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
16#define MPTCP_SUPPORTED_VERSION 1
17
18/* MPTCP option bits */
19#define OPTION_MPTCP_MPC_SYN BIT(0)
20#define OPTION_MPTCP_MPC_SYNACK BIT(1)
21#define OPTION_MPTCP_MPC_ACK BIT(2)
22#define OPTION_MPTCP_MPJ_SYN BIT(3)
23#define OPTION_MPTCP_MPJ_SYNACK BIT(4)
24#define OPTION_MPTCP_MPJ_ACK BIT(5)
25#define OPTION_MPTCP_ADD_ADDR BIT(6)
26#define OPTION_MPTCP_RM_ADDR BIT(7)
27#define OPTION_MPTCP_FASTCLOSE BIT(8)
28#define OPTION_MPTCP_PRIO BIT(9)
29#define OPTION_MPTCP_RST BIT(10)
30#define OPTION_MPTCP_DSS BIT(11)
31#define OPTION_MPTCP_FAIL BIT(12)
32
33#define OPTION_MPTCP_CSUMREQD BIT(13)
34
35#define OPTIONS_MPTCP_MPC (OPTION_MPTCP_MPC_SYN | OPTION_MPTCP_MPC_SYNACK | \
36 OPTION_MPTCP_MPC_ACK)
37#define OPTIONS_MPTCP_MPJ (OPTION_MPTCP_MPJ_SYN | OPTION_MPTCP_MPJ_SYNACK | \
38 OPTION_MPTCP_MPJ_ACK)
39
40/* MPTCP option subtypes */
41#define MPTCPOPT_MP_CAPABLE 0
42#define MPTCPOPT_MP_JOIN 1
43#define MPTCPOPT_DSS 2
44#define MPTCPOPT_ADD_ADDR 3
45#define MPTCPOPT_RM_ADDR 4
46#define MPTCPOPT_MP_PRIO 5
47#define MPTCPOPT_MP_FAIL 6
48#define MPTCPOPT_MP_FASTCLOSE 7
49#define MPTCPOPT_RST 8
50
51/* MPTCP suboption lengths */
52#define TCPOLEN_MPTCP_MPC_SYN 4
53#define TCPOLEN_MPTCP_MPC_SYNACK 12
54#define TCPOLEN_MPTCP_MPC_ACK 20
55#define TCPOLEN_MPTCP_MPC_ACK_DATA 22
56#define TCPOLEN_MPTCP_MPJ_SYN 12
57#define TCPOLEN_MPTCP_MPJ_SYNACK 16
58#define TCPOLEN_MPTCP_MPJ_ACK 24
59#define TCPOLEN_MPTCP_DSS_BASE 4
60#define TCPOLEN_MPTCP_DSS_ACK32 4
61#define TCPOLEN_MPTCP_DSS_ACK64 8
62#define TCPOLEN_MPTCP_DSS_MAP32 10
63#define TCPOLEN_MPTCP_DSS_MAP64 14
64#define TCPOLEN_MPTCP_DSS_CHECKSUM 2
65#define TCPOLEN_MPTCP_ADD_ADDR 16
66#define TCPOLEN_MPTCP_ADD_ADDR_PORT 18
67#define TCPOLEN_MPTCP_ADD_ADDR_BASE 8
68#define TCPOLEN_MPTCP_ADD_ADDR_BASE_PORT 10
69#define TCPOLEN_MPTCP_ADD_ADDR6 28
70#define TCPOLEN_MPTCP_ADD_ADDR6_PORT 30
71#define TCPOLEN_MPTCP_ADD_ADDR6_BASE 20
72#define TCPOLEN_MPTCP_ADD_ADDR6_BASE_PORT 22
73#define TCPOLEN_MPTCP_PORT_LEN 2
74#define TCPOLEN_MPTCP_PORT_ALIGN 2
75#define TCPOLEN_MPTCP_RM_ADDR_BASE 3
76#define TCPOLEN_MPTCP_PRIO 3
77#define TCPOLEN_MPTCP_PRIO_ALIGN 4
78#define TCPOLEN_MPTCP_FASTCLOSE 12
79#define TCPOLEN_MPTCP_RST 4
80#define TCPOLEN_MPTCP_FAIL 12
81
82#define TCPOLEN_MPTCP_MPC_ACK_DATA_CSUM (TCPOLEN_MPTCP_DSS_CHECKSUM + TCPOLEN_MPTCP_MPC_ACK_DATA)
83
84/* MPTCP MP_JOIN flags */
85#define MPTCPOPT_BACKUP BIT(0)
86#define MPTCPOPT_THMAC_LEN 8
87
88/* MPTCP MP_CAPABLE flags */
89#define MPTCP_VERSION_MASK (0x0F)
90#define MPTCP_CAP_CHECKSUM_REQD BIT(7)
91#define MPTCP_CAP_EXTENSIBILITY BIT(6)
92#define MPTCP_CAP_DENY_JOIN_ID0 BIT(5)
93#define MPTCP_CAP_HMAC_SHA256 BIT(0)
94#define MPTCP_CAP_FLAG_MASK (0x1F)
95
96/* MPTCP DSS flags */
97#define MPTCP_DSS_DATA_FIN BIT(4)
98#define MPTCP_DSS_DSN64 BIT(3)
99#define MPTCP_DSS_HAS_MAP BIT(2)
100#define MPTCP_DSS_ACK64 BIT(1)
