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
before64(__u64 seq1,__u64 seq2)129 static 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
136 struct 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
mptcp_option(u8 subopt,u8 len,u8 nib,u8 field)168 static 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
174 enum 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
186 enum 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
197 enum 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
206 struct 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
231 struct 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
239 struct 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 */
250 struct 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 fastopening:1;
290 int connect_flags;
291 struct work_struct work;
292 struct sk_buff *ooo_last_skb;
293 struct rb_root out_of_order_queue;
294 struct sk_buff_head receive_queue;
295 struct list_head conn_list;
296 struct list_head rtx_queue;
297 struct mptcp_data_frag *first_pending;
298 struct list_head join_list;
299 struct socket *subflow; /* outgoing connect/listener/!mp_capable */
300 struct sock *first;
301 struct mptcp_pm_data pm;
302 struct {
303 u32 space; /* bytes copied in last measurement window */
304 u32 copied; /* bytes copied in this measurement window */
305 u64 time; /* start time of measurement window */
306 u64 rtt_us; /* last maximum rtt of subflows */
307 } rcvq_space;
308
309 u32 setsockopt_seq;
310 char ca_name[TCP_CA_NAME_MAX];
311 struct mptcp_sock *dl_next;
312 };
313
314 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock)
315 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock)
316
317 #define mptcp_for_each_subflow(__msk, __subflow) \
318 list_for_each_entry(__subflow, &((__msk)->conn_list), node)
319 #define mptcp_for_each_subflow_safe(__msk, __subflow, __tmp) \
320 list_for_each_entry_safe(__subflow, __tmp, &((__msk)->conn_list), node)
321
msk_owned_by_me(const struct mptcp_sock * msk)322 static inline void msk_owned_by_me(const struct mptcp_sock *msk)
323 {
324 sock_owned_by_me((const struct sock *)msk);
325 }
326
mptcp_sk(const struct sock * sk)327 static inline struct mptcp_sock *mptcp_sk(const struct sock *sk)
328 {
329 return (struct mptcp_sock *)sk;
330 }
331
332 /* the msk socket don't use the backlog, also account for the bulk
333 * free memory
334 */
__mptcp_rmem(const struct sock * sk)335 static inline int __mptcp_rmem(const struct sock *sk)
336 {
337 return atomic_read(&sk->sk_rmem_alloc) - READ_ONCE(mptcp_sk(sk)->rmem_released);
338 }
339
__mptcp_space(const struct sock * sk)340 static inline int __mptcp_space(const struct sock *sk)
341 {
342 return tcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - __mptcp_rmem(sk));
343 }
344
mptcp_send_head(const struct sock * sk)345 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk)
346 {
347 const struct mptcp_sock *msk = mptcp_sk(sk);
348
349 return READ_ONCE(msk->first_pending);
350 }
351
mptcp_send_next(struct sock * sk)352 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk)
353 {
354 struct mptcp_sock *msk = mptcp_sk(sk);
355 struct mptcp_data_frag *cur;
356
357 cur = msk->first_pending;
358 return list_is_last(&cur->list, &msk->rtx_queue) ? NULL :
359 list_next_entry(cur, list);
360 }
361
mptcp_pending_tail(const struct sock * sk)362 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk)
363 {
364 struct mptcp_sock *msk = mptcp_sk(sk);
365
366 if (!msk->first_pending)
367 return NULL;
368
369 if (WARN_ON_ONCE(list_empty(&msk->rtx_queue)))
370 return NULL;
371
372 return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list);
373 }
374
mptcp_rtx_head(const struct sock * sk)375 static inline struct mptcp_data_frag *mptcp_rtx_head(const struct sock *sk)
376 {
377 struct mptcp_sock *msk = mptcp_sk(sk);
378
379 if (msk->snd_una == READ_ONCE(msk->snd_nxt))
380 return NULL;
381
382 return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list);
383 }
384
385 struct csum_pseudo_header {
386 __be64 data_seq;
387 __be32 subflow_seq;
388 __be16 data_len;
389 __sum16 csum;
390 };
391
392 struct mptcp_subflow_request_sock {
393 struct tcp_request_sock sk;
394 u16 mp_capable : 1,
395 mp_join : 1,
396 backup : 1,
397 csum_reqd : 1,
398 allow_join_id0 : 1;
399 u8 local_id;
400 u8 remote_id;
401 u64 local_key;
402 u64 idsn;
403 u32 token;
404 u32 ssn_offset;
405 u64 thmac;
406 u32 local_nonce;
407 u32 remote_nonce;
408 struct mptcp_sock *msk;
409 struct hlist_nulls_node token_node;
410 };
411
412 static inline struct mptcp_subflow_request_sock *
mptcp_subflow_rsk(const struct request_sock * rsk)413 mptcp_subflow_rsk(const struct request_sock *rsk)
414 {
415 return (struct mptcp_subflow_request_sock *)rsk;
416 }
417
418 enum mptcp_data_avail {
419 MPTCP_SUBFLOW_NODATA,
420 MPTCP_SUBFLOW_DATA_AVAIL,
421 };
422
423 struct mptcp_delegated_action {
424 struct napi_struct napi;
425 struct list_head head;
426 };
427
428 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions);
429
430 #define MPTCP_DELEGATE_SEND 0
431 #define MPTCP_DELEGATE_ACK 1
432
433 /* MPTCP subflow context */
434 struct mptcp_subflow_context {
435 struct list_head node;/* conn_list of subflows */
436
437 struct_group(reset,
438
439 unsigned long avg_pacing_rate; /* protected by msk socket lock */
440 u64 local_key;
441 u64 remote_key;
442 u64 idsn;
443 u64 map_seq;
444 u32 snd_isn;
445 u32 token;
446 u32 rel_write_seq;
447 u32 map_subflow_seq;
448 u32 ssn_offset;
449 u32 map_data_len;
450 __wsum map_data_csum;
451 u32 map_csum_len;
452 u32 request_mptcp : 1, /* send MP_CAPABLE */
453 request_join : 1, /* send MP_JOIN */
454 request_bkup : 1,
455 mp_capable : 1, /* remote is MPTCP capable */
456 mp_join : 1, /* remote is JOINing */
457 fully_established : 1, /* path validated */
458 pm_notified : 1, /* PM hook called for established status */
459 conn_finished : 1,
460 map_valid : 1,
461 map_csum_reqd : 1,
462 map_data_fin : 1,
463 mpc_map : 1,
464 backup : 1,
465 send_mp_prio : 1,
466 send_mp_fail : 1,
467 send_fastclose : 1,
468 send_infinite_map : 1,
469 rx_eof : 1,
470 can_ack : 1, /* only after processing the remote a key */
471 disposable : 1, /* ctx can be free at ulp release time */
472 stale : 1, /* unable to snd/rcv data, do not use for xmit */
473 local_id_valid : 1, /* local_id is correctly initialized */
474 valid_csum_seen : 1; /* at least one csum validated */
475 enum mptcp_data_avail data_avail;
476 u32 remote_nonce;
477 u64 thmac;
478 u32 local_nonce;
479 u32 remote_token;
480 u8 hmac[MPTCPOPT_HMAC_LEN];
481 u8 local_id;
482 u8 remote_id;
483 u8 reset_seen:1;
484 u8 reset_transient:1;
485 u8 reset_reason:4;
486 u8 stale_count;
487
488 long delegated_status;
489 unsigned long fail_tout;
490
491 );
492
493 struct list_head delegated_node; /* link into delegated_action, protected by local BH */
494
495 u32 setsockopt_seq;
496 u32 stale_rcv_tstamp;
497
498 struct sock *tcp_sock; /* tcp sk backpointer */
499 struct sock *conn; /* parent mptcp_sock */
500 const struct inet_connection_sock_af_ops *icsk_af_ops;
501 void (*tcp_state_change)(struct sock *sk);
502 void (*tcp_error_report)(struct sock *sk);
503
504 struct rcu_head rcu;
505 };
506
507 static inline struct mptcp_subflow_context *
mptcp_subflow_ctx(const struct sock * sk)508 mptcp_subflow_ctx(const struct sock *sk)
509 {
510 struct inet_connection_sock *icsk = inet_csk(sk);
511
512 /* Use RCU on icsk_ulp_data only for sock diag code */
513 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data;
514 }
515
516 static inline struct sock *
mptcp_subflow_tcp_sock(const struct mptcp_subflow_context * subflow)517 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow)
518 {
519 return subflow->tcp_sock;
520 }
521
522 static inline void
mptcp_subflow_ctx_reset(struct mptcp_subflow_context * subflow)523 mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow)
524 {
525 memset(&subflow->reset, 0, sizeof(subflow->reset));
526 subflow->request_mptcp = 1;
527 }
528
529 static inline u64
mptcp_subflow_get_map_offset(const struct mptcp_subflow_context * subflow)530 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow)
531 {
532 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq -
533 subflow->ssn_offset -
534 subflow->map_subflow_seq;
535 }
536
537 static inline u64
mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context * subflow)538 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow)
539 {
540 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow);
541 }
542
543 void mptcp_subflow_process_delegated(struct sock *ssk);
544
mptcp_subflow_delegate(struct mptcp_subflow_context * subflow,int action)545 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action)
546 {
547 struct mptcp_delegated_action *delegated;
548 bool schedule;
549
550 /* the caller held the subflow bh socket lock */
551 lockdep_assert_in_softirq();
552
553 /* The implied barrier pairs with mptcp_subflow_delegated_done(), and
554 * ensures the below list check sees list updates done prior to status
555 * bit changes
556 */
557 if (!test_and_set_bit(action, &subflow->delegated_status)) {
558 /* still on delegated list from previous scheduling */
559 if (!list_empty(&subflow->delegated_node))
560 return;
561
562 delegated = this_cpu_ptr(&mptcp_delegated_actions);
563 schedule = list_empty(&delegated->head);
564 list_add_tail(&subflow->delegated_node, &delegated->head);
565 sock_hold(mptcp_subflow_tcp_sock(subflow));
566 if (schedule)
567 napi_schedule(&delegated->napi);
568 }
569 }
570
571 static inline struct mptcp_subflow_context *
mptcp_subflow_delegated_next(struct mptcp_delegated_action * delegated)572 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated)
573 {
574 struct mptcp_subflow_context *ret;
575
576 if (list_empty(&delegated->head))
577 return NULL;
578
579 ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node);
580 list_del_init(&ret->delegated_node);
581 return ret;
582 }
583
mptcp_subflow_has_delegated_action(const struct mptcp_subflow_context * subflow)584 static inline bool mptcp_subflow_has_delegated_action(const struct mptcp_subflow_context *subflow)
585 {
586 return !!READ_ONCE(subflow->delegated_status);
587 }
588
mptcp_subflow_delegated_done(struct mptcp_subflow_context * subflow,int action)589 static inline void mptcp_subflow_delegated_done(struct mptcp_subflow_context *subflow, int action)
590 {
591 /* pairs with mptcp_subflow_delegate, ensures delegate_node is updated before
592 * touching the status bit
593 */
594 smp_wmb();
595 clear_bit(action, &subflow->delegated_status);
596 }
597
598 int mptcp_is_enabled(const struct net *net);
599 unsigned int mptcp_get_add_addr_timeout(const struct net *net);
600 int mptcp_is_checksum_enabled(const struct net *net);
601 int mptcp_allow_join_id0(const struct net *net);
602 unsigned int mptcp_stale_loss_cnt(const struct net *net);
603 int mptcp_get_pm_type(const struct net *net);
604 void mptcp_copy_inaddrs(struct sock *msk, const struct sock *ssk);
605 void mptcp_subflow_fully_established(struct mptcp_subflow_context *subflow,
606 struct mptcp_options_received *mp_opt);
607 bool __mptcp_retransmit_pending_data(struct sock *sk);
608 void mptcp_check_and_set_pending(struct sock *sk);
609 void __mptcp_push_pending(struct sock *sk, unsigned int flags);
610 bool mptcp_subflow_data_available(struct sock *sk);
611 void __init mptcp_subflow_init(void);
612 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how);
613 void mptcp_close_ssk(struct sock *sk, struct sock *ssk,
614 struct mptcp_subflow_context *subflow);
615 void __mptcp_subflow_send_ack(struct sock *ssk);
616 void mptcp_subflow_reset(struct sock *ssk);
617 void mptcp_subflow_queue_clean(struct sock *sk, struct sock *ssk);
618 void mptcp_sock_graft(struct sock *sk, struct socket *parent);
619 struct socket *__mptcp_nmpc_socket(const struct mptcp_sock *msk);
620 bool __mptcp_close(struct sock *sk, long timeout);
621 void mptcp_cancel_work(struct sock *sk);
622
623 bool mptcp_addresses_equal(const struct mptcp_addr_info *a,
624 const struct mptcp_addr_info *b, bool use_port);
625
626 /* called with sk socket lock held */
627 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc,
628 const struct mptcp_addr_info *remote);
629 int mptcp_subflow_create_socket(struct sock *sk, unsigned short family,
630 struct socket **new_sock);
631 void mptcp_info2sockaddr(const struct mptcp_addr_info *info,
632 struct sockaddr_storage *addr,
633 unsigned short family);
634
__tcp_can_send(const struct sock * ssk)635 static inline bool __tcp_can_send(const struct sock *ssk)
636 {
637 /* only send if our side has not closed yet */
638 return ((1 << inet_sk_state_load(ssk)) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT));
639 }
640
__mptcp_subflow_active(struct mptcp_subflow_context * subflow)641 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow)
642 {
643 /* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */
644 if (subflow->request_join && !subflow->fully_established)
645 return false;
646
647 return __tcp_can_send(mptcp_subflow_tcp_sock(subflow));
648 }
649
650 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow);
651
652 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow);
653
mptcp_subflow_tcp_fallback(struct sock * sk,struct mptcp_subflow_context * ctx)654 static inline void mptcp_subflow_tcp_fallback(struct sock *sk,
655 struct mptcp_subflow_context *ctx)
656 {
657 sk->sk_data_ready = sock_def_readable;
658 sk->sk_state_change = ctx->tcp_state_change;
659 sk->sk_write_space = sk_stream_write_space;
660 sk->sk_error_report = ctx->tcp_error_report;
661
662 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops;
663 }
664
665 void __init mptcp_proto_init(void);
666 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
667 int __init mptcp_proto_v6_init(void);
668 #endif
669
670 struct sock *mptcp_sk_clone(const struct sock *sk,
671 const struct mptcp_options_received *mp_opt,
672 struct request_sock *req);
673 void mptcp_get_options(const struct sk_buff *skb,
674 struct mptcp_options_received *mp_opt);
675
676 void mptcp_finish_connect(struct sock *sk);
677 void __mptcp_set_connected(struct sock *sk);
678 void mptcp_reset_timeout(struct mptcp_sock *msk, unsigned long fail_tout);
mptcp_is_fully_established(struct sock * sk)679 static inline bool mptcp_is_fully_established(struct sock *sk)
680 {
681 return inet_sk_state_load(sk) == TCP_ESTABLISHED &&
682 READ_ONCE(mptcp_sk(sk)->fully_established);
683 }
684 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk);
685 void mptcp_data_ready(struct sock *sk, struct sock *ssk);
686 bool mptcp_finish_join(struct sock *sk);
687 bool mptcp_schedule_work(struct sock *sk);
688 int mptcp_setsockopt(struct sock *sk, int level, int optname,
689 sockptr_t optval, unsigned int optlen);
690 int mptcp_getsockopt(struct sock *sk, int level, int optname,
691 char __user *optval, int __user *option);
692
693 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq);
mptcp_expand_seq(u64 old_seq,u64 cur_seq,bool use_64bit)694 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit)
695 {
696 if (use_64bit)
697 return cur_seq;
698
699 return __mptcp_expand_seq(old_seq, cur_seq);
700 }
701 void __mptcp_check_push(struct sock *sk, struct sock *ssk);
702 void __mptcp_data_acked(struct sock *sk);
703 void __mptcp_error_report(struct sock *sk);
704 void mptcp_subflow_eof(struct sock *sk);
705 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit);
mptcp_data_fin_enabled(const struct mptcp_sock * msk)706 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk)
707 {
708 return READ_ONCE(msk->snd_data_fin_enable) &&
709 READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt);
710 }
711
mptcp_propagate_sndbuf(struct sock * sk,struct sock * ssk)712 static inline bool mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
713 {
714 if ((sk->sk_userlocks & SOCK_SNDBUF_LOCK) || ssk->sk_sndbuf <= READ_ONCE(sk->sk_sndbuf))
715 return false;
716
717 WRITE_ONCE(sk->sk_sndbuf, ssk->sk_sndbuf);
718 return true;
719 }
720
mptcp_write_space(struct sock * sk)721 static inline void mptcp_write_space(struct sock *sk)
722 {
723 if (sk_stream_is_writeable(sk)) {
724 /* pairs with memory barrier in mptcp_poll */
725 smp_mb();
726 if (test_and_clear_bit(MPTCP_NOSPACE, &mptcp_sk(sk)->flags))
727 sk_stream_write_space(sk);
728 }
729 }
730
731 void mptcp_destroy_common(struct mptcp_sock *msk, unsigned int flags);
732
733 #define MPTCP_TOKEN_MAX_RETRIES 4
734
735 void __init mptcp_token_init(void);
mptcp_token_init_request(struct request_sock * req)736 static inline void mptcp_token_init_request(struct request_sock *req)
737 {
738 mptcp_subflow_rsk(req)->token_node.pprev = NULL;
739 }
740
741 int mptcp_token_new_request(struct request_sock *req);
742 void mptcp_token_destroy_request(struct request_sock *req);
743 int mptcp_token_new_connect(struct sock *sk);
744 void mptcp_token_accept(struct mptcp_subflow_request_sock *r,
745 struct mptcp_sock *msk);
746 bool mptcp_token_exists(u32 token);
747 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token);
748 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot,
749 long *s_num);
750 void mptcp_token_destroy(struct mptcp_sock *msk);
751
752 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn);
753
754 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac);
755 __sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum);
756
757 void __init mptcp_pm_init(void);
758 void mptcp_pm_data_init(struct mptcp_sock *msk);
759 void mptcp_pm_data_reset(struct mptcp_sock *msk);
760 int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info,
761 struct mptcp_addr_info *addr);
762 int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info,
763 bool require_family,
764 struct mptcp_pm_addr_entry *entry);
765 bool mptcp_pm_addr_families_match(const struct sock *sk,
766 const struct mptcp_addr_info *loc,
767 const struct mptcp_addr_info *rem);
768 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
769 void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
770 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side);
771 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk, gfp_t gfp);
772 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk);
773 void mptcp_pm_connection_closed(struct mptcp_sock *msk);
774 void mptcp_pm_subflow_established(struct mptcp_sock *msk);
775 bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk);
776 void mptcp_pm_subflow_check_next(struct mptcp_sock *msk, const struct sock *ssk,
777 const struct mptcp_subflow_context *subflow);
778 void mptcp_pm_add_addr_received(const struct sock *ssk,
779 const struct mptcp_addr_info *addr);
780 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk,
781 const struct mptcp_addr_info *addr);
782 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk);
783 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk);
784 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk,
785 const struct mptcp_rm_list *rm_list);
786 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup);
787 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq);
788 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk,
789 struct mptcp_addr_info *addr,
790 struct mptcp_addr_info *rem,
791 u8 bkup);
792 bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk,
793 const struct mptcp_pm_addr_entry *entry);
794 void mptcp_pm_free_anno_list(struct mptcp_sock *msk);
795 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk);
796 struct mptcp_pm_add_entry *
797 mptcp_pm_del_add_timer(struct mptcp_sock *msk,
798 const struct mptcp_addr_info *addr, bool check_id);
799 struct mptcp_pm_add_entry *
800 mptcp_lookup_anno_list_by_saddr(const struct mptcp_sock *msk,
801 const struct mptcp_addr_info *addr);
802 int mptcp_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk,
803 unsigned int id,
804 u8 *flags, int *ifindex);
805 int mptcp_userspace_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk,
806 unsigned int id,
807 u8 *flags, int *ifindex);
808 int mptcp_userspace_pm_set_flags(struct net *net, struct nlattr *token,
809 struct mptcp_pm_addr_entry *loc,
810 struct mptcp_pm_addr_entry *rem, u8 bkup);
811 int mptcp_pm_announce_addr(struct mptcp_sock *msk,
812 const struct mptcp_addr_info *addr,
813 bool echo);
814 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
815 int mptcp_pm_remove_subflow(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
816 void mptcp_pm_remove_addrs_and_subflows(struct mptcp_sock *msk,
817 struct list_head *rm_list);
818
819 int mptcp_userspace_pm_append_new_local_addr(struct mptcp_sock *msk,
820 struct mptcp_pm_addr_entry *entry);
821 void mptcp_free_local_addr_list(struct mptcp_sock *msk);
822 int mptcp_nl_cmd_announce(struct sk_buff *skb, struct genl_info *info);
823 int mptcp_nl_cmd_remove(struct sk_buff *skb, struct genl_info *info);
824 int mptcp_nl_cmd_sf_create(struct sk_buff *skb, struct genl_info *info);
825 int mptcp_nl_cmd_sf_destroy(struct sk_buff *skb, struct genl_info *info);
826
827 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk,
828 const struct sock *ssk, gfp_t gfp);
829 void mptcp_event_addr_announced(const struct sock *ssk, const struct mptcp_addr_info *info);
830 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id);
831 bool mptcp_userspace_pm_active(const struct mptcp_sock *msk);
832
mptcp_pm_should_add_signal(struct mptcp_sock * msk)833 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk)
834 {
835 return READ_ONCE(msk->pm.addr_signal) &
836 (BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO));
837 }
838
mptcp_pm_should_add_signal_addr(struct mptcp_sock * msk)839 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk)
840 {
841 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL);
842 }
843
mptcp_pm_should_add_signal_echo(struct mptcp_sock * msk)844 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk)
845 {
846 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO);
847 }
848
mptcp_pm_should_rm_signal(struct mptcp_sock * msk)849 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk)
850 {
851 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL);
852 }
853
mptcp_pm_is_userspace(const struct mptcp_sock * msk)854 static inline bool mptcp_pm_is_userspace(const struct mptcp_sock *msk)
855 {
856 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_USERSPACE;
857 }
858
mptcp_pm_is_kernel(const struct mptcp_sock * msk)859 static inline bool mptcp_pm_is_kernel(const struct mptcp_sock *msk)
860 {
861 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_KERNEL;
862 }
863
mptcp_add_addr_len(int family,bool echo,bool port)864 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port)
865 {
866 u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE;
867
868 if (family == AF_INET6)
869 len = TCPOLEN_MPTCP_ADD_ADDR6_BASE;
870 if (!echo)
871 len += MPTCPOPT_THMAC_LEN;
872 /* account for 2 trailing 'nop' options */
873 if (port)
874 len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN;
875
876 return len;
877 }
878
mptcp_rm_addr_len(const struct mptcp_rm_list * rm_list)879 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list)
880 {
881 if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX)
882 return -EINVAL;
883
884 return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1;
885 }
886
887 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, const struct sk_buff *skb,
888 unsigned int opt_size, unsigned int remaining,
889 struct mptcp_addr_info *addr, bool *echo,
890 bool *drop_other_suboptions);
891 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining,
892 struct mptcp_rm_list *rm_list);
893 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
894 int mptcp_userspace_pm_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
895
896 void __init mptcp_pm_nl_init(void);
897 void mptcp_pm_nl_work(struct mptcp_sock *msk);
898 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk,
899 const struct mptcp_rm_list *rm_list);
900 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
901 unsigned int mptcp_pm_get_add_addr_signal_max(const struct mptcp_sock *msk);
902 unsigned int mptcp_pm_get_add_addr_accept_max(const struct mptcp_sock *msk);
903 unsigned int mptcp_pm_get_subflows_max(const struct mptcp_sock *msk);
904 unsigned int mptcp_pm_get_local_addr_max(const struct mptcp_sock *msk);
905
906 /* called under PM lock */
__mptcp_pm_close_subflow(struct mptcp_sock * msk)907 static inline void __mptcp_pm_close_subflow(struct mptcp_sock *msk)
908 {
909 if (--msk->pm.subflows < mptcp_pm_get_subflows_max(msk))
910 WRITE_ONCE(msk->pm.accept_subflow, true);
911 }
912
mptcp_pm_close_subflow(struct mptcp_sock * msk)913 static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk)
914 {
915 spin_lock_bh(&msk->pm.lock);
916 __mptcp_pm_close_subflow(msk);
917 spin_unlock_bh(&msk->pm.lock);
918 }
919
920 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk);
921 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk);
922
mptcp_get_ext(const struct sk_buff * skb)923 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb)
924 {
925 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP);
926 }
927
928 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops);
929
__mptcp_check_fallback(const struct mptcp_sock * msk)930 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk)
931 {
932 return test_bit(MPTCP_FALLBACK_DONE, &msk->flags);
933 }
934
mptcp_check_fallback(const struct sock * sk)935 static inline bool mptcp_check_fallback(const struct sock *sk)
936 {
937 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
938 struct mptcp_sock *msk = mptcp_sk(subflow->conn);
939
940 return __mptcp_check_fallback(msk);
941 }
942
__mptcp_do_fallback(struct mptcp_sock * msk)943 static inline void __mptcp_do_fallback(struct mptcp_sock *msk)
944 {
945 if (test_bit(MPTCP_FALLBACK_DONE, &msk->flags)) {
946 pr_debug("TCP fallback already done (msk=%p)", msk);
947 return;
948 }
949 set_bit(MPTCP_FALLBACK_DONE, &msk->flags);
950 }
951
mptcp_do_fallback(struct sock * ssk)952 static inline void mptcp_do_fallback(struct sock *ssk)
953 {
954 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
955 struct sock *sk = subflow->conn;
956 struct mptcp_sock *msk;
957
958 msk = mptcp_sk(sk);
959 __mptcp_do_fallback(msk);
960 if (READ_ONCE(msk->snd_data_fin_enable) && !(ssk->sk_shutdown & SEND_SHUTDOWN)) {
961 gfp_t saved_allocation = ssk->sk_allocation;
962
963 /* we are in a atomic (BH) scope, override ssk default for data
964 * fin allocation
965 */
966 ssk->sk_allocation = GFP_ATOMIC;
967 ssk->sk_shutdown |= SEND_SHUTDOWN;
968 tcp_shutdown(ssk, SEND_SHUTDOWN);
969 ssk->sk_allocation = saved_allocation;
970 }
971 }
972
973 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a)
974
mptcp_check_infinite_map(struct sk_buff * skb)975 static inline bool mptcp_check_infinite_map(struct sk_buff *skb)
976 {
977 struct mptcp_ext *mpext;
978
979 mpext = skb ? mptcp_get_ext(skb) : NULL;
980 if (mpext && mpext->infinite_map)
981 return true;
982
983 return false;
984 }
985
is_active_ssk(struct mptcp_subflow_context * subflow)986 static inline bool is_active_ssk(struct mptcp_subflow_context *subflow)
987 {
988 return (subflow->request_mptcp || subflow->request_join);
989 }
990
subflow_simultaneous_connect(struct sock * sk)991 static inline bool subflow_simultaneous_connect(struct sock *sk)
992 {
993 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
994
995 return sk->sk_state == TCP_ESTABLISHED &&
996 is_active_ssk(subflow) &&
997 !subflow->conn_finished;
998 }
999
1000 #ifdef CONFIG_SYN_COOKIES
1001 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1002 struct sk_buff *skb);
1003 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1004 struct sk_buff *skb);
1005 void __init mptcp_join_cookie_init(void);
1006 #else
1007 static inline void
subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock * subflow_req,struct sk_buff * skb)1008 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1009 struct sk_buff *skb) {}
1010 static inline bool
mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock * subflow_req,struct sk_buff * skb)1011 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1012 struct sk_buff *skb)
1013 {
1014 return false;
1015 }
1016
mptcp_join_cookie_init(void)1017 static inline void mptcp_join_cookie_init(void) {}
1018 #endif
1019
1020 #endif /* __MPTCP_PROTOCOL_H */
1021