Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
tjh.dev
kernel
os
linux
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
128struct mptcp_skb_cb {
129 u64 map_seq;
130 u64 end_seq;
131 u32 offset;
132 u8 has_rxtstamp;
133 u8 cant_coalesce;
134};
135
136#define MPTCP_SKB_CB(__skb) ((struct mptcp_skb_cb *)&((__skb)->cb[0]))
137
138static inline bool before64(__u64 seq1, __u64 seq2)
139{
140 return (__s64)(seq1 - seq2) < 0;
141}
142
143#define after64(seq2, seq1) before64(seq1, seq2)
144
145struct mptcp_options_received {
146 u64 sndr_key;
147 u64 rcvr_key;
148 u64 data_ack;
149 u64 data_seq;
150 u32 subflow_seq;
151 u16 data_len;
152 __sum16 csum;
153 struct_group(status,
154 u16 suboptions;
155 u16 use_map:1,
156 dsn64:1,
157 data_fin:1,
158 use_ack:1,
159 ack64:1,
160 mpc_map:1,
161 reset_reason:4,
162 reset_transient:1,
163 echo:1,
164 backup:1,
165 deny_join_id0:1,
166 __unused:2;
167 );
168 u8 join_id;
169 u32 token;
170 u32 nonce;
171 u64 thmac;
172 u8 hmac[MPTCPOPT_HMAC_LEN];
173 struct mptcp_addr_info addr;
174 struct mptcp_rm_list rm_list;
175 u64 ahmac;
176 u64 fail_seq;
177};
178
179static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field)
180{
181 return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) |
182 ((nib & 0xF) << 8) | field);
183}
184
185enum mptcp_pm_status {
186 MPTCP_PM_ADD_ADDR_RECEIVED,
187 MPTCP_PM_ADD_ADDR_SEND_ACK,
188 MPTCP_PM_RM_ADDR_RECEIVED,
189 MPTCP_PM_ESTABLISHED,
190 MPTCP_PM_SUBFLOW_ESTABLISHED,
191 MPTCP_PM_ALREADY_ESTABLISHED, /* persistent status, set after ESTABLISHED event */
192 MPTCP_PM_MPC_ENDPOINT_ACCOUNTED /* persistent status, set after MPC local address is
193 * accounted int id_avail_bitmap
194 */
195};
196
197enum mptcp_pm_type {
198 MPTCP_PM_TYPE_KERNEL = 0,
199 MPTCP_PM_TYPE_USERSPACE,
200
201 __MPTCP_PM_TYPE_NR,
202 __MPTCP_PM_TYPE_MAX = __MPTCP_PM_TYPE_NR - 1,
203};
204
205/* Status bits below MPTCP_PM_ALREADY_ESTABLISHED need pm worker actions */
206#define MPTCP_PM_WORK_MASK ((1 << MPTCP_PM_ALREADY_ESTABLISHED) - 1)
207
208enum mptcp_addr_signal_status {
209 MPTCP_ADD_ADDR_SIGNAL,
210 MPTCP_ADD_ADDR_ECHO,
211 MPTCP_RM_ADDR_SIGNAL,
212};
213
214/* max value of mptcp_addr_info.id */
215#define MPTCP_PM_MAX_ADDR_ID U8_MAX
216
217struct mptcp_pm_data {
218 struct mptcp_addr_info local;
219 struct mptcp_addr_info remote;
220 struct list_head anno_list;
221 struct list_head userspace_pm_local_addr_list;
222
223 spinlock_t lock; /*protects the whole PM data */
224
225 struct_group(reset,
226
227 u8 addr_signal;
228 bool server_side;
229 bool work_pending;
230 bool accept_addr;
231 bool accept_subflow;
232 bool remote_deny_join_id0;
233 u8 add_addr_signaled;
234 u8 add_addr_accepted;
235 u8 local_addr_used;
236 u8 pm_type;
237 u8 extra_subflows;
238 u8 status;
239
240 );
241
242 DECLARE_BITMAP(id_avail_bitmap, MPTCP_PM_MAX_ADDR_ID + 1);
243 struct mptcp_rm_list rm_list_tx;
244 struct mptcp_rm_list rm_list_rx;
245};
246
247struct mptcp_pm_local {
248 struct mptcp_addr_info addr;
249 u8 flags;
250 int ifindex;
251};
252
253struct mptcp_pm_addr_entry {
254 struct list_head list;
255 struct mptcp_addr_info addr;
256 u8 flags;
257 int ifindex;
258 struct socket *lsk;
259};
260
261struct mptcp_data_frag {
262 struct list_head list;
263 u64 data_seq;
264 u16 data_len;
265 u16 offset;
266 u16 overhead;
267 u16 already_sent;
268 struct page *page;
269};
270
271/* MPTCP connection sock */
272struct mptcp_sock {
273 /* inet_connection_sock must be the first member */
274 struct inet_connection_sock sk;
275 u64 local_key; /* protected by the first subflow socket lock
276 * lockless access read
277 */
278 u64 remote_key; /* same as above */
279 u64 write_seq;
280 u64 bytes_sent;
281 u64 snd_nxt;
282 u64 bytes_received;
283 u64 ack_seq;
284 atomic64_t rcv_wnd_sent;
285 u64 rcv_data_fin_seq;
286 u64 bytes_retrans;
287 u64 bytes_consumed;
288 int snd_burst;
289 int old_wspace;
290 u64 recovery_snd_nxt; /* in recovery mode accept up to this seq;
291 * recovery related fields are under data_lock
292 * protection
293 */
294 u64 bytes_acked;
295 u64 snd_una;
296 u64 wnd_end;
297 u32 last_data_sent;
298 u32 last_data_recv;
299 u32 last_ack_recv;
300 unsigned long timer_ival;
301 u32 token;
302 unsigned long flags;
303 unsigned long cb_flags;
304 bool recovery; /* closing subflow write queue reinjected */
305 bool can_ack;
306 bool fully_established;
307 bool rcv_data_fin;
308 bool snd_data_fin_enable;
309 bool rcv_fastclose;
310 bool use_64bit_ack; /* Set when we received a 64-bit DSN */
311 bool csum_enabled;
312 bool allow_infinite_fallback;
313 u8 pending_state; /* A subflow asked to set this sk_state,
314 * protected by the msk data lock
315 */
316 u8 mpc_endpoint_id;
317 u8 recvmsg_inq:1,
318 cork:1,
319 nodelay:1,
320 fastopening:1,
321 in_accept_queue:1,
322 free_first:1,
323 rcvspace_init:1,
324 fastclosing:1;
325 u32 notsent_lowat;
326 int keepalive_cnt;
327 int keepalive_idle;
328 int keepalive_intvl;
329 int maxseg;
330 struct work_struct work;
331 struct sk_buff *ooo_last_skb;
332 struct rb_root out_of_order_queue;
333 struct list_head conn_list;
334 struct list_head rtx_queue;
335 struct mptcp_data_frag *first_pending;
336 struct list_head join_list;
337 struct sock *first; /* The mptcp ops can safely dereference, using suitable
338 * ONCE annotation, the subflow outside the socket
339 * lock as such sock is freed after close().
340 */
341 struct mptcp_pm_data pm;
342 struct mptcp_sched_ops *sched;
343 struct {
344 int space; /* bytes copied in last measurement window */
345 int copied; /* bytes copied in this measurement window */
346 u64 time; /* start time of measurement window */
347 u64 rtt_us; /* last maximum rtt of subflows */
348 } rcvq_space;
349 u8 scaling_ratio;
350 bool allow_subflows;
351
352 u32 subflow_id;
353 u32 setsockopt_seq;
354 char ca_name[TCP_CA_NAME_MAX];
355
356 spinlock_t fallback_lock; /* protects fallback,
357 * allow_infinite_fallback and
358 * allow_join
359 */
360
361 struct list_head backlog_list; /* protected by the data lock */
362 u32 backlog_len;
363 u32 backlog_unaccounted;
364};
365
366#define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock)
367#define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock)
368
369#define mptcp_for_each_subflow(__msk, __subflow) \
370 list_for_each_entry(__subflow, &((__msk)->conn_list), node)
371#define mptcp_for_each_subflow_safe(__msk, __subflow, __tmp) \
372 list_for_each_entry_safe(__subflow, __tmp, &((__msk)->conn_list), node)
373#define mptcp_next_subflow(__msk, __subflow) \
374 list_next_entry_circular(__subflow, &((__msk)->conn_list), node)
375
376extern struct genl_family mptcp_genl_family;
377
378static inline void msk_owned_by_me(const struct mptcp_sock *msk)
379{
380 sock_owned_by_me((const struct sock *)msk);
381}
382
383#ifdef CONFIG_DEBUG_NET
384/* MPTCP-specific: we might (indirectly) call this helper with the wrong sk */
385#undef tcp_sk
386#define tcp_sk(ptr) ({ \
387 typeof(ptr) _ptr = (ptr); \
388 WARN_ON(_ptr->sk_protocol != IPPROTO_TCP); \
389 container_of_const(_ptr, struct tcp_sock, inet_conn.icsk_inet.sk); \
390})
391#define mptcp_sk(ptr) ({ \
392 typeof(ptr) _ptr = (ptr); \
393 WARN_ON(_ptr->sk_protocol != IPPROTO_MPTCP); \
394 container_of_const(_ptr, struct mptcp_sock, sk.icsk_inet.sk); \
395})
396
397#else /* !CONFIG_DEBUG_NET */
398#define mptcp_sk(ptr) container_of_const(ptr, struct mptcp_sock, sk.icsk_inet.sk)
399#endif
400
401static inline int mptcp_win_from_space(const struct sock *sk, int space)
402{
403 return __tcp_win_from_space(mptcp_sk(sk)->scaling_ratio, space);
404}
405
406static inline int mptcp_space_from_win(const struct sock *sk, int win)
407{
408 return __tcp_space_from_win(mptcp_sk(sk)->scaling_ratio, win);
409}
410
411static inline int __mptcp_space(const struct sock *sk)
412{
413 return mptcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) -
414 READ_ONCE(mptcp_sk(sk)->backlog_len) -
415 sk_rmem_alloc_get(sk));
416}
417
418static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk)
419{
420 const struct mptcp_sock *msk = mptcp_sk(sk);
421
422 return msk->first_pending;
423}
424
425static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk)
426{
427 struct mptcp_sock *msk = mptcp_sk(sk);
428 struct mptcp_data_frag *cur;
429
430 cur = msk->first_pending;
431 return list_is_last(&cur->list, &msk->rtx_queue) ? NULL :
432 list_next_entry(cur, list);
433}
434
435static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk)
436{
437 const struct mptcp_sock *msk = mptcp_sk(sk);
438
439 if (!msk->first_pending)
440 return NULL;
441
442 if (WARN_ON_ONCE(list_empty(&msk->rtx_queue)))
443 return NULL;
444
445 return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list);
446}
447
448static inline struct mptcp_data_frag *mptcp_rtx_head(struct sock *sk)
449{
450 struct mptcp_sock *msk = mptcp_sk(sk);
451
452 if (msk->snd_una == msk->snd_nxt)
453 return NULL;
454
455 return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list);
456}
457
458struct csum_pseudo_header {
459 __be64 data_seq;
460 __be32 subflow_seq;
461 __be16 data_len;
462 __sum16 csum;
463};
464
465struct mptcp_subflow_request_sock {
466 struct tcp_request_sock sk;
467 u16 mp_capable : 1,
468 mp_join : 1,
469 backup : 1,
470 request_bkup : 1,
471 csum_reqd : 1,
472 allow_join_id0 : 1;
473 u8 local_id;
474 u8 remote_id;
475 u64 local_key;
476 u64 idsn;
477 u32 token;
478 u32 ssn_offset;
479 u64 thmac;
480 u32 local_nonce;
481 u32 remote_nonce;
482 struct mptcp_sock *msk;
483 struct hlist_nulls_node token_node;
484};
485
486static inline struct mptcp_subflow_request_sock *
487mptcp_subflow_rsk(const struct request_sock *rsk)
488{
489 return (struct mptcp_subflow_request_sock *)rsk;
490}
491
492struct mptcp_delegated_action {
493 struct napi_struct napi;
494 local_lock_t bh_lock;
495 struct list_head head;
496};
497
498DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions);
499
500#define MPTCP_DELEGATE_SCHEDULED 0
501#define MPTCP_DELEGATE_SEND 1
502#define MPTCP_DELEGATE_ACK 2
503#define MPTCP_DELEGATE_SNDBUF 3
504
505#define MPTCP_DELEGATE_ACTIONS_MASK (~BIT(MPTCP_DELEGATE_SCHEDULED))
506/* MPTCP subflow context */
507struct mptcp_subflow_context {
508 struct list_head node;/* conn_list of subflows */
509
510 struct_group(reset,
511
512 unsigned long avg_pacing_rate; /* protected by msk socket lock */
513 u64 local_key;
514 u64 remote_key;
515 u64 idsn;
516 u64 map_seq;
517 u64 rcv_wnd_sent;
518 u32 snd_isn;
519 u32 token;
520 u32 rel_write_seq;
521 u32 map_subflow_seq;
522 u32 ssn_offset;
523 u32 map_data_len;
524 __wsum map_data_csum;
525 u32 map_csum_len;
526 u32 request_mptcp : 1, /* send MP_CAPABLE */
527 request_join : 1, /* send MP_JOIN */
528 request_bkup : 1,
529 mp_capable : 1, /* remote is MPTCP capable */
530 mp_join : 1, /* remote is JOINing */
531 pm_notified : 1, /* PM hook called for established status */
532 conn_finished : 1,
533 map_valid : 1,
534 map_csum_reqd : 1,
535 map_data_fin : 1,
536 mpc_map : 1,
537 backup : 1,
538 send_mp_prio : 1,
539 send_mp_fail : 1,
540 send_fastclose : 1,
541 send_infinite_map : 1,
542 remote_key_valid : 1, /* received the peer key from */
543 disposable : 1, /* ctx can be free at ulp release time */
544 closing : 1, /* must not pass rx data to msk anymore */
545 stale : 1, /* unable to snd/rcv data, do not use for xmit */
546 valid_csum_seen : 1, /* at least one csum validated */
547 is_mptfo : 1, /* subflow is doing TFO */
548 close_event_done : 1, /* has done the post-closed part */
549 mpc_drop : 1, /* the MPC option has been dropped in a rtx */
550 __unused : 8;
551 bool data_avail;
552 bool scheduled;
553 bool pm_listener; /* a listener managed by the kernel PM? */
554 bool fully_established; /* path validated */
555 u32 lent_mem_frag;
556 u32 remote_nonce;
557 u64 thmac;
558 u32 local_nonce;
559 u32 remote_token;
560 union {
561 u8 hmac[MPTCPOPT_HMAC_LEN]; /* MPJ subflow only */
562 u64 iasn; /* initial ack sequence number, MPC subflows only */
563 };
564 s16 local_id; /* if negative not initialized yet */
565 u8 remote_id;
566 u8 reset_seen:1;
567 u8 reset_transient:1;
568 u8 reset_reason:4;
569 u8 stale_count;
570
571 u32 subflow_id;
572
573 long delegated_status;
574 unsigned long fail_tout;
575
576 );
577
578 struct list_head delegated_node; /* link into delegated_action, protected by local BH */
579
580 u32 setsockopt_seq;
581 u32 stale_rcv_tstamp;
582 int cached_sndbuf; /* sndbuf size when last synced with the msk sndbuf,
583 * protected by the msk socket lock
584 */
585
586 struct sock *tcp_sock; /* tcp sk backpointer */
587 struct sock *conn; /* parent mptcp_sock */
588 const struct inet_connection_sock_af_ops *icsk_af_ops;
589 void (*tcp_state_change)(struct sock *sk);
590 void (*tcp_error_report)(struct sock *sk);
591
592 struct rcu_head rcu;
593};
594
595static inline struct mptcp_subflow_context *
596mptcp_subflow_ctx(const struct sock *sk)
597{
598 const struct inet_connection_sock *icsk = inet_csk(sk);
599
600 /* Use RCU on icsk_ulp_data only for sock diag code */
601 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data;
602}
603
604static inline struct sock *
605mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow)
606{
607 return subflow->tcp_sock;
608}
609
610static inline void
611mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow)
612{
613 memset(&subflow->reset, 0, sizeof(subflow->reset));
614 subflow->request_mptcp = 1;
615 WRITE_ONCE(subflow->local_id, -1);
616}
617
618/* Convert reset reasons in MPTCP to enum sk_rst_reason type */
619static inline enum sk_rst_reason
620sk_rst_convert_mptcp_reason(u32 reason)
621{
622 switch (reason) {
623 case MPTCP_RST_EUNSPEC:
624 return SK_RST_REASON_MPTCP_RST_EUNSPEC;
625 case MPTCP_RST_EMPTCP:
626 return SK_RST_REASON_MPTCP_RST_EMPTCP;
627 case MPTCP_RST_ERESOURCE:
628 return SK_RST_REASON_MPTCP_RST_ERESOURCE;
629 case MPTCP_RST_EPROHIBIT:
630 return SK_RST_REASON_MPTCP_RST_EPROHIBIT;
631 case MPTCP_RST_EWQ2BIG:
632 return SK_RST_REASON_MPTCP_RST_EWQ2BIG;
633 case MPTCP_RST_EBADPERF:
634 return SK_RST_REASON_MPTCP_RST_EBADPERF;
635 case MPTCP_RST_EMIDDLEBOX:
636 return SK_RST_REASON_MPTCP_RST_EMIDDLEBOX;
637 default:
638 /* It should not happen, or else errors may occur
639 * in MPTCP layer
640 */
641 return SK_RST_REASON_ERROR;
642 }
643}
644
645static inline void
646mptcp_send_active_reset_reason(struct sock *sk)
647{
648 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
649 enum sk_rst_reason reason;
650
651 reason = sk_rst_convert_mptcp_reason(subflow->reset_reason);
652 tcp_send_active_reset(sk, GFP_ATOMIC, reason);
653}
654
655/* Made the fwd mem carried by the given skb available to the msk,
656 * To be paired with a previous mptcp_subflow_lend_fwdmem() before freeing
657 * the skb or setting the skb ownership.
658 */
659static inline void mptcp_borrow_fwdmem(struct sock *sk, struct sk_buff *skb)
660{
661 struct sock *ssk = skb->sk;
662
663 /* The subflow just lend the skb fwd memory; if the subflow meanwhile
664 * closed, mptcp_close_ssk() already released the ssk rcv memory.
665 */
666 DEBUG_NET_WARN_ON_ONCE(skb->destructor);
667 sk_forward_alloc_add(sk, skb->truesize);
668 if (!ssk)
669 return;
670
671 atomic_sub(skb->truesize, &ssk->sk_rmem_alloc);
672 skb->sk = NULL;
673}
674
675static inline void
676__mptcp_subflow_lend_fwdmem(struct mptcp_subflow_context *subflow, int size)
677{
678 int frag = (subflow->lent_mem_frag + size) & (PAGE_SIZE - 1);
679
680 subflow->lent_mem_frag = frag;
681}
682
683static inline void
684mptcp_subflow_lend_fwdmem(struct mptcp_subflow_context *subflow,
685 struct sk_buff *skb)
686{
687 __mptcp_subflow_lend_fwdmem(subflow, skb->truesize);
688 skb->destructor = NULL;
689}
690
691static inline u64
692mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow)
693{
694 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq -
695 subflow->ssn_offset -
696 subflow->map_subflow_seq;
697}
698
699static inline u64
700mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow)
701{
702 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow);
703}
704
705void mptcp_subflow_process_delegated(struct sock *ssk, long actions);
706
707static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action)
708{
709 long old, set_bits = BIT(MPTCP_DELEGATE_SCHEDULED) | BIT(action);
710 struct mptcp_delegated_action *delegated;
711 bool schedule;
712
713 /* the caller held the subflow bh socket lock */
714 lockdep_assert_in_softirq();
715
716 /* The implied barrier pairs with tcp_release_cb_override()
717 * mptcp_napi_poll(), and ensures the below list check sees list
718 * updates done prior to delegated status bits changes
719 */
720 old = set_mask_bits(&subflow->delegated_status, 0, set_bits);
721 if (!(old & BIT(MPTCP_DELEGATE_SCHEDULED))) {
722 if (WARN_ON_ONCE(!list_empty(&subflow->delegated_node)))
723 return;
724
725 local_lock_nested_bh(&mptcp_delegated_actions.bh_lock);
726 delegated = this_cpu_ptr(&mptcp_delegated_actions);
727 schedule = list_empty(&delegated->head);
728 list_add_tail(&subflow->delegated_node, &delegated->head);
729 local_unlock_nested_bh(&mptcp_delegated_actions.bh_lock);
730 sock_hold(mptcp_subflow_tcp_sock(subflow));
731 if (schedule)
732 napi_schedule(&delegated->napi);
733 }
734}
735
736static inline struct mptcp_subflow_context *
737mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated)
738{
739 struct mptcp_subflow_context *ret;
740
741 local_lock_nested_bh(&mptcp_delegated_actions.bh_lock);
742 if (list_empty(&delegated->head)) {
743 local_unlock_nested_bh(&mptcp_delegated_actions.bh_lock);
744 return NULL;
745 }
746
747 ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node);
748 list_del_init(&ret->delegated_node);
749 local_unlock_nested_bh(&mptcp_delegated_actions.bh_lock);
750 return ret;
751}
752
753void __mptcp_inherit_memcg(struct sock *sk, struct sock *ssk, gfp_t gfp);
754void __mptcp_inherit_cgrp_data(struct sock *sk, struct sock *ssk);
755
756int mptcp_is_enabled(const struct net *net);
757unsigned int mptcp_get_add_addr_timeout(const struct net *net);
758int mptcp_is_checksum_enabled(const struct net *net);
759int mptcp_allow_join_id0(const struct net *net);
760unsigned int mptcp_stale_loss_cnt(const struct net *net);
761unsigned int mptcp_close_timeout(const struct sock *sk);
762int mptcp_get_pm_type(const struct net *net);
763const char *mptcp_get_path_manager(const struct net *net);
764const char *mptcp_get_scheduler(const struct net *net);
765
766void mptcp_active_disable(struct sock *sk);
767bool mptcp_active_should_disable(struct sock *ssk);
768void mptcp_active_enable(struct sock *sk);
769
770void mptcp_get_available_schedulers(char *buf, size_t maxlen);
771void __mptcp_subflow_fully_established(struct mptcp_sock *msk,
772 struct mptcp_subflow_context *subflow,
773 const struct mptcp_options_received *mp_opt);
774bool __mptcp_retransmit_pending_data(struct sock *sk);
775void mptcp_check_and_set_pending(struct sock *sk);
776void __mptcp_push_pending(struct sock *sk, unsigned int flags);
777bool mptcp_subflow_data_available(struct sock *sk);
778void __init mptcp_subflow_init(void);
779void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how);
780void mptcp_close_ssk(struct sock *sk, struct sock *ssk,
781 struct mptcp_subflow_context *subflow);
782void __mptcp_subflow_send_ack(struct sock *ssk);
783void mptcp_subflow_reset(struct sock *ssk);
784void mptcp_subflow_queue_clean(struct sock *sk, struct sock *ssk);
785void mptcp_sock_graft(struct sock *sk, struct socket *parent);
786struct sock *__mptcp_nmpc_sk(struct mptcp_sock *msk);
787bool __mptcp_close(struct sock *sk, long timeout);
788void mptcp_cancel_work(struct sock *sk);
789void __mptcp_unaccepted_force_close(struct sock *sk);
790void mptcp_set_state(struct sock *sk, int state);
791
792bool mptcp_addresses_equal(const struct mptcp_addr_info *a,
793 const struct mptcp_addr_info *b, bool use_port);
794void mptcp_local_address(const struct sock_common *skc,
795 struct mptcp_addr_info *addr);
796void mptcp_remote_address(const struct sock_common *skc,
797 struct mptcp_addr_info *addr);
798
799/* called with sk socket lock held */
800int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_pm_local *local,
801 const struct mptcp_addr_info *remote);
802int mptcp_subflow_create_socket(struct sock *sk, unsigned short family,
803 struct socket **new_sock);
804void mptcp_info2sockaddr(const struct mptcp_addr_info *info,
805 struct sockaddr_storage *addr,
806 unsigned short family);
807struct mptcp_sched_ops *mptcp_sched_find(const char *name);
808int mptcp_validate_scheduler(struct mptcp_sched_ops *sched);
809int mptcp_register_scheduler(struct mptcp_sched_ops *sched);
810void mptcp_unregister_scheduler(struct mptcp_sched_ops *sched);
811void mptcp_sched_init(void);
812int mptcp_init_sched(struct mptcp_sock *msk,
813 struct mptcp_sched_ops *sched);
814void mptcp_release_sched(struct mptcp_sock *msk);
815void mptcp_subflow_set_scheduled(struct mptcp_subflow_context *subflow,
816 bool scheduled);
817struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk);
818struct sock *mptcp_subflow_get_retrans(struct mptcp_sock *msk);
819int mptcp_sched_get_send(struct mptcp_sock *msk);
820int mptcp_sched_get_retrans(struct mptcp_sock *msk);
821
822static inline u64 mptcp_data_avail(const struct mptcp_sock *msk)
823{
824 return READ_ONCE(msk->bytes_received) - READ_ONCE(msk->bytes_consumed);
825}
826
827static inline bool mptcp_epollin_ready(const struct sock *sk)
828{
829 u64 data_avail = mptcp_data_avail(mptcp_sk(sk));
830
831 if (!data_avail)
832 return false;
833
834 /* mptcp doesn't have to deal with small skbs in the receive queue,
835 * as it can always coalesce them
836 */
837 return (data_avail >= sk->sk_rcvlowat) ||
838 tcp_under_memory_pressure(sk);
839}
840
841int mptcp_set_rcvlowat(struct sock *sk, int val);
842
843static inline bool __tcp_can_send(const struct sock *ssk)
844{
845 /* only send if our side has not closed yet */
846 return ((1 << inet_sk_state_load(ssk)) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT));
847}
848
849static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow)
850{
851 /* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */
852 if (subflow->request_join && !READ_ONCE(subflow->fully_established))
853 return false;
854
855 return __tcp_can_send(mptcp_subflow_tcp_sock(subflow));
856}
857
858void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow);
859
860bool mptcp_subflow_active(struct mptcp_subflow_context *subflow);
861
862void mptcp_subflow_drop_ctx(struct sock *ssk);
863
864static inline void mptcp_subflow_tcp_fallback(struct sock *sk,
865 struct mptcp_subflow_context *ctx)
866{
867 sk->sk_data_ready = sock_def_readable;
868 sk->sk_state_change = ctx->tcp_state_change;
869 sk->sk_write_space = sk_stream_write_space;
870 sk->sk_error_report = ctx->tcp_error_report;
871
872 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops;
873}
874
875void __init mptcp_proto_init(void);
876#if IS_ENABLED(CONFIG_MPTCP_IPV6)
877int __init mptcp_proto_v6_init(void);
878#endif
879
880struct sock *mptcp_sk_clone_init(const struct sock *sk,
881 const struct mptcp_options_received *mp_opt,
882 struct sock *ssk,
883 struct request_sock *req);
884void mptcp_get_options(const struct sk_buff *skb,
885 struct mptcp_options_received *mp_opt);
886
887void mptcp_finish_connect(struct sock *sk);
888void __mptcp_sync_state(struct sock *sk, int state);
889void mptcp_reset_tout_timer(struct mptcp_sock *msk, unsigned long fail_tout);
890
891static inline void mptcp_stop_tout_timer(struct sock *sk)
892{
893 if (!inet_csk(sk)->icsk_mtup.probe_timestamp)
894 return;
895
896 sk_stop_timer(sk, &inet_csk(sk)->mptcp_tout_timer);
897 inet_csk(sk)->icsk_mtup.probe_timestamp = 0;
898}
899
900static inline void mptcp_set_close_tout(struct sock *sk, unsigned long tout)
901{
902 /* avoid 0 timestamp, as that means no close timeout */
903 inet_csk(sk)->icsk_mtup.probe_timestamp = tout ? : 1;
904}
905
906static inline void mptcp_start_tout_timer(struct sock *sk)
907{
908 mptcp_set_close_tout(sk, tcp_jiffies32);
909 mptcp_reset_tout_timer(mptcp_sk(sk), 0);
910}
911
912static inline bool mptcp_is_fully_established(struct sock *sk)
913{
914 return inet_sk_state_load(sk) == TCP_ESTABLISHED &&
915 READ_ONCE(mptcp_sk(sk)->fully_established);
916}
917
918void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk);
919void mptcp_data_ready(struct sock *sk, struct sock *ssk);
920bool mptcp_finish_join(struct sock *sk);
921bool mptcp_schedule_work(struct sock *sk);
922int mptcp_setsockopt(struct sock *sk, int level, int optname,
923 sockptr_t optval, unsigned int optlen);
924int mptcp_getsockopt(struct sock *sk, int level, int optname,
925 char __user *optval, int __user *option);
926
927u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq);
928static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit)
929{
930 if (use_64bit)
931 return cur_seq;
932
933 return __mptcp_expand_seq(old_seq, cur_seq);
934}
935void __mptcp_check_push(struct sock *sk, struct sock *ssk);
936void __mptcp_data_acked(struct sock *sk);
937void __mptcp_error_report(struct sock *sk);
938bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit);
939static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk)
940{
941 return READ_ONCE(msk->snd_data_fin_enable) &&
942 READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt);
943}
944
945static inline u32 mptcp_notsent_lowat(const struct sock *sk)
946{
947 struct net *net = sock_net(sk);
948 u32 val;
949
950 val = READ_ONCE(mptcp_sk(sk)->notsent_lowat);
951 return val ?: READ_ONCE(net->ipv4.sysctl_tcp_notsent_lowat);
952}
953
954static inline bool mptcp_stream_memory_free(const struct sock *sk, int wake)
955{
956 const struct mptcp_sock *msk = mptcp_sk(sk);
957 u32 notsent_bytes;
958
959 notsent_bytes = READ_ONCE(msk->write_seq) - READ_ONCE(msk->snd_nxt);
960 return (notsent_bytes << wake) < mptcp_notsent_lowat(sk);
961}
962
963static inline bool __mptcp_stream_is_writeable(const struct sock *sk, int wake)
964{
965 return mptcp_stream_memory_free(sk, wake) &&
966 __sk_stream_is_writeable(sk, wake);
967}
968
969static inline void mptcp_write_space(struct sock *sk)
970{
971 /* pairs with memory barrier in mptcp_poll */
972 smp_mb();
973 if (mptcp_stream_memory_free(sk, 1))
974 sk_stream_write_space(sk);
975}
976
977static inline void __mptcp_sync_sndbuf(struct sock *sk)
978{
979 struct mptcp_subflow_context *subflow;
980 int ssk_sndbuf, new_sndbuf;
981
982 if (sk->sk_userlocks & SOCK_SNDBUF_LOCK)
983 return;
984
985 new_sndbuf = READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_wmem[0]);
986 mptcp_for_each_subflow(mptcp_sk(sk), subflow) {
987 ssk_sndbuf = READ_ONCE(mptcp_subflow_tcp_sock(subflow)->sk_sndbuf);
988
989 subflow->cached_sndbuf = ssk_sndbuf;
990 new_sndbuf += ssk_sndbuf;
991 }
992
993 /* the msk max wmem limit is <nr_subflows> * tcp wmem[2] */
994 WRITE_ONCE(sk->sk_sndbuf, new_sndbuf);
995 mptcp_write_space(sk);
996}
997
998/* The called held both the msk socket and the subflow socket locks,
999 * possibly under BH
1000 */
1001static inline void __mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
1002{
1003 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
1004
1005 if (READ_ONCE(ssk->sk_sndbuf) != subflow->cached_sndbuf)
1006 __mptcp_sync_sndbuf(sk);
1007}
1008
1009/* the caller held only the subflow socket lock, either in process or
1010 * BH context. Additionally this can be called under the msk data lock,
1011 * so we can't acquire such lock here: let the delegate action acquires
1012 * the needed locks in suitable order.
1013 */
1014static inline void mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
1015{
1016 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
1017
1018 if (likely(READ_ONCE(ssk->sk_sndbuf) == subflow->cached_sndbuf))
1019 return;
1020
1021 local_bh_disable();
1022 mptcp_subflow_delegate(subflow, MPTCP_DELEGATE_SNDBUF);
1023 local_bh_enable();
1024}
1025
1026#define MPTCP_TOKEN_MAX_RETRIES 4
1027
1028void __init mptcp_token_init(void);
1029static inline void mptcp_token_init_request(struct request_sock *req)
1030{
1031 mptcp_subflow_rsk(req)->token_node.pprev = NULL;
1032}
1033
1034int mptcp_token_new_request(struct request_sock *req);
1035void mptcp_token_destroy_request(struct request_sock *req);
1036int mptcp_token_new_connect(struct sock *ssk);
1037void mptcp_token_accept(struct mptcp_subflow_request_sock *r,
1038 struct mptcp_sock *msk);
1039bool mptcp_token_exists(u32 token);
1040struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token);
1041struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot,
1042 long *s_num);
1043void mptcp_token_destroy(struct mptcp_sock *msk);
1044
1045void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn);
1046
1047void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac);
1048__sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum);
1049
1050void __init mptcp_pm_init(void);
1051void mptcp_pm_data_init(struct mptcp_sock *msk);
1052void mptcp_pm_data_reset(struct mptcp_sock *msk);
1053void mptcp_pm_destroy(struct mptcp_sock *msk);
1054int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info,
1055 struct mptcp_addr_info *addr);
1056int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info,
1057 bool require_family,
1058 struct mptcp_pm_addr_entry *entry);
1059bool mptcp_pm_addr_families_match(const struct sock *sk,
1060 const struct mptcp_addr_info *loc,
1061 const struct mptcp_addr_info *rem);
1062void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
1063void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side);
1064void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk);
1065bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk);
1066void mptcp_pm_connection_closed(struct mptcp_sock *msk);
1067void mptcp_pm_subflow_established(struct mptcp_sock *msk);
1068bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk);
1069void mptcp_pm_subflow_check_next(struct mptcp_sock *msk,
1070 const struct mptcp_subflow_context *subflow);
1071void mptcp_pm_add_addr_received(const struct sock *ssk,
1072 const struct mptcp_addr_info *addr);
1073void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk,
1074 const struct mptcp_addr_info *addr);
1075void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk);
1076void mptcp_pm_send_ack(struct mptcp_sock *msk,
1077 struct mptcp_subflow_context *subflow,
1078 bool prio, bool backup);
1079void mptcp_pm_addr_send_ack(struct mptcp_sock *msk);
