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
14#define MPTCP_SUPPORTED_VERSION 1
15
16/* MPTCP option bits */
17#define OPTION_MPTCP_MPC_SYN BIT(0)
18#define OPTION_MPTCP_MPC_SYNACK BIT(1)
19#define OPTION_MPTCP_MPC_ACK BIT(2)
20#define OPTION_MPTCP_MPJ_SYN BIT(3)
21#define OPTION_MPTCP_MPJ_SYNACK BIT(4)
22#define OPTION_MPTCP_MPJ_ACK BIT(5)
23#define OPTION_MPTCP_ADD_ADDR BIT(6)
24#define OPTION_MPTCP_ADD_ADDR6 BIT(7)
25#define OPTION_MPTCP_RM_ADDR BIT(8)
26
27/* MPTCP option subtypes */
28#define MPTCPOPT_MP_CAPABLE 0
29#define MPTCPOPT_MP_JOIN 1
30#define MPTCPOPT_DSS 2
31#define MPTCPOPT_ADD_ADDR 3
32#define MPTCPOPT_RM_ADDR 4
33#define MPTCPOPT_MP_PRIO 5
34#define MPTCPOPT_MP_FAIL 6
35#define MPTCPOPT_MP_FASTCLOSE 7
36
37/* MPTCP suboption lengths */
38#define TCPOLEN_MPTCP_MPC_SYN 4
39#define TCPOLEN_MPTCP_MPC_SYNACK 12
40#define TCPOLEN_MPTCP_MPC_ACK 20
41#define TCPOLEN_MPTCP_MPC_ACK_DATA 22
42#define TCPOLEN_MPTCP_MPJ_SYN 12
43#define TCPOLEN_MPTCP_MPJ_SYNACK 16
44#define TCPOLEN_MPTCP_MPJ_ACK 24
45#define TCPOLEN_MPTCP_DSS_BASE 4
46#define TCPOLEN_MPTCP_DSS_ACK32 4
47#define TCPOLEN_MPTCP_DSS_ACK64 8
48#define TCPOLEN_MPTCP_DSS_MAP32 10
49#define TCPOLEN_MPTCP_DSS_MAP64 14
50#define TCPOLEN_MPTCP_DSS_CHECKSUM 2
51#define TCPOLEN_MPTCP_ADD_ADDR 16
52#define TCPOLEN_MPTCP_ADD_ADDR_PORT 18
53#define TCPOLEN_MPTCP_ADD_ADDR_BASE 8
54#define TCPOLEN_MPTCP_ADD_ADDR_BASE_PORT 10
55#define TCPOLEN_MPTCP_ADD_ADDR6 28
56#define TCPOLEN_MPTCP_ADD_ADDR6_PORT 30
57#define TCPOLEN_MPTCP_ADD_ADDR6_BASE 20
58#define TCPOLEN_MPTCP_ADD_ADDR6_BASE_PORT 22
59#define TCPOLEN_MPTCP_PORT_LEN 2
60#define TCPOLEN_MPTCP_RM_ADDR_BASE 4
61
62/* MPTCP MP_JOIN flags */
63#define MPTCPOPT_BACKUP BIT(0)
64#define MPTCPOPT_HMAC_LEN 20
65#define MPTCPOPT_THMAC_LEN 8
66
67/* MPTCP MP_CAPABLE flags */
68#define MPTCP_VERSION_MASK (0x0F)
69#define MPTCP_CAP_CHECKSUM_REQD BIT(7)
70#define MPTCP_CAP_EXTENSIBILITY BIT(6)
71#define MPTCP_CAP_HMAC_SHA256 BIT(0)
72#define MPTCP_CAP_FLAG_MASK (0x3F)
73
74/* MPTCP DSS flags */
75#define MPTCP_DSS_DATA_FIN BIT(4)
76#define MPTCP_DSS_DSN64 BIT(3)
77#define MPTCP_DSS_HAS_MAP BIT(2)
78#define MPTCP_DSS_ACK64 BIT(1)
79#define MPTCP_DSS_HAS_ACK BIT(0)
80#define MPTCP_DSS_FLAG_MASK (0x1F)
81
82/* MPTCP ADD_ADDR flags */
83#define MPTCP_ADDR_ECHO BIT(0)
84#define MPTCP_ADDR_IPVERSION_4 4
85#define MPTCP_ADDR_IPVERSION_6 6
86
87/* MPTCP socket flags */
88#define MPTCP_DATA_READY 0
89#define MPTCP_SEND_SPACE 1
90#define MPTCP_WORK_RTX 2
91#define MPTCP_WORK_EOF 3
92
93struct mptcp_options_received {
94 u64 sndr_key;
95 u64 rcvr_key;
96 u64 data_ack;
97 u64 data_seq;
98 u32 subflow_seq;
99 u16 data_len;
100 u16 mp_capable : 1,
101 mp_join : 1,
102 dss : 1,
103 add_addr : 1,
104 rm_addr : 1,
105 family : 4,
106 echo : 1,
107 backup : 1;
108 u32 token;
109 u32 nonce;
110 u64 thmac;
111 u8 hmac[20];
112 u8 join_id;
113 u8 use_map:1,
114 dsn64:1,
115 data_fin:1,
116 use_ack:1,
117 ack64:1,
118 mpc_map:1,
119 __unused:2;
120 u8 addr_id;
121 u8 rm_id;
122 union {
123 struct in_addr addr;
124#if IS_ENABLED(CONFIG_MPTCP_IPV6)
125 struct in6_addr addr6;
126#endif
127 };
128 u64 ahmac;
129 u16 port;
130};
131
132static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field)
133{
134 return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) |
135 ((nib & 0xF) << 8) | field);
136}
137
138struct mptcp_addr_info {
139 sa_family_t family;
140 __be16 port;
141 u8 id;
142 union {
143 struct in_addr addr;
144#if IS_ENABLED(CONFIG_MPTCP_IPV6)
145 struct in6_addr addr6;
146#endif
147 };
148};
149
150enum mptcp_pm_status {
151 MPTCP_PM_ADD_ADDR_RECEIVED,
152 MPTCP_PM_ESTABLISHED,
153 MPTCP_PM_SUBFLOW_ESTABLISHED,
154};
155
156struct mptcp_pm_data {
157 struct mptcp_addr_info local;
158 struct mptcp_addr_info remote;
159
160 spinlock_t lock; /*protects the whole PM data */
161
162 bool addr_signal;
163 bool server_side;
164 bool work_pending;
165 bool accept_addr;
166 bool accept_subflow;
167 u8 add_addr_signaled;
168 u8 add_addr_accepted;
169 u8 local_addr_used;
170 u8 subflows;
171 u8 add_addr_signal_max;
172 u8 add_addr_accept_max;
173 u8 local_addr_max;
174 u8 subflows_max;
175 u8 status;
176
177 struct work_struct work;
178};
179
180struct mptcp_data_frag {
181 struct list_head list;
182 u64 data_seq;
183 int data_len;
184 int offset;
185 int overhead;
186 struct page *page;
187};
188
189/* MPTCP connection sock */
190struct mptcp_sock {
191 /* inet_connