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_WORK_EOF 3
117#define MPTCP_FALLBACK_DONE 4
118#define MPTCP_WORK_CLOSE_SUBFLOW 5
119
120/* MPTCP socket release cb flags */
121#define MPTCP_PUSH_PENDING 1
122#define MPTCP_CLEAN_UNA 2
123#define MPTCP_ERROR_REPORT 3
124#define MPTCP_RETRANSMIT 4
125#define MPTCP_FLUSH_JOIN_LIST 5
126#define MPTCP_CONNECTED 6
127#define MPTCP_RESET_SCHEDULER 7
128
129static inline bool before64(__u64 seq1, __u64 seq2)
130{
131 return (__s64)(seq1 - seq2) < 0;
132}
133
134#define after64(seq2, seq1) before64(seq1, seq2)
135
136struct mptcp_options_received {
137 u64 sndr_key;
138 u64 rcvr_key;
139 u64 data_ack;
140 u64 data_seq;
141 u32 subflow_seq;
142 u16 data_len;
143 __sum16 csum;
144 u16 suboptions;
145 u32 token;
146 u32 nonce;
147 u16 use_map:1,
148 dsn64:1,
149 data_fin:1,
150 use_ack:1,
151 ack64:1,
152 mpc_map:1,
153 reset_reason:4,
154 reset_transient:1,
155 echo:1,
156 backup:1,
157 deny_join_id0:1,
158 __unused:2;
159 u8 join_id;
160 u64 thmac;
161 u8 hmac[MPTCPOPT_HMAC_LEN];
162 struct mptcp_addr_info addr;
163 struct mptcp_rm_list rm_list;
164 u64 ahmac;
165 u64 fail_seq;
166};
167
168static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field)
169{
170 return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) |
171 ((nib & 0xF) << 8) | field);
172}
173
174enum mptcp_pm_status {
175 MPTCP_PM_ADD_ADDR_RECEIVED,
176 MPTCP_PM_ADD_ADDR_SEND_ACK,
177 MPTCP_PM_RM_ADDR_RECEIVED,
178 MPTCP_PM_ESTABLISHED,
179 MPTCP_PM_SUBFLOW_ESTABLISHED,
180 MPTCP_PM_ALREADY_ESTABLISHED, /* persistent status, set after ESTABLISHED event */
181 MPTCP_PM_MPC_ENDPOINT_ACCOUNTED /* persistent status, set after MPC local address is
182 * accounted int id_avail_bitmap
183 */
184};
185
186enum mptcp_pm_type {
187 MPTCP_PM_TYPE_KERNEL = 0,
188 MPTCP_PM_TYPE_USERSPACE,
189
190 __MPTCP_PM_TYPE_NR,
191 __MPTCP_PM_TYPE_MAX = __MPTCP_PM_TYPE_NR - 1,
192};
193
194/* Status bits below MPTCP_PM_ALREADY_ESTABLISHED need pm worker actions */
195#define MPTCP_PM_WORK_MASK ((1 << MPTCP_PM_ALREADY_ESTABLISHED) - 1)
196
197enum mptcp_addr_signal_status {
198 MPTCP_ADD_ADDR_SIGNAL,
199 MPTCP_ADD_ADDR_ECHO,
200 MPTCP_RM_ADDR_SIGNAL,
201};
202
203/* max value of mptcp_addr_info.id */
204#define MPTCP_PM_MAX_ADDR_ID U8_MAX
205
206struct mptcp_pm_data {
207 struct mptcp_addr_info local;
208 struct mptcp_addr_info remote;
209 struct list_head anno_list;
210 struct list_head userspace_pm_local_addr_list;
211
212 spinlock_t lock; /*protects the whole PM data */
213
214 u8 addr_signal;
215 bool server_side;
216 bool work_pending;
217 bool accept_addr;
218 bool accept_subflow;
219 bool remote_deny_join_id0;
220 u8 add_addr_signaled;
221 u8 add_addr_accepted;
222 u8 local_addr_used;
223 u8 pm_type;
224 u8 subflows;
225 u8 status;
226 DECLARE_BITMAP(id_avail_bitmap, MPTCP_PM_MAX_ADDR_ID + 1);
227 struct mptcp_rm_list rm_list_tx;
228 struct mptcp_rm_list rm_list_rx;
229};
230
231struct mptcp_pm_addr_entry {
232 struct list_head list;
233 struct mptcp_addr_info addr;
234 u8 flags;
235 int ifindex;
236 struct socket *lsk;
237};
238
239struct mptcp_data_frag {
240 struct list_head list;
241 u64 data_seq;
242 u16 data_len;
243 u16 offset;
244 u16 overhead;
245 u16 already_sent;
246 struct page *page;
247};
248
249/* MPTCP connection sock */
250struct mptcp_sock {
251 /* inet_connection_sock must be the first member */
252 struct inet_connection_sock sk;
253 u64 local_key;
254 u64 remote_key;
255 u64 write_seq;
256 u64 snd_nxt;
257 u64 ack_seq;
258 atomic64_t rcv_wnd_sent;
259 u64 rcv_data_fin_seq;
260 int rmem_fwd_alloc;
261 struct sock *last_snd;
262 int snd_burst;
263 int old_wspace;
264 u64 recovery_snd_nxt; /* in recovery mode accept up to this seq;
265 * recovery related fields are under data_lock
266 * protection
267 */
268 u64 snd_una;
269 u64 wnd_end;
270 unsigned long timer_ival;
271 u32 token;
272 int rmem_released;
273 unsigned long flags;
274 unsigned long cb_flags;
275 unsigned long push_pending;
276 bool recovery; /* closing subflow write queue reinjected */
277 bool can_ack;
278 bool fully_established;
279 bool rcv_data_fin;
280 bool snd_data_fin_enable;
281 bool rcv_fastclose;
282 bool use_64bit_ack; /* Set when we received a 64-bit DSN */
283 bool csum_enabled;
284 bool allow_infinite_fallback;
285 u8 mpc_endpoint_id;
286 u8 recvmsg_inq:1,
287 cork:1,
288 nodelay:1;
289 struct work_struct work;
290 struct sk_buff *ooo_last_skb;
291 struct rb_root out_of_order_queue;
292 struct sk_buff_head receive_queue;
293 struct list_head conn_list;
294 struct list_head rtx_queue;
295 struct mptcp_data_frag *first_pending;
296 struct list_head join_list;
297 struct socket *subflow; /* outgoing connect/listener/!mp_capable */
298 struct sock *first;
299 struct mptcp_pm_data pm;
300 struct {
301 u32 space; /* bytes copied in last measurement window */
302 u32 copied; /* bytes copied in this measurement window */
303 u64 time; /* start time of measurement window */
304 u64 rtt_us; /* last maximum rtt of subflows */
305 } rcvq_space;
306
307 u32 setsockopt_seq;
308 char ca_name[TCP_CA_NAME_MAX];
309 struct mptcp_sock *dl_next;
310};
311
312#define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock)
313#define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock)
314
315#define mptcp_for_each_subflow(__msk, __subflow) \
316 list_for_each_entry(__subflow, &((__msk)->conn_list), node)
317
318static inline void msk_owned_by_me(const struct mptcp_sock *msk)
319{
320 sock_owned_by_me((const struct sock *)msk);
321}
322
323static inline struct mptcp_sock *mptcp_sk(const struct sock *sk)
324{
325 return (struct mptcp_sock *)sk;
326}
327
328/* the msk socket don't use the backlog, also account for the bulk
329 * free memory
330 */
331static inline int __mptcp_rmem(const struct sock *sk)
332{
333 return atomic_read(&sk->sk_rmem_alloc) - READ_ONCE(mptcp_sk(sk)->rmem_released);
334}
335
336static inline int __mptcp_space(const struct sock *sk)
337{
338 return tcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - __mptcp_rmem(sk));
339}
340
341static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk)
342{
343 const struct mptcp_sock *msk = mptcp_sk(sk);
344
345 return READ_ONCE(msk->first_pending);
346}
347
348static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk)
349{
350 struct mptcp_sock *msk = mptcp_sk(sk);
351 struct mptcp_data_frag *cur;
352
353 cur = msk->first_pending;
354 return list_is_last(&cur->list, &msk->rtx_queue) ? NULL :
355 list_next_entry(cur, list);
356}
357
358static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk)
359{
360 struct mptcp_sock *msk = mptcp_sk(sk);
361
362 if (!msk->first_pending)
363 return NULL;
364
365 if (WARN_ON_ONCE(list_empty(&msk->rtx_queue)))
366 return NULL;
367
368 return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list);
369}
370
371static inline struct mptcp_data_frag *mptcp_rtx_head(const struct sock *sk)
372{
373 struct mptcp_sock *msk = mptcp_sk(sk);
374
375 if (msk->snd_una == READ_ONCE(msk->snd_nxt))
376 return NULL;
377
378 return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list);
379}
380
381struct csum_pseudo_header {
382 __be64 data_seq;
383 __be32 subflow_seq;
384 __be16 data_len;
385 __sum16 csum;
386};
387
388struct mptcp_subflow_request_sock {
389 struct tcp_request_sock sk;
390 u16 mp_capable : 1,
391 mp_join : 1,
392 backup : 1,
393 csum_reqd : 1,
394 allow_join_id0 : 1;
395 u8 local_id;
396 u8 remote_id;
397 u64 local_key;
398 u64 idsn;
399 u32 token;
400 u32 ssn_offset;
401 u64 thmac;
402 u32 local_nonce;
403 u32 remote_nonce;
404 struct mptcp_sock *msk;
405 struct hlist_nulls_node token_node;
406};
407
408static inline struct mptcp_subflow_request_sock *
409mptcp_subflow_rsk(const struct request_sock *rsk)
410{
411 return (struct mptcp_subflow_request_sock *)rsk;
412}
413
414enum mptcp_data_avail {
415 MPTCP_SUBFLOW_NODATA,
416 MPTCP_SUBFLOW_DATA_AVAIL,
417};
418
419struct mptcp_delegated_action {
420 struct napi_struct napi;
421 struct list_head head;
422};
423
424DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions);
425
426#define MPTCP_DELEGATE_SEND 0
427#define MPTCP_DELEGATE_ACK 1
428
429/* MPTCP subflow context */
430struct mptcp_subflow_context {
431 struct list_head node;/* conn_list of subflows */
