk, wake) &&
877 __sk_stream_is_writeable(sk, wake);
878 }
879
mptcp_write_space(struct sock * sk)880 static inline void mptcp_write_space(struct sock *sk)
881 {
882 /* pairs with memory barrier in mptcp_poll */
883 smp_mb();
884 if (mptcp_stream_memory_free(sk, 1))
885 sk_stream_write_space(sk);
886 }
887
__mptcp_sync_sndbuf(struct sock * sk)888 static inline void __mptcp_sync_sndbuf(struct sock *sk)
889 {
890 struct mptcp_subflow_context *subflow;
891 int ssk_sndbuf, new_sndbuf;
892
893 if (sk->sk_userlocks & SOCK_SNDBUF_LOCK)
894 return;
895
896 new_sndbuf = READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_wmem[0]);
897 mptcp_for_each_subflow(mptcp_sk(sk), subflow) {
898 ssk_sndbuf = READ_ONCE(mptcp_subflow_tcp_sock(subflow)->sk_sndbuf);
899
900 subflow->cached_sndbuf = ssk_sndbuf;
901 new_sndbuf += ssk_sndbuf;
902 }
903
904 /* the msk max wmem limit is <nr_subflows> * tcp wmem[2] */
905 WRITE_ONCE(sk->sk_sndbuf, new_sndbuf);
906 mptcp_write_space(sk);
907 }
908
909 /* The called held both the msk socket and the subflow socket locks,
910 * possibly under BH
911 */
__mptcp_propagate_sndbuf(struct sock * sk,struct sock * ssk)912 static inline void __mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
913 {
914 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
915
916 if (READ_ONCE(ssk->sk_sndbuf) != subflow->cached_sndbuf)
917 __mptcp_sync_sndbuf(sk);
918 }
919
920 /* the caller held only the subflow socket lock, either in process or
921 * BH context. Additionally this can be called under the msk data lock,
922 * so we can't acquire such lock here: let the delegate action acquires
923 * the needed locks in suitable order.
924 */
mptcp_propagate_sndbuf(struct sock * sk,struct sock * ssk)925 static inline void mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
926 {
927 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
928
929 if (likely(READ_ONCE(ssk->sk_sndbuf) == subflow->cached_sndbuf))
930 return;
931
932 local_bh_disable();
933 mptcp_subflow_delegate(subflow, MPTCP_DELEGATE_SNDBUF);
934 local_bh_enable();
935 }
936
937 void mptcp_destroy_common(struct mptcp_sock *msk, unsigned int flags);
938
939 #define MPTCP_TOKEN_MAX_RETRIES 4
940
941 void __init mptcp_token_init(void);
mptcp_token_init_request(struct request_sock * req)942 static inline void mptcp_token_init_request(struct request_sock *req)
943 {
944 mptcp_subflow_rsk(req)->token_node.pprev = NULL;
945 }
946
947 int mptcp_token_new_request(struct request_sock *req);
948 void mptcp_token_destroy_request(struct request_sock *req);
949 int mptcp_token_new_connect(struct sock *ssk);
950 void mptcp_token_accept(struct mptcp_subflow_request_sock *r,
951 struct mptcp_sock *msk);
952 bool mptcp_token_exists(u32 token);
953 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token);
954 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot,
955 long *s_num);
956 void mptcp_token_destroy(struct mptcp_sock *msk);
957
958 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn);
959
960 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac);
961 __sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum);
962
963 void __init mptcp_pm_init(void);
964 void mptcp_pm_data_init(struct mptcp_sock *msk);
965 void mptcp_pm_data_reset(struct mptcp_sock *msk);
966 int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info,
967 struct mptcp_addr_info *addr);
968 int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info,
969 bool require_family,
970 struct mptcp_pm_addr_entry *entry);
971 bool mptcp_pm_addr_families_match(const struct sock *sk,
972 const struct mptcp_addr_info *loc,
973 const struct mptcp_addr_info *rem);
974 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
975 void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
976 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side);
977 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk);