101#define MPTCP_DSS_HAS_ACK BIT(0)
102#define MPTCP_DSS_FLAG_MASK (0x1F)
103
104/* MPTCP ADD_ADDR flags */
105#define MPTCP_ADDR_ECHO BIT(0)
106
107/* MPTCP MP_PRIO flags */
108#define MPTCP_PRIO_BKUP BIT(0)
109
110/* MPTCP TCPRST flags */
111#define MPTCP_RST_TRANSIENT BIT(0)
112
113/* MPTCP socket atomic flags */
114#define MPTCP_NOSPACE 1
115#define MPTCP_WORK_RTX 2
116#define MPTCP_FALLBACK_DONE 4
117#define MPTCP_WORK_CLOSE_SUBFLOW 5
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_CONNECTED 6
126#define MPTCP_RESET_SCHEDULER 7
127
128struct 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
137static 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
144struct 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
176static 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
182enum 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
194enum 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
205enum 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
214struct 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
239struct 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
247struct 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 */
258struct mptcp_sock {
259 /* inet_connection_sock must be the first member */
260 struct inet_connection_sock sk;
261 u64 local_key;
262 u64 remote_key;
263 u64 write_seq;
264 u64 bytes_sent;
265 u64 snd_nxt;
266 u64 bytes_received;
267 u64 ack_seq;
268 atomic64_t rcv_wnd_sent;
269 u64 rcv_data_fin_seq;
270 u64 bytes_retrans;
271 int rmem_fwd_alloc;
272 struct sock *last_snd;
273 int snd_burst;
274 int old_wspace;
275 u64 recovery_snd_nxt; /* in recovery mode accept up to this seq;
276 * recovery related fields are under data_lock
277 * protection
278 */
279 u64 bytes_acked;
280 u64 snd_una;
281 u64 wnd_end;
282 unsigned long timer_ival;
283 u32 token;
284 int rmem_released;
285 unsigned long flags;
286 unsigned long cb_flags;
287 unsigned long push_pending;
288 bool recovery; /* closing subflow write queue reinjected */
289 bool can_ack;
290 bool fully_established;
291 bool rcv_data_fin;
292 bool snd_data_fin_enable;
293 bool rcv_fastclose;
294 bool use_64bit_ack; /* Set when we received a 64-bit DSN */
295 bool csum_enabled;
296 bool allow_infinite_fallback;
297 u8 mpc_endpoint_id;
298 u8 recvmsg_inq:1,
299 cork:1,
300 nodelay:1,
301 fastopening:1,
302 in_accept_queue:1;
303 struct work_struct work;
304 struct sk_buff *ooo_last_skb;
305 struct rb_root out_of_order_queue;
306 struct sk_buff_head receive_queue;
307 struct list_head conn_list;
308 struct list_head rtx_queue;
309 struct mptcp_data_frag *first_pending;
310 struct list_head join_list;
311 struct socket *subflow; /* outgoing connect/listener/!mp_capable
312 * The mptcp ops can safely dereference, using suitable
313 * ONCE annotation, the subflow outside the socket
314 * lock as such sock is freed after close().
315 */
316 struct sock *first;
317 struct mptcp_pm_data pm;
318 struct {
319 u32 space; /* bytes copied in last measurement window */
320 u32 copied; /* bytes copied in this measurement window */
321 u64 time; /* start time of measurement window */
322 u64 rtt_us; /* last maximum rtt of subflows */
323 } rcvq_space;
324
325 u32 subflow_id;
326 u32 setsockopt_seq;
327 char ca_name[TCP_CA_NAME_MAX];
328};
329
330#define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock)
331#define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock)
332
333#define mptcp_for_each_subflow(__msk, __subflow) \
334 list_for_each_entry(__subflow, &((__msk)->conn_list), node)
335#define mptcp_for_each_subflow_safe(__msk, __subflow, __tmp) \
336 list_for_each_entry_safe(__subflow, __tmp, &((__msk)->conn_list), node)
337
338static inline void msk_owned_by_me(const struct mptcp_sock *msk)
339{
340 sock_owned_by_me((const struct sock *)msk);
341}
342
343#define mptcp_sk(ptr) container_of_const(ptr, struct mptcp_sock, sk.icsk_inet.sk)
344
345/* the msk socket don't use the backlog, also account for the bulk
346 * free memory
347 */
348static inline int __mptcp_rmem(const struct sock *sk)
349{
350 return atomic_read(&sk->sk_rmem_alloc) - READ_ONCE(mptcp_sk(sk)->rmem_released);
351}
352
353static inline int __mptcp_space(const struct sock *sk)
354{
355 return tcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - __mptcp_rmem(sk));
356}
357
358static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk)
359{
360 const struct mptcp_sock *msk = mptcp_sk(sk);
361
362 return READ_ONCE(msk->first_pending);
363}
364
365static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk)
366{
367 struct mptcp_sock *msk = mptcp_sk(sk);
368 struct mptcp_data_frag *cur;
369
370 cur = msk->first_pending;
371 return list_is_last(&cur->list, &msk->rtx_queue) ? NULL :
372 list_next_entry(cur, list);
373}
374
375static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk)
376{
377 const struct mptcp_sock *msk = mptcp_sk(sk);
378
379 if (!msk->first_pending)
380 return NULL;
381
382 if (WARN_ON_ONCE(list_empty(&msk->rtx_queue)))
383 return NULL;
384
385 return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list);
386}
387
388static inline struct mptcp_data_frag *mptcp_rtx_head(struct sock *sk)
389{
390 struct mptcp_sock *msk = mptcp_sk(sk);
391
392 if (msk->snd_una == READ_ONCE(msk->snd_nxt))
393 return NULL;
394
395 return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list);
396}
397
398struct csum_pseudo_header {
399 __be64 data_seq;
400 __be32 subflow_seq;
401 __be16 data_len;
402 __sum16 csum;
403};
404
405struct mptcp_subflow_request_sock {
406 struct tcp_request_sock sk;
407 u16 mp_capable : 1,
408 mp_join : 1,
409 backup : 1,
410 csum_reqd : 1,
411 allow_join_id0 : 1;
412 u8 local_id;
413 u8 remote_id;
414 u64 local_key;
415 u64 idsn;
416 u32 token;
417 u32 ssn_offset;
418 u64 thmac;
419 u32 local_nonce;
420 u32 remote_nonce;
421 struct mptcp_sock *msk;
422 struct hlist_nulls_node token_node;
423};
424
425static inline struct mptcp_subflow_request_sock *
426mptcp_subflow_rsk(const struct request_sock *rsk)
427{
428 return (struct mptcp_subflow_request_sock *)rsk;
429}
430
431enum mptcp_data_avail {
432 MPTCP_SUBFLOW_NODATA,
433 MPTCP_SUBFLOW_DATA_AVAIL,
434};
435
436struct mptcp_delegated_action {
437 struct napi_struct napi;
438 struct list_head head;
439};
440
441DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions);
442
443#define MPTCP_DELEGATE_SEND 0
444#define MPTCP_DELEGATE_ACK 1
445
446/* MPTCP subflow context */
447struct mptcp_subflow_context {
448 struct list_head node;/* conn_list of subflows */
449
450 struct_group(reset,
451
452 unsigned long avg_pacing_rate; /* protected by msk socket lock */
453 u64 local_key;
454 u64 remote_key;
455 u64 idsn;
456 u64 map_seq;
457 u32 snd_isn;
458 u32 token;
459 u32 rel_write_seq;
460 u32 map_subflow_seq;
461 u32 ssn_offset;
462 u32 map_data_len;
463 __wsum map_data_csum;
464 u32 map_csum_len;
465 u32 request_mptcp : 1, /* send MP_CAPABLE */
466 request_join : 1, /* send MP_JOIN */
467 request_bkup : 1,
468 mp_capable : 1, /* remote is MPTCP capable */
469 mp_join : 1, /* remote is JOINing */
470 fully_established : 1, /* path validated */