1080void mptcp_pm_nl_rm_addr(struct mptcp_sock *msk, u8 rm_id);
1081void mptcp_pm_rm_subflow(struct mptcp_sock *msk,
1082 const struct mptcp_rm_list *rm_list);
1083void mptcp_pm_rm_addr_received(struct mptcp_sock *msk,
1084 const struct mptcp_rm_list *rm_list);
1085void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup);
1086void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq);
1087int mptcp_pm_mp_prio_send_ack(struct mptcp_sock *msk,
1088 struct mptcp_addr_info *addr,
1089 struct mptcp_addr_info *rem,
1090 u8 bkup);
1091bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk,
1092 const struct mptcp_addr_info *addr);
1093bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk);
1094struct mptcp_pm_add_entry *
1095mptcp_pm_del_add_timer(struct mptcp_sock *msk,
1096 const struct mptcp_addr_info *addr, bool check_id);
1097bool mptcp_lookup_subflow_by_saddr(const struct list_head *list,
1098 const struct mptcp_addr_info *saddr);
1099bool mptcp_remove_anno_list_by_saddr(struct mptcp_sock *msk,
1100 const struct mptcp_addr_info *addr);
1101int mptcp_pm_nl_set_flags(struct mptcp_pm_addr_entry *local,
1102 struct genl_info *info);
1103int mptcp_userspace_pm_set_flags(struct mptcp_pm_addr_entry *local,
1104 struct genl_info *info);
1105int mptcp_pm_announce_addr(struct mptcp_sock *msk,
1106 const struct mptcp_addr_info *addr,
1107 bool echo);
1108int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
1109void mptcp_pm_remove_addr_entry(struct mptcp_sock *msk,
1110 struct mptcp_pm_addr_entry *entry);
1111
1112/* the default path manager, used in mptcp_pm_unregister */
1113extern struct mptcp_pm_ops mptcp_pm_kernel;
1114
1115struct mptcp_pm_ops *mptcp_pm_find(const char *name);
1116int mptcp_pm_register(struct mptcp_pm_ops *pm_ops);
1117void mptcp_pm_unregister(struct mptcp_pm_ops *pm_ops);
1118int mptcp_pm_validate(struct mptcp_pm_ops *pm_ops);
1119void mptcp_pm_get_available(char *buf, size_t maxlen);
1120
1121void mptcp_userspace_pm_free_local_addr_list(struct mptcp_sock *msk);
1122
1123void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk,
1124 const struct sock *ssk, gfp_t gfp);
1125void mptcp_event_addr_announced(const struct sock *ssk, const struct mptcp_addr_info *info);
1126void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id);
1127void mptcp_event_pm_listener(const struct sock *ssk,
1128 enum mptcp_event_type event);
1129bool mptcp_userspace_pm_active(const struct mptcp_sock *msk);
1130
1131void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subflow,
1132 struct request_sock *req);
1133int mptcp_pm_genl_fill_addr(struct sk_buff *msg,
1134 struct netlink_callback *cb,
1135 struct mptcp_pm_addr_entry *entry);
1136
1137static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk)
1138{
1139 return READ_ONCE(msk->pm.addr_signal) &
1140 (BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO));
1141}
1142
1143static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk)
1144{
1145 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL);
1146}
1147
1148static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk)
1149{
1150 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO);
1151}
1152
1153static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk)
1154{
1155 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL);
1156}
1157
1158static inline bool mptcp_pm_is_userspace(const struct mptcp_sock *msk)
1159{
1160 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_USERSPACE;
1161}
1162
1163static inline bool mptcp_pm_is_kernel(const struct mptcp_sock *msk)
1164{
1165 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_KERNEL;
1166}
1167
1168static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port)
1169{
1170 u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE;
1171
1172 if (family == AF_INET6)
1173 len = TCPOLEN_MPTCP_ADD_ADDR6_BASE;
1174 if (!echo)
1175 len += MPTCPOPT_THMAC_LEN;
1176 /* account for 2 trailing 'nop' options */
1177 if (port)
1178 len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN;
1179
1180 return len;
1181}
1182
1183static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list)
1184{
1185 if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX)
1186 return -EINVAL;
1187
1188 return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1;
1189}
1190
1191bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, const struct sk_buff *skb,
1192 unsigned int opt_size, unsigned int remaining,
1193 struct mptcp_addr_info *addr, bool *echo,
1194 bool *drop_other_suboptions);
1195bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining,
1196 struct mptcp_rm_list *rm_list);
1197int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
1198int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk,
1199 struct mptcp_pm_addr_entry *skc);
1200int mptcp_userspace_pm_get_local_id(struct mptcp_sock *msk,
1201 struct mptcp_pm_addr_entry *skc);
1202bool mptcp_pm_is_backup(struct mptcp_sock *msk, struct sock_common *skc);
1203bool mptcp_pm_nl_is_backup(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1204bool mptcp_userspace_pm_is_backup(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1205int mptcp_pm_nl_dump_addr(struct sk_buff *msg,
1206 struct netlink_callback *cb);
1207int mptcp_userspace_pm_dump_addr(struct sk_buff *msg,
1208 struct netlink_callback *cb);
1209int mptcp_pm_nl_get_addr(u8 id, struct mptcp_pm_addr_entry *addr,
1210 struct genl_info *info);
1211int mptcp_userspace_pm_get_addr(u8 id, struct mptcp_pm_addr_entry *addr,
1212 struct genl_info *info);
1213
1214static inline u8 subflow_get_local_id(const struct mptcp_subflow_context *subflow)
1215{
1216 int local_id = READ_ONCE(subflow->local_id);
1217
1218 if (local_id < 0)
1219 return 0;
1220 return local_id;
1221}
1222
1223void __init mptcp_pm_kernel_register(void);
1224void __init mptcp_pm_userspace_register(void);
1225void __init mptcp_pm_nl_init(void);
1226void mptcp_pm_worker(struct mptcp_sock *msk);
1227void __mptcp_pm_kernel_worker(struct mptcp_sock *msk);
1228u8 mptcp_pm_get_endp_signal_max(const struct mptcp_sock *msk);
1229u8 mptcp_pm_get_endp_subflow_max(const struct mptcp_sock *msk);
1230u8 mptcp_pm_get_endp_laminar_max(const struct mptcp_sock *msk);
1231u8 mptcp_pm_get_endp_fullmesh_max(const struct mptcp_sock *msk);
1232u8 mptcp_pm_get_limit_add_addr_accepted(const struct mptcp_sock *msk);
1233u8 mptcp_pm_get_limit_extra_subflows(const struct mptcp_sock *msk);
1234
1235/* called under PM lock */
1236static inline void __mptcp_pm_close_subflow(struct mptcp_sock *msk)
1237{
1238 if (--msk->pm.extra_subflows < mptcp_pm_get_limit_extra_subflows(msk))
1239 WRITE_ONCE(msk->pm.accept_subflow, true);
1240}
1241
1242static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk)
1243{
1244 spin_lock_bh(&msk->pm.lock);
1245 __mptcp_pm_close_subflow(msk);
1246 spin_unlock_bh(&msk->pm.lock);
1247}
1248
1249static inline bool mptcp_pm_add_addr_c_flag_case(struct mptcp_sock *msk)
1250{
1251 return READ_ONCE(msk->pm.remote_deny_join_id0) &&
1252 msk->pm.local_addr_used == 0 &&
1253 mptcp_pm_get_limit_add_addr_accepted(msk) == 0 &&
1254 msk->pm.extra_subflows < mptcp_pm_get_limit_extra_subflows(msk);
1255}
1256
1257void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk);
1258
1259static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb)
1260{
1261 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP);
1262}
1263
1264void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops);
1265
1266static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk)
1267{
1268 return test_bit(MPTCP_FALLBACK_DONE, &msk->flags);
1269}
1270
1271static inline bool mptcp_check_fallback(const struct sock *sk)
1272{
1273 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1274 struct mptcp_sock *msk = mptcp_sk(subflow->conn);
1275
1276 return __mptcp_check_fallback(msk);
1277}
1278
1279static inline bool __mptcp_has_initial_subflow(const struct mptcp_sock *msk)
1280{
1281 struct sock *ssk = READ_ONCE(msk->first);
1282
1283 return ssk && ((1 << inet_sk_state_load(ssk)) &
1284 (TCPF_ESTABLISHED | TCPF_SYN_SENT |
1285 TCPF_SYN_RECV | TCPF_LISTEN));
1286}
1287
1288bool __mptcp_try_fallback(struct mptcp_sock *msk, int fb_mib);
1289
1290static inline bool mptcp_try_fallback(struct sock *ssk, int fb_mib)
1291{
1292 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
1293 struct sock *sk = subflow->conn;
1294 struct mptcp_sock *msk;
1295
1296 msk = mptcp_sk(sk);
1297 if (!__mptcp_try_fallback(msk, fb_mib))
1298 return false;
1299 if (READ_ONCE(msk->snd_data_fin_enable) && !(ssk->sk_shutdown & SEND_SHUTDOWN)) {
1300 gfp_t saved_allocation = ssk->sk_allocation;
1301
1302 /* we are in a atomic (BH) scope, override ssk default for data
1303 * fin allocation
1304 */
1305 ssk->sk_allocation = GFP_ATOMIC;
1306 ssk->sk_shutdown |= SEND_SHUTDOWN;
1307 tcp_shutdown(ssk, SEND_SHUTDOWN);
1308 ssk->sk_allocation = saved_allocation;
1309 }
1310 return true;
1311}
1312
1313static inline void mptcp_early_fallback(struct mptcp_sock *msk,
1314 struct mptcp_subflow_context *subflow,
1315 int fb_mib)
1316{
1317 subflow->request_mptcp = 0;
1318 WARN_ON_ONCE(!__mptcp_try_fallback(msk, fb_mib));
1319}
1320
1321static inline bool mptcp_check_infinite_map(struct sk_buff *skb)
1322{
1323 struct mptcp_ext *mpext;
1324
1325 mpext = skb ? mptcp_get_ext(skb) : NULL;
1326 if (mpext && mpext->infinite_map)
1327 return true;
1328
1329 return false;
1330}
1331
1332static inline bool is_active_ssk(struct mptcp_subflow_context *subflow)
1333{
1334 return (subflow->request_mptcp || subflow->request_join);
1335}
1336
1337static inline bool subflow_simultaneous_connect(struct sock *sk)
1338{
1339 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1340
1341 /* Note that the sk state implies !subflow->conn_finished. */
1342 return sk->sk_state == TCP_SYN_RECV && is_active_ssk(subflow);
1343}
1344
1345#ifdef CONFIG_SYN_COOKIES
1346void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1347 struct sk_buff *skb);
1348bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1349 struct sk_buff *skb);
1350void __init mptcp_join_cookie_init(void);
1351#else
1352static inline void
1353subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1354 struct sk_buff *skb) {}
1355static inline bool
1356mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1357 struct sk_buff *skb)
1358{
1359 return false;
1360}
1361
1362static inline void mptcp_join_cookie_init(void) {}
1363#endif
1364
1365#endif /* __MPTCP_PROTOCOL_H */