978 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk);
979 void mptcp_pm_connection_closed(struct mptcp_sock *msk);
980 void mptcp_pm_subflow_established(struct mptcp_sock *msk);
981 bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk);
982 void mptcp_pm_subflow_check_next(struct mptcp_sock *msk,
983 const struct mptcp_subflow_context *subflow);
984 void mptcp_pm_add_addr_received(const struct sock *ssk,
985 const struct mptcp_addr_info *addr);
986 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk,
987 const struct mptcp_addr_info *addr);
988 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk);
989 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk);
990 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk,
991 const struct mptcp_rm_list *rm_list);
992 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup);
993 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq);
994 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk,
995 struct mptcp_addr_info *addr,
996 struct mptcp_addr_info *rem,
997 u8 bkup);
998 bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk,
999 const struct mptcp_addr_info *addr);
1000 void mptcp_pm_free_anno_list(struct mptcp_sock *msk);
1001 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk);
1002 struct mptcp_pm_add_entry *
1003 mptcp_pm_del_add_timer(struct mptcp_sock *msk,
1004 const struct mptcp_addr_info *addr, bool check_id);
1005 struct mptcp_pm_add_entry *
1006 mptcp_lookup_anno_list_by_saddr(const struct mptcp_sock *msk,
1007 const struct mptcp_addr_info *addr);
1008 int mptcp_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk,
1009 unsigned int id,
1010 u8 *flags, int *ifindex);
1011 int mptcp_pm_nl_get_flags_and_ifindex_by_id(struct mptcp_sock *msk, unsigned int id,
1012 u8 *flags, int *ifindex);
1013 int mptcp_userspace_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk,
1014 unsigned int id,
1015 u8 *flags, int *ifindex);
1016 int mptcp_pm_set_flags(struct sk_buff *skb, struct genl_info *info);
1017 int mptcp_pm_nl_set_flags(struct sk_buff *skb, struct genl_info *info);
1018 int mptcp_userspace_pm_set_flags(struct sk_buff *skb, struct genl_info *info);
1019 int mptcp_pm_announce_addr(struct mptcp_sock *msk,
1020 const struct mptcp_addr_info *addr,
1021 bool echo);
1022 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
1023 int mptcp_pm_remove_subflow(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
1024 void mptcp_pm_remove_addrs(struct mptcp_sock *msk, struct list_head *rm_list);
1025
1026 void mptcp_free_local_addr_list(struct mptcp_sock *msk);
1027
1028 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk,
1029 const struct sock *ssk, gfp_t gfp);
1030 void mptcp_event_addr_announced(const struct sock *ssk, const struct mptcp_addr_info *info);
1031 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id);
1032 void mptcp_event_pm_listener(const struct sock *ssk,
1033 enum mptcp_event_type event);
1034 bool mptcp_userspace_pm_active(const struct mptcp_sock *msk);
1035
1036 void __mptcp_fastopen_gen_msk_ackseq(struct mptcp_sock *msk, struct mptcp_subflow_context *subflow,
1037 const struct mptcp_options_received *mp_opt);
1038 void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subflow,
1039 struct request_sock *req);
1040 int mptcp_nl_fill_addr(struct sk_buff *skb,
1041 struct mptcp_pm_addr_entry *entry);
1042
mptcp_pm_should_add_signal(struct mptcp_sock * msk)1043 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk)
1044 {
1045 return READ_ONCE(msk->pm.addr_signal) &
1046 (BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO));
1047 }
1048
mptcp_pm_should_add_signal_addr(struct mptcp_sock * msk)1049 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk)
1050 {
1051 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL);
1052 }
1053
mptcp_pm_should_add_signal_echo(struct mptcp_sock * msk)1054 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk)