432
433 struct_group(reset,
434
435 unsigned long avg_pacing_rate; /* protected by msk socket lock */
436 u64 local_key;
437 u64 remote_key;
438 u64 idsn;
439 u64 map_seq;
440 u32 snd_isn;
441 u32 token;
442 u32 rel_write_seq;
443 u32 map_subflow_seq;
444 u32 ssn_offset;
445 u32 map_data_len;
446 __wsum map_data_csum;
447 u32 map_csum_len;
448 u32 request_mptcp : 1, /* send MP_CAPABLE */
449 request_join : 1, /* send MP_JOIN */
450 request_bkup : 1,
451 mp_capable : 1, /* remote is MPTCP capable */
452 mp_join : 1, /* remote is JOINing */
453 fully_established : 1, /* path validated */
454 pm_notified : 1, /* PM hook called for established status */
455 conn_finished : 1,
456 map_valid : 1,
457 map_csum_reqd : 1,
458 map_data_fin : 1,
459 mpc_map : 1,
460 backup : 1,
461 send_mp_prio : 1,
462 send_mp_fail : 1,
463 send_fastclose : 1,
464 send_infinite_map : 1,
465 rx_eof : 1,
466 can_ack : 1, /* only after processing the remote a key */
467 disposable : 1, /* ctx can be free at ulp release time */
468 stale : 1, /* unable to snd/rcv data, do not use for xmit */
469 local_id_valid : 1, /* local_id is correctly initialized */
470 valid_csum_seen : 1; /* at least one csum validated */
471 enum mptcp_data_avail data_avail;
472 u32 remote_nonce;
473 u64 thmac;
474 u32 local_nonce;
475 u32 remote_token;
476 u8 hmac[MPTCPOPT_HMAC_LEN];
477 u8 local_id;
478 u8 remote_id;
479 u8 reset_seen:1;
480 u8 reset_transient:1;
481 u8 reset_reason:4;
482 u8 stale_count;
483
484 long delegated_status;
485 unsigned long fail_tout;
486
487 );
488
489 struct list_head delegated_node; /* link into delegated_action, protected by local BH */
490
491 u32 setsockopt_seq;
492 u32 stale_rcv_tstamp;
493
494 struct sock *tcp_sock; /* tcp sk backpointer */
495 struct sock *conn; /* parent mptcp_sock */
496 const struct inet_connection_sock_af_ops *icsk_af_ops;
497 void (*tcp_state_change)(struct sock *sk);
498 void (*tcp_error_report)(struct sock *sk);
499
500 struct rcu_head rcu;
501};
502
503static inline struct mptcp_subflow_context *
504mptcp_subflow_ctx(const struct sock *sk)
505{
506 struct inet_connection_sock *icsk = inet_csk(sk);
507
508 /* Use RCU on icsk_ulp_data only for sock diag code */
509 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data;
510}
511
512static inline struct sock *
513mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow)
514{
515 return subflow->tcp_sock;
516}
517
518static inline void
519mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow)
520{
521 memset(&subflow->reset, 0, sizeof(subflow->reset));
522 subflow->request_mptcp = 1;
523}
524
525static inline u64
526mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow)
527{
528 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq -
529 subflow->ssn_offset -
530 subflow->map_subflow_seq;
531}
532
533static inline u64
534mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow)
535{
536 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow);
537}
538
539void mptcp_subflow_process_delegated(struct sock *ssk);
540
541static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action)
542{
543 struct mptcp_delegated_action *delegated;
544 bool schedule;
545
546 /* the caller held the subflow bh socket lock */
547 lockdep_assert_in_softirq();
548
549 /* The implied barrier pairs with mptcp_subflow_delegated_done(), and
550 * ensures the below list check sees list updates done prior to status
551 * bit changes
552 */
553 if (!test_and_set_bit(action, &subflow->delegated_status)) {
554 /* still on delegated list from previous scheduling */
555 if (!list_empty(&subflow->delegated_node))
556 return;
557
558 delegated = this_cpu_ptr(&mptcp_delegated_actions);
559 schedule = list_empty(&delegated->head);
560 list_add_tail(&subflow->delegated_node, &delegated->head);
561 sock_hold(mptcp_subflow_tcp_sock(subflow));
562 if (schedule)
563 napi_schedule(&delegated->napi);
564 }
565}
566
567static inline struct mptcp_subflow_context *
568mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated)
569{
570 struct mptcp_subflow_context *ret;
571
572 if (list_empty(&delegated->head))
573 return NULL;
574
575 ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node);
576 list_del_init(&ret->delegated_node);
577 return ret;
578}
579
580static inline bool mptcp_subflow_has_delegated_action(const struct mptcp_subflow_context *subflow)
581{
582 return !!READ_ONCE(subflow->delegated_status);
583}
584
585static inline void mptcp_subflow_delegated_done(struct mptcp_subflow_context *subflow, int action)
586{
587 /* pairs with mptcp_subflow_delegate, ensures delegate_node is updated before
588 * touching the status bit
589 */
590 smp_wmb();
591 clear_bit(action, &subflow->delegated_status);
592}
593
594int mptcp_is_enabled(const struct net *net);
595unsigned int mptcp_get_add_addr_timeout(const struct net *net);
596int mptcp_is_checksum_enabled(const struct net *net);
597int mptcp_allow_join_id0(const struct net *net);
598unsigned int mptcp_stale_loss_cnt(const struct net *net);
599int mptcp_get_pm_type(const struct net *net);
600void mptcp_subflow_fully_established(struct mptcp_subflow_context *subflow,
601 struct mptcp_options_received *mp_opt);
602bool __mptcp_retransmit_pending_data(struct sock *sk);
603void mptcp_check_and_set_pending(struct sock *sk);
604void __mptcp_push_pending(struct sock *sk, unsigned int flags);
605bool mptcp_subflow_data_available(struct sock *sk);
606void __init mptcp_subflow_init(void);
607void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how);
608void mptcp_close_ssk(struct sock *sk, struct sock *ssk,
609 struct mptcp_subflow_context *subflow);
610void __mptcp_subflow_send_ack(struct sock *ssk);
611void mptcp_subflow_reset(struct sock *ssk);
612void mptcp_subflow_queue_clean(struct sock *ssk);
613void mptcp_sock_graft(struct sock *sk, struct socket *parent);
614struct socket *__mptcp_nmpc_socket(const struct mptcp_sock *msk);
615bool __mptcp_close(struct sock *sk, long timeout);
616void mptcp_cancel_work(struct sock *sk);
617
618bool mptcp_addresses_equal(const struct mptcp_addr_info *a,
619 const struct mptcp_addr_info *b, bool use_port);
620
621/* called with sk socket lock held */
622int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc,
623 const struct mptcp_addr_info *remote);
624int mptcp_subflow_create_socket(struct sock *sk, struct socket **new_sock);
625void mptcp_info2sockaddr(const struct mptcp_addr_info *info,
626 struct sockaddr_storage *addr,
627 unsigned short family);
628
629static inline bool __tcp_can_send(const struct sock *ssk)
630{
631 /* only send if our side has not closed yet */
632 return ((1 << inet_sk_state_load(ssk)) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT));
633}
634
635static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow)
636{
637 /* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */
638 if (subflow->request_join && !subflow->fully_established)
639 return false;
640
641 return __tcp_can_send(mptcp_subflow_tcp_sock(subflow));
642}
643
644void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow);
645
646bool mptcp_subflow_active(struct mptcp_subflow_context *subflow);
647
648static inline void mptcp_subflow_tcp_fallback(struct sock *sk,
649 struct mptcp_subflow_context *ctx)
650{
651 sk->sk_data_ready = sock_def_readable;
652 sk->sk_state_change = ctx->tcp_state_change;
653 sk->sk_write_space = sk_stream_write_space;
654 sk->sk_error_report = ctx->tcp_error_report;
655
656 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops;
657}
658
659void __init mptcp_proto_init(void);
660#if IS_ENABLED(CONFIG_MPTCP_IPV6)
661int __init mptcp_proto_v6_init(void);
662#endif
663
664struct sock *mptcp_sk_clone(const struct sock *sk,
665 const struct mptcp_options_received *mp_opt,
666 struct request_sock *req);
667void mptcp_get_options(const struct sk_buff *skb,
668 struct mptcp_options_received *mp_opt);
669
670void mptcp_finish_connect(struct sock *sk);
671void __mptcp_set_connected(struct sock *sk);
672void mptcp_reset_timeout(struct mptcp_sock *msk, unsigned long fail_tout);
673static inline bool mptcp_is_fully_established(struct sock *sk)
674{
675 return inet_sk_state_load(sk) == TCP_ESTABLISHED &&
676 READ_ONCE(mptcp_sk(sk)->fully_established);
677}
678void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk);
679void mptcp_data_ready(struct sock *sk, struct sock *ssk);
680bool mptcp_finish_join(struct sock *sk);
681bool mptcp_schedule_work(struct sock *sk);
682int mptcp_setsockopt(struct sock *sk, int level, int optname,
683 sockptr_t optval, unsigned int optlen);
684int mptcp_getsockopt(struct sock *sk, int level, int optname,
685 char __user *optval, int __user *option);
686
687u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq);
688static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit)
689{
690 if (use_64bit)
691 return cur_seq;
692
693 return __mptcp_expand_seq(old_seq, cur_seq);
694}
695void __mptcp_check_push(struct sock *sk, struct sock *ssk);
696void __mptcp_data_acked(struct sock *sk);
697void __mptcp_error_report(struct sock *sk);
698void mptcp_subflow_eof(struct sock *sk);
699bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit);
700static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk)
701{
702 return READ_ONCE(msk->snd_data_fin_enable) &&
703 READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt);
704}
705
706static inline bool mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
707{
708 if ((sk->sk_userlocks & SOCK_SNDBUF_LOCK) || ssk->sk_sndbuf <= READ_ONCE(sk->sk_sndbuf))
709 return false;
710
711 WRITE_ONCE(sk->sk_sndbuf, ssk->sk_sndbuf);
712 return true;
713}
714
715static inline void mptcp_write_space(struct sock *sk)
716{
717 if (sk_stream_is_writeable(sk)) {
718 /* pairs with memory barrier in mptcp_poll */
719 smp_mb();
720 if (test_and_clear_bit(MPTCP_NOSPACE, &mptcp_sk(sk)->flags))
721 sk_stream_write_space(sk);
722 }
723}
724
725void mptcp_destroy_common(struct mptcp_sock *msk, unsigned int flags);
726
727#define MPTCP_TOKEN_MAX_RETRIES 4
728
729void __init mptcp_token_init(void);
730static inline void mptcp_token_init_request(struct request_sock *req)
731{
732 mptcp_subflow_rsk(req)->token_node.pprev = NULL;
733}
734
735int mptcp_token_new_request(struct request_sock *req);
736void mptcp_token_destroy_request(struct request_sock *req);
737int mptcp_token_new_connect(struct sock *sk);
738void mptcp_token_accept(struct mptcp_subflow_request_sock *r,
739 struct mptcp_sock *msk);
740bool mptcp_token_exists(u32 token);
741struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token);
742struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot,
743 long *s_num);
744void mptcp_token_destroy(struct mptcp_sock *msk);
745
746void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn);
747
748void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac);
749__sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum);
750
751void __init mptcp_pm_init(void);
752void mptcp_pm_data_init(struct mptcp_sock *msk);
753void mptcp_pm_data_reset(struct mptcp_sock *msk);
754int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info,
755 struct mptcp_addr_info *addr);
756int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info,
757 bool require_family,
758 struct mptcp_pm_addr_entry *entry);
759void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
760void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
761void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side);
762void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk, gfp_t gfp);
763bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk);
764void mptcp_pm_connection_closed(struct mptcp_sock *msk);
765void mptcp_pm_subflow_established(struct mptcp_sock *msk);
766bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk);
767void mptcp_pm_subflow_check_next(struct mptcp_sock *msk, const struct sock *ssk,
768 const struct mptcp_subflow_context *subflow);
769void mptcp_pm_add_addr_received(const struct sock *ssk,
770 const struct mptcp_addr_info *addr);
771void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk,
772 const struct mptcp_addr_info *addr);
773void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk);
774void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk);
775void mptcp_pm_rm_addr_received(struct mptcp_sock *msk,
776 const struct mptcp_rm_list *rm_list);
777void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup);
778void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq);
779int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk,
780 struct mptcp_addr_info *addr,
781 struct mptcp_addr_info *rem,
782 u8 bkup);
783bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk,
784 const struct mptcp_pm_addr_entry *entry);
785void mptcp_pm_free_anno_list(struct mptcp_sock *msk);
786bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk);
787struct mptcp_pm_add_entry *
788mptcp_pm_del_add_timer(struct mptcp_sock *msk,
789 const struct mptcp_addr_info *addr, bool check_id);
790struct mptcp_pm_add_entry *
791mptcp_lookup_anno_list_by_saddr(const struct mptcp_sock *msk,
792 const struct mptcp_addr_info *addr);
793int mptcp_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk,
794 unsigned int id,
795 u8 *flags, int *ifindex);
796int mptcp_userspace_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk,
797 unsigned int id,
798 u8 *flags, int *ifindex);
799int mptcp_userspace_pm_set_flags(struct net *net, struct nlattr *token,
800 struct mptcp_pm_addr_entry *loc,
801 struct mptcp_pm_addr_entry *rem, u8 bkup);
802int mptcp_pm_announce_addr(struct mptcp_sock *msk,
803 const struct mptcp_addr_info *addr,
804 bool echo);
805int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
806int mptcp_pm_remove_subflow(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
807void mptcp_pm_remove_addrs_and_subflows(struct mptcp_sock *msk,
808 struct list_head *rm_list);
809
810int mptcp_userspace_pm_append_new_local_addr(struct mptcp_sock *msk,
811 struct mptcp_pm_addr_entry *entry);
812void mptcp_free_local_addr_list(struct mptcp_sock *msk);
813int mptcp_nl_cmd_announce(struct sk_buff *skb, struct genl_info *info);
814int mptcp_nl_cmd_remove(struct sk_buff *skb, struct genl_info *info);
815int mptcp_nl_cmd_sf_create(struct sk_buff *skb, struct genl_info *info);
816int mptcp_nl_cmd_sf_destroy(struct sk_buff *skb, struct genl_info *info);
817
818void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk,
819 const struct sock *ssk, gfp_t gfp);
820void mptcp_event_addr_announced(const struct sock *ssk, const struct mptcp_addr_info *info);
821void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id);
822bool mptcp_userspace_pm_active(const struct mptcp_sock *msk);
823
824static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk)
825{
826 return READ_ONCE(msk->pm.addr_signal) &
827 (BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO));
828}
829
830static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk)
831{
832 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL);
833}
834
835static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk)
836{
837 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO);
838}
839
840static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk)
841{
842 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL);
843}
844
845static inline bool mptcp_pm_is_userspace(const struct mptcp_sock *msk)
846{
847 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_USERSPACE;
848}
849
850static inline bool mptcp_pm_is_kernel(const struct mptcp_sock *msk)
851{
852 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_KERNEL;
853}
854
855static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port)
856{
857 u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE;
858
859 if (family == AF_INET6)
860 len = TCPOLEN_MPTCP_ADD_ADDR6_BASE;
861 if (!echo)
862 len += MPTCPOPT_THMAC_LEN;
863 /* account for 2 trailing 'nop' options */
864 if (port)
865 len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN;
866
867 return len;
868}
869
870static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list)
871{
872 if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX)
873 return -EINVAL;
874
875 return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1;
876}
877
878bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, const struct sk_buff *skb,
879 unsigned int opt_size, unsigned int remaining,
880 struct mptcp_addr_info *addr, bool *echo,
881 bool *drop_other_suboptions);
882bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining,
883 struct mptcp_rm_list *rm_list);
884int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
885int mptcp_userspace_pm_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
886
887void __init mptcp_pm_nl_init(void);
888void mptcp_pm_nl_work(struct mptcp_sock *msk);
889void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk,
890 const struct mptcp_rm_list *rm_list);
891int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
892unsigned int mptcp_pm_get_add_addr_signal_max(const struct mptcp_sock *msk);
893unsigned int mptcp_pm_get_add_addr_accept_max(const struct mptcp_sock *msk);
894unsigned int mptcp_pm_get_subflows_max(const struct mptcp_sock *msk);
895unsigned int mptcp_pm_get_local_addr_max(const struct mptcp_sock *msk);
896
897/* called under PM lock */
898static inline void __mptcp_pm_close_subflow(struct mptcp_sock *msk)
899{
900 if (--msk->pm.subflows < mptcp_pm_get_subflows_max(msk))
901 WRITE_ONCE(msk->pm.accept_subflow, true);
902}
903
904static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk)
905{
906 spin_lock_bh(&msk->pm.lock);
907 __mptcp_pm_close_subflow(msk);
908 spin_unlock_bh(&msk->pm.lock);
909}
910
911void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk);
912void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk);
913
914static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb)
915{
916 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP);
917}
918
919void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops);
920
921static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk)
922{
923 return test_bit(MPTCP_FALLBACK_DONE, &msk->flags);
924}
925
926static inline bool mptcp_check_fallback(const struct sock *sk)
927{
928 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
929 struct mptcp_sock *msk = mptcp_sk(subflow->conn);
930
931 return __mptcp_check_fallback(msk);
932}
933
934static inline void __mptcp_do_fallback(struct mptcp_sock *msk)
935{
936 if (test_bit(MPTCP_FALLBACK_DONE, &msk->flags)) {
937 pr_debug("TCP fallback already done (msk=%p)", msk);
938 return;
939 }
940 set_bit(MPTCP_FALLBACK_DONE, &msk->flags);
941}
942
943static inline void mptcp_do_fallback(struct sock *ssk)
944{
945 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
946 struct sock *sk = subflow->conn;
947 struct mptcp_sock *msk;
948
949 msk = mptcp_sk(sk);
950 __mptcp_do_fallback(msk);
951 if (READ_ONCE(msk->snd_data_fin_enable) && !(ssk->sk_shutdown & SEND_SHUTDOWN)) {
952 gfp_t saved_allocation = ssk->sk_allocation;
953
954 /* we are in a atomic (BH) scope, override ssk default for data
955 * fin allocation
956 */
957 ssk->sk_allocation = GFP_ATOMIC;
958 ssk->sk_shutdown |= SEND_SHUTDOWN;
959 tcp_shutdown(ssk, SEND_SHUTDOWN);
960 ssk->sk_allocation = saved_allocation;
961 }
962}
963
964#define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a)
965
966static inline bool mptcp_check_infinite_map(struct sk_buff *skb)
967{
968 struct mptcp_ext *mpext;
969
970 mpext = skb ? mptcp_get_ext(skb) : NULL;
971 if (mpext && mpext->infinite_map)
972 return true;
973
974 return false;
975}
976
977static inline bool is_active_ssk(struct mptcp_subflow_context *subflow)
978{
979 return (subflow->request_mptcp || subflow->request_join);
980}
981
982static inline bool subflow_simultaneous_connect(struct sock *sk)
983{
984 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
985
986 return sk->sk_state == TCP_ESTABLISHED &&
987 is_active_ssk(subflow) &&
988 !subflow->conn_finished;
989}
990
991#ifdef CONFIG_SYN_COOKIES
992void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
993 struct sk_buff *skb);
994bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
995 struct sk_buff *skb);
996void __init mptcp_join_cookie_init(void);
997#else
998static inline void
999subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1000 struct sk_buff *skb) {}
1001static inline bool
1002mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1003 struct sk_buff *skb)
1004{
1005 return false;
1006}
1007
1008static inline void mptcp_join_cookie_init(void) {}
1009#endif
1010
1011#endif /* __MPTCP_PROTOCOL_H */