471 pm_notified : 1, /* PM hook called for established status */
472 conn_finished : 1,
473 map_valid : 1,
474 map_csum_reqd : 1,
475 map_data_fin : 1,
476 mpc_map : 1,
477 backup : 1,
478 send_mp_prio : 1,
479 send_mp_fail : 1,
480 send_fastclose : 1,
481 send_infinite_map : 1,
482 remote_key_valid : 1, /* received the peer key from */
483 disposable : 1, /* ctx can be free at ulp release time */
484 stale : 1, /* unable to snd/rcv data, do not use for xmit */
485 local_id_valid : 1, /* local_id is correctly initialized */
486 valid_csum_seen : 1, /* at least one csum validated */
487 is_mptfo : 1, /* subflow is doing TFO */
488 __unused : 9;
489 enum mptcp_data_avail data_avail;
490 u32 remote_nonce;
491 u64 thmac;
492 u32 local_nonce;
493 u32 remote_token;
494 union {
495 u8 hmac[MPTCPOPT_HMAC_LEN]; /* MPJ subflow only */
496 u64 iasn; /* initial ack sequence number, MPC subflows only */
497 };
498 u8 local_id;
499 u8 remote_id;
500 u8 reset_seen:1;
501 u8 reset_transient:1;
502 u8 reset_reason:4;
503 u8 stale_count;
504
505 u32 subflow_id;
506
507 long delegated_status;
508 unsigned long fail_tout;
509
510 );
511
512 struct list_head delegated_node; /* link into delegated_action, protected by local BH */
513
514 u32 setsockopt_seq;
515 u32 stale_rcv_tstamp;
516
517 struct sock *tcp_sock; /* tcp sk backpointer */
518 struct sock *conn; /* parent mptcp_sock */
519 const struct inet_connection_sock_af_ops *icsk_af_ops;
520 void (*tcp_state_change)(struct sock *sk);
521 void (*tcp_error_report)(struct sock *sk);
522
523 struct rcu_head rcu;
524};
525
526static inline struct mptcp_subflow_context *
527mptcp_subflow_ctx(const struct sock *sk)
528{
529 struct inet_connection_sock *icsk = inet_csk(sk);
530
531 /* Use RCU on icsk_ulp_data only for sock diag code */
532 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data;
533}
534
535static inline struct sock *
536mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow)
537{
538 return subflow->tcp_sock;
539}
540
541static inline void
542mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow)
543{
544 memset(&subflow->reset, 0, sizeof(subflow->reset));
545 subflow->request_mptcp = 1;
546}
547
548static inline u64
549mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow)
550{
551 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq -
552 subflow->ssn_offset -
553 subflow->map_subflow_seq;
554}
555
556static inline u64
557mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow)
558{
559 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow);
560}
561
562void mptcp_subflow_process_delegated(struct sock *ssk);
563
564static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action)
565{
566 struct mptcp_delegated_action *delegated;
567 bool schedule;
568
569 /* the caller held the subflow bh socket lock */
570 lockdep_assert_in_softirq();
571
572 /* The implied barrier pairs with mptcp_subflow_delegated_done(), and
573 * ensures the below list check sees list updates done prior to status
574 * bit changes
575 */
576 if (!test_and_set_bit(action, &subflow->delegated_status)) {
577 /* still on delegated list from previous scheduling */
578 if (!list_empty(&subflow->delegated_node))
579 return;
580
581 delegated = this_cpu_ptr(&mptcp_delegated_actions);
582 schedule = list_empty(&delegated->head);
583 list_add_tail(&subflow->delegated_node, &delegated->head);
584 sock_hold(mptcp_subflow_tcp_sock(subflow));
585 if (schedule)
586 napi_schedule(&delegated->napi);
587 }
588}
589
590static inline struct mptcp_subflow_context *
591mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated)
592{
593 struct mptcp_subflow_context *ret;
594
595 if (list_empty(&delegated->head))
596 return NULL;
597
598 ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node);
599 list_del_init(&ret->delegated_node);
600 return ret;
601}
602
603static inline bool mptcp_subflow_has_delegated_action(const struct mptcp_subflow_context *subflow)
604{
605 return !!READ_ONCE(subflow->delegated_status);
606}
607
608static inline void mptcp_subflow_delegated_done(struct mptcp_subflow_context *subflow, int action)
609{
610 /* pairs with mptcp_subflow_delegate, ensures delegate_node is updated before
611 * touching the status bit
612 */
613 smp_wmb();
614 clear_bit(action, &subflow->delegated_status);
615}
616
617int mptcp_is_enabled(const struct net *net);
618unsigned int mptcp_get_add_addr_timeout(const struct net *net);
619int mptcp_is_checksum_enabled(const struct net *net);
620int mptcp_allow_join_id0(const struct net *net);
621unsigned int mptcp_stale_loss_cnt(const struct net *net);
622int mptcp_get_pm_type(const struct net *net);
623void mptcp_subflow_fully_established(struct mptcp_subflow_context *subflow,
624 const struct mptcp_options_received *mp_opt);
625bool __mptcp_retransmit_pending_data(struct sock *sk);
626void mptcp_check_and_set_pending(struct sock *sk);
627void __mptcp_push_pending(struct sock *sk, unsigned int flags);
628bool mptcp_subflow_data_available(struct sock *sk);
629void __init mptcp_subflow_init(void);
630void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how);
631void mptcp_close_ssk(struct sock *sk, struct sock *ssk,
632 struct mptcp_subflow_context *subflow);
633void __mptcp_subflow_send_ack(struct sock *ssk);
634void mptcp_subflow_reset(struct sock *ssk);
635void mptcp_subflow_queue_clean(struct sock *sk, struct sock *ssk);
636void mptcp_sock_graft(struct sock *sk, struct socket *parent);
637struct socket *__mptcp_nmpc_socket(struct mptcp_sock *msk);
638bool __mptcp_close(struct sock *sk, long timeout);
639void mptcp_cancel_work(struct sock *sk);
640void __mptcp_unaccepted_force_close(struct sock *sk);
641void mptcp_set_owner_r(struct sk_buff *skb, struct sock *sk);
642
643bool mptcp_addresses_equal(const struct mptcp_addr_info *a,
644 const struct mptcp_addr_info *b, bool use_port);
645void mptcp_local_address(const struct sock_common *skc, struct mptcp_addr_info *addr);
646
647/* called with sk socket lock held */
648int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc,
649 const struct mptcp_addr_info *remote);
650int mptcp_subflow_create_socket(struct sock *sk, unsigned short family,
651 struct socket **new_sock);
652void mptcp_info2sockaddr(const struct mptcp_addr_info *info,
653 struct sockaddr_storage *addr,
654 unsigned short family);
655
656static inline bool __tcp_can_send(const struct sock *ssk)
657{
658 /* only send if our side has not closed yet */
659 return ((1 << inet_sk_state_load(ssk)) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT));
660}
661
662static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow)
663{
664 /* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */
665 if (subflow->request_join && !subflow->fully_established)
666 return false;
667
668 return __tcp_can_send(mptcp_subflow_tcp_sock(subflow));
669}
670
671void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow);
672
673bool mptcp_subflow_active(struct mptcp_subflow_context *subflow);
674
675void mptcp_subflow_drop_ctx(struct sock *ssk);
676
677static inline void mptcp_subflow_tcp_fallback(struct sock *sk,
678 struct mptcp_subflow_context *ctx)
679{
680 sk->sk_data_ready = sock_def_readable;
681 sk->sk_state_change = ctx->tcp_state_change;
682 sk->sk_write_space = sk_stream_write_space;
683 sk->sk_error_report = ctx->tcp_error_report;
684
685 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops;
686}
687
688void __init mptcp_proto_init(void);
689#if IS_ENABLED(CONFIG_MPTCP_IPV6)
690int __init mptcp_proto_v6_init(void);
691#endif
692
693struct sock *mptcp_sk_clone_init(const struct sock *sk,
694 const struct mptcp_options_received *mp_opt,
695 struct sock *ssk,
696 struct request_sock *req);
697void mptcp_get_options(const struct sk_buff *skb,
698 struct mptcp_options_received *mp_opt);
699
700void mptcp_finish_connect(struct sock *sk);
701void __mptcp_set_connected(struct sock *sk);
702void mptcp_reset_timeout(struct mptcp_sock *msk, unsigned long fail_tout);
703static inline bool mptcp_is_fully_established(struct sock *sk)
704{
705 return inet_sk_state_load(sk) == TCP_ESTABLISHED &&
706 READ_ONCE(mptcp_sk(sk)->fully_established);
707}
708void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk);
709void mptcp_data_ready(struct sock *sk, struct sock *ssk);
710bool mptcp_finish_join(struct sock *sk);
711bool mptcp_schedule_work(struct sock *sk);
712int mptcp_setsockopt(struct sock *sk, int level, int optname,
713 sockptr_t optval, unsigned int optlen);
714int mptcp_getsockopt(struct sock *sk, int level, int optname,
715 char __user *optval, int __user *option);
716
717u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq);
718static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit)
719{
720 if (use_64bit)
721 return cur_seq;
722
723 return __mptcp_expand_seq(old_seq, cur_seq);
724}
725void __mptcp_check_push(struct sock *sk, struct sock *ssk);
726void __mptcp_data_acked(struct sock *sk);
727void __mptcp_error_report(struct sock *sk);
728bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit);
729static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk)
730{
731 return READ_ONCE(msk->snd_data_fin_enable) &&
732 READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt);
733}
734
735static inline bool mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
736{
737 if ((sk->sk_userlocks & SOCK_SNDBUF_LOCK) || ssk->sk_sndbuf <= READ_ONCE(sk->sk_sndbuf))
738 return false;
739
740 WRITE_ONCE(sk->sk_sndbuf, ssk->sk_sndbuf);
741 return true;
742}
743
744static inline void mptcp_write_space(struct sock *sk)
745{
746 if (sk_stream_is_writeable(sk)) {
747 /* pairs with memory barrier in mptcp_poll */
748 smp_mb();
749 if (test_and_clear_bit(MPTCP_NOSPACE, &mptcp_sk(sk)->flags))
750 sk_stream_write_space(sk);
751 }
752}
753
754void mptcp_destroy_common(struct mptcp_sock *msk, unsigned int flags);
755
756#define MPTCP_TOKEN_MAX_RETRIES 4
757
758void __init mptcp_token_init(void);
759static inline void mptcp_token_init_request(struct request_sock *req)
760{
761 mptcp_subflow_rsk(req)->token_node.pprev = NULL;
762}
763
764int mptcp_token_new_request(struct request_sock *req);
765void mptcp_token_destroy_request(struct request_sock *req);
766int mptcp_token_new_connect(struct sock *ssk);
767void mptcp_token_accept(struct mptcp_subflow_request_sock *r,
768 struct mptcp_sock *msk);
769bool mptcp_token_exists(u32 token);
770struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token);
771struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot,
772 long *s_num);
773void mptcp_token_destroy(struct mptcp_sock *msk);
774
775void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn);
776
777void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac);
778__sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum);
779
780void __init mptcp_pm_init(void);
781void mptcp_pm_data_init(struct mptcp_sock *msk);
782void mptcp_pm_data_reset(struct mptcp_sock *msk);
783int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info,
784 struct mptcp_addr_info *addr);
785int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info,
786 bool require_family,
787 struct mptcp_pm_addr_entry *entry);
788bool mptcp_pm_addr_families_match(const struct sock *sk,
789 const struct mptcp_addr_info *loc,
790 const struct mptcp_addr_info *rem);
791void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
792void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
793void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side);
794void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk);
795bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk);
796void mptcp_pm_connection_closed(struct mptcp_sock *msk);
797void mptcp_pm_subflow_established(struct mptcp_sock *msk);
798bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk);
799void mptcp_pm_subflow_check_next(struct mptcp_sock *msk, const struct sock *ssk,
800 const struct mptcp_subflow_context *subflow);
801void mptcp_pm_add_addr_received(const struct sock *ssk,
802 const struct mptcp_addr_info *addr);
803void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk,
804 const struct mptcp_addr_info *addr);
805void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk);
806void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk);
807void mptcp_pm_rm_addr_received(struct mptcp_sock *msk,
808 const struct mptcp_rm_list *rm_list);
809void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup);
810void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq);
811int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk,
812 struct mptcp_addr_info *addr,
813 struct mptcp_addr_info *rem,
814 u8 bkup);
815bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk,
816 const struct mptcp_addr_info *addr);
817void mptcp_pm_free_anno_list(struct mptcp_sock *msk);
818bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk);
819struct mptcp_pm_add_entry *
820mptcp_pm_del_add_timer(struct mptcp_sock *msk,
821 const struct mptcp_addr_info *addr, bool check_id);
822struct mptcp_pm_add_entry *
823mptcp_lookup_anno_list_by_saddr(const struct mptcp_sock *msk,
824 const struct mptcp_addr_info *addr);
825int mptcp_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk,
826 unsigned int id,
827 u8 *flags, int *ifindex);
828int mptcp_pm_nl_get_flags_and_ifindex_by_id(struct mptcp_sock *msk, unsigned int id,
829 u8 *flags, int *ifindex);
830int mptcp_userspace_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk,
831 unsigned int id,
832 u8 *flags, int *ifindex);
833int mptcp_pm_set_flags(struct net *net, struct nlattr *token,
834 struct mptcp_pm_addr_entry *loc,
835 struct mptcp_pm_addr_entry *rem, u8 bkup);
836int mptcp_pm_nl_set_flags(struct net *net, struct mptcp_pm_addr_entry *addr, u8 bkup);
837int mptcp_userspace_pm_set_flags(struct net *net, struct nlattr *token,
838 struct mptcp_pm_addr_entry *loc,
839 struct mptcp_pm_addr_entry *rem, u8 bkup);
840int mptcp_pm_announce_addr(struct mptcp_sock *msk,
841 const struct mptcp_addr_info *addr,
842 bool echo);
843int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
844int mptcp_pm_remove_subflow(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
845void mptcp_pm_remove_addrs(struct mptcp_sock *msk, struct list_head *rm_list);
846void mptcp_pm_remove_addrs_and_subflows(struct mptcp_sock *msk,
847 struct list_head *rm_list);
848
849void mptcp_free_local_addr_list(struct mptcp_sock *msk);
850int mptcp_nl_cmd_announce(struct sk_buff *skb, struct genl_info *info);
851int mptcp_nl_cmd_remove(struct sk_buff *skb, struct genl_info *info);
852int mptcp_nl_cmd_sf_create(struct sk_buff *skb, struct genl_info *info);
853int mptcp_nl_cmd_sf_destroy(struct sk_buff *skb, struct genl_info *info);
854
855void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk,
856 const struct sock *ssk, gfp_t gfp);
857void mptcp_event_addr_announced(const struct sock *ssk, const struct mptcp_addr_info *info);
858void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id);
859void mptcp_event_pm_listener(const struct sock *ssk,
860 enum mptcp_event_type event);
861bool mptcp_userspace_pm_active(const struct mptcp_sock *msk);
862
863void mptcp_fastopen_gen_msk_ackseq(struct mptcp_sock *msk, struct mptcp_subflow_context *subflow,
864 const struct mptcp_options_received *mp_opt);
865void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subflow,
866 struct request_sock *req);
867
868static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk)
869{
870 return READ_ONCE(msk->pm.addr_signal) &
871 (BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO));
872}
873
874static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk)
875{
876 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL);
877}
878
879static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk)
880{
881 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO);
882}
883
884static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk)
885{
886 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL);
887}
888
889static inline bool mptcp_pm_is_userspace(const struct mptcp_sock *msk)
890{
891 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_USERSPACE;
892}
893
894static inline bool mptcp_pm_is_kernel(const struct mptcp_sock *msk)
895{
896 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_KERNEL;
897}
898
899static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port)
900{
901 u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE;
902
903 if (family == AF_INET6)
904 len = TCPOLEN_MPTCP_ADD_ADDR6_BASE;
905 if (!echo)
906 len += MPTCPOPT_THMAC_LEN;
907 /* account for 2 trailing 'nop' options */
908 if (port)
909 len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN;
910
911 return len;
912}
913
914static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list)
915{
916 if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX)
917 return -EINVAL;
918
919 return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1;
920}
921
922bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, const struct sk_buff *skb,
923 unsigned int opt_size, unsigned int remaining,
924 struct mptcp_addr_info *addr, bool *echo,
925 bool *drop_other_suboptions);
926bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining,
927 struct mptcp_rm_list *rm_list);
928int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
929int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
930int mptcp_userspace_pm_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
931
932void __init mptcp_pm_nl_init(void);
933void mptcp_pm_nl_work(struct mptcp_sock *msk);
934void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk,
935 const struct mptcp_rm_list *rm_list);
936unsigned int mptcp_pm_get_add_addr_signal_max(const struct mptcp_sock *msk);
937unsigned int mptcp_pm_get_add_addr_accept_max(const struct mptcp_sock *msk);
938unsigned int mptcp_pm_get_subflows_max(const struct mptcp_sock *msk);
939unsigned int mptcp_pm_get_local_addr_max(const struct mptcp_sock *msk);
940
941/* called under PM lock */
942static inline void __mptcp_pm_close_subflow(struct mptcp_sock *msk)
943{
944 if (--msk->pm.subflows < mptcp_pm_get_subflows_max(msk))
945 WRITE_ONCE(msk->pm.accept_subflow, true);
946}
947
948static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk)
949{
950 spin_lock_bh(&msk->pm.lock);
951 __mptcp_pm_close_subflow(msk);
952 spin_unlock_bh(&msk->pm.lock);
953}
954
955void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk);
956void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk);
957
958static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb)
959{
960 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP);
961}
962
963void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops);
964
965static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk)
966{
967 return test_bit(MPTCP_FALLBACK_DONE, &msk->flags);
968}
969
970static inline bool mptcp_check_fallback(const struct sock *sk)
971{
972 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
973 struct mptcp_sock *msk = mptcp_sk(subflow->conn);
974
975 return __mptcp_check_fallback(msk);
976}
977
978static inline void __mptcp_do_fallback(struct mptcp_sock *msk)
979{
980 if (test_bit(MPTCP_FALLBACK_DONE, &msk->flags)) {
981 pr_debug("TCP fallback already done (msk=%p)", msk);
982 return;
983 }
984 set_bit(MPTCP_FALLBACK_DONE, &msk->flags);
985}
986
987static inline void mptcp_do_fallback(struct sock *ssk)
988{
989 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
990 struct sock *sk = subflow->conn;
991 struct mptcp_sock *msk;
992
993 msk = mptcp_sk(sk);
994 __mptcp_do_fallback(msk);
995 if (READ_ONCE(msk->snd_data_fin_enable) && !(ssk->sk_shutdown & SEND_SHUTDOWN)) {
996 gfp_t saved_allocation = ssk->sk_allocation;
997
998 /* we are in a atomic (BH) scope, override ssk default for data
999 * fin allocation
1000 */
1001 ssk->sk_allocation = GFP_ATOMIC;
1002 ssk->sk_shutdown |= SEND_SHUTDOWN;
1003 tcp_shutdown(ssk, SEND_SHUTDOWN);
1004 ssk->sk_allocation = saved_allocation;
1005 }
1006}
1007
1008#define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a)
1009
1010static inline bool mptcp_check_infinite_map(struct sk_buff *skb)
1011{
1012 struct mptcp_ext *mpext;
1013
1014 mpext = skb ? mptcp_get_ext(skb) : NULL;
1015 if (mpext && mpext->infinite_map)
1016 return true;
1017
1018 return false;
1019}
1020
1021static inline bool is_active_ssk(struct mptcp_subflow_context *subflow)
1022{
1023 return (subflow->request_mptcp || subflow->request_join);
1024}
1025
1026static inline bool subflow_simultaneous_connect(struct sock *sk)
1027{
1028 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1029
1030 return sk->sk_state == TCP_ESTABLISHED &&
1031 is_active_ssk(subflow) &&
1032 !subflow->conn_finished;
1033}
1034
1035#ifdef CONFIG_SYN_COOKIES
1036void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1037 struct sk_buff *skb);
1038bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1039 struct sk_buff *skb);
1040void __init mptcp_join_cookie_init(void);
1041#else
1042static inline void
1043subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1044 struct sk_buff *skb) {}
1045static inline bool
1046mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1047 struct sk_buff *skb)
1048{
1049 return false;
1050}
1051
1052static inline void mptcp_join_cookie_init(void) {}
1053#endif
1054
1055#endif /* __MPTCP_PROTOCOL_H */