1055 {
1056 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO);
1057 }
1058
mptcp_pm_should_rm_signal(struct mptcp_sock * msk)1059 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk)
1060 {
1061 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL);
1062 }
1063
mptcp_pm_is_userspace(const struct mptcp_sock * msk)1064 static inline bool mptcp_pm_is_userspace(const struct mptcp_sock *msk)
1065 {
1066 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_USERSPACE;
1067 }
1068
mptcp_pm_is_kernel(const struct mptcp_sock * msk)1069 static inline bool mptcp_pm_is_kernel(const struct mptcp_sock *msk)
1070 {
1071 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_KERNEL;
1072 }
1073
mptcp_add_addr_len(int family,bool echo,bool port)1074 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port)
1075 {
1076 u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE;
1077
1078 if (family == AF_INET6)
1079 len = TCPOLEN_MPTCP_ADD_ADDR6_BASE;
1080 if (!echo)
1081 len += MPTCPOPT_THMAC_LEN;
1082 /* account for 2 trailing 'nop' options */
1083 if (port)
1084 len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN;
1085
1086 return len;
1087 }
1088
mptcp_rm_addr_len(const struct mptcp_rm_list * rm_list)1089 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list)
1090 {
1091 if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX)
1092 return -EINVAL;
1093
1094 return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1;
1095 }
1096
1097 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, const struct sk_buff *skb,
1098 unsigned int opt_size, unsigned int remaining,
1099 struct mptcp_addr_info *addr, bool *echo,
1100 bool *drop_other_suboptions);
1101 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining,
1102 struct mptcp_rm_list *rm_list);
1103 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
1104 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1105 int mptcp_userspace_pm_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1106 int mptcp_pm_dump_addr(struct sk_buff *msg, struct netlink_callback *cb);
1107 int mptcp_pm_nl_dump_addr(struct sk_buff *msg,
1108 struct netlink_callback *cb);
1109 int mptcp_userspace_pm_dump_addr(struct sk_buff *msg,
1110 struct netlink_callback *cb);
1111 int mptcp_pm_get_addr(struct sk_buff *skb, struct genl_info *info);
1112 int mptcp_pm_nl_get_addr(struct sk_buff *skb, struct genl_info *info);
1113 int mptcp_userspace_pm_get_addr(struct sk_buff *skb,
1114 struct genl_info *info);
1115
subflow_get_local_id(const struct mptcp_subflow_context * subflow)1116 static inline u8 subflow_get_local_id(const struct mptcp_subflow_context *subflow)
1117 {
1118 int local_id = READ_ONCE(subflow->local_id);
1119
1120 if (local_id < 0)
1121 return 0;
1122 return local_id;
1123 }
1124
1125 void __init mptcp_pm_nl_init(void);
1126 void mptcp_pm_nl_work(struct mptcp_sock *msk);
1127 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk,
1128 const struct mptcp_rm_list *rm_list);
1129 unsigned int mptcp_pm_get_add_addr_signal_max(const struct mptcp_sock *msk);
1130 unsigned int mptcp_pm_get_add_addr_accept_max(const struct mptcp_sock *msk);
1131 unsigned int mptcp_pm_get_subflows_max(const struct mptcp_sock *msk);
1132 unsigned int mptcp_pm_get_local_addr_max(const struct mptcp_sock *msk);
1133
1134 /* called under PM lock */
__mptcp_pm_close_subflow(struct mptcp_sock * msk)1135 static inline void __mptcp_pm_close_subflow(struct mptcp_sock *msk)
1136 {
1137 if (--msk->pm.subflows < mptcp_pm_get_subflows_max(msk))
1138 WRITE_ONCE(msk->pm.accept_subflow, true);
1139 }
1140
mptcp_pm_close_subflow(struct mptcp_sock * msk)1141 static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk)
1142 {
1143 spin_lock_bh(&msk->pm.lock);
1144 __mptcp_pm_close_subflow(msk);
1145 spin_unlock_bh(&msk->pm.lock);
1146 }
1147
1148 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk);
1149 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk);
1150
mptcp_get_ext(const struct sk_buff * skb)1151 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb)
1152 {
1153 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP);
1154 }
1155
1156 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops);
1157
__mptcp_check_fallback(const struct mptcp_sock * msk)1158 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk)
1159 {
1160 return test_bit(MPTCP_FALLBACK_DONE, &msk->flags);
1161 }
1162
mptcp_check_fallback(const struct sock * sk)1163 static inline bool mptcp_check_fallback(const struct sock *sk)
1164 {
1165 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1166 struct mptcp_sock *msk = mptcp_sk(subflow->conn);
1167
1168 return __mptcp_check_fallback(msk);
1169 }
1170
__mptcp_do_fallback(struct mptcp_sock * msk)1171 static inline void __mptcp_do_fallback(struct mptcp_sock *msk)
1172 {
1173 if (__mptcp_check_fallback(msk)) {
1174 pr_debug("TCP fallback already done (msk=%p)", msk);
1175 return;
1176 }
1177 set_bit(MPTCP_FALLBACK_DONE, &msk->flags);
1178 }
1179
__mptcp_has_initial_subflow(const struct mptcp_sock * msk)1180 static inline bool __mptcp_has_initial_subflow(const struct mptcp_sock *msk)
1181 {
1182 struct sock *ssk = READ_ONCE(msk->first);
1183
1184 return ssk && ((1 << inet_sk_state_load(ssk)) &
1185 (TCPF_ESTABLISHED | TCPF_SYN_SENT |
1186 TCPF_SYN_RECV | TCPF_LISTEN));
1187 }
1188
mptcp_do_fallback(struct sock * ssk)1189 static inline void mptcp_do_fallback(struct sock *ssk)
1190 {
1191 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
1192 struct sock *sk = subflow->conn;
1193 struct mptcp_sock *msk;
1194
1195 msk = mptcp_sk(sk);
1196 __mptcp_do_fallback(msk);
1197 if (READ_ONCE(msk->snd_data_fin_enable) && !(ssk->sk_shutdown & SEND_SHUTDOWN)) {
1198 gfp_t saved_allocation = ssk->sk_allocation;
1199
1200 /* we are in a atomic (BH) scope, override ssk default for data
1201 * fin allocation
1202 */
1203 ssk->sk_allocation = GFP_ATOMIC;
1204 ssk->sk_shutdown |= SEND_SHUTDOWN;
1205 tcp_shutdown(ssk, SEND_SHUTDOWN);
1206 ssk->sk_allocation = saved_allocation;
1207 }
1208 }
1209
1210 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a)
1211
mptcp_check_infinite_map(struct sk_buff * skb)1212 static inline bool mptcp_check_infinite_map(struct sk_buff *skb)
1213 {
1214 struct mptcp_ext *mpext;
1215
1216 mpext = skb ? mptcp_get_ext(skb) : NULL;
1217 if (mpext && mpext->infinite_map)
1218 return true;
1219
1220 return false;
1221 }
1222
is_active_ssk(struct mptcp_subflow_context * subflow)1223 static inline bool is_active_ssk(struct mptcp_subflow_context *subflow)
1224 {
1225 return (subflow->request_mptcp || subflow->request_join);
1226 }
1227
subflow_simultaneous_connect(struct sock * sk)1228 static inline bool subflow_simultaneous_connect(struct sock *sk)
1229 {
1230 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1231
1232 return (1 << sk->sk_state) &
1233 (TCPF_ESTABLISHED | TCPF_FIN_WAIT1 | TCPF_FIN_WAIT2 | TCPF_CLOSING) &&
1234 is_active_ssk(subflow) &&
1235 !subflow->conn_finished;
1236 }
1237
1238 #ifdef CONFIG_SYN_COOKIES
1239 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1240 struct sk_buff *skb);
1241 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1242 struct sk_buff *skb);
1243 void __init mptcp_join_cookie_init(void);
1244 #else
1245 static inline void
subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock * subflow_req,struct sk_buff * skb)1246 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1247 struct sk_buff *skb) {}
1248 static inline bool
mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock * subflow_req,struct sk_buff * skb)1249 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1250 struct sk_buff *skb)
1251 {
1252 return false;
1253 }
1254
mptcp_join_cookie_init(void)1255 static inline void mptcp_join_cookie_init(void) {}
1256 #endif
1257
1258 #endif /* __MPTCP_PROTOCOL_H */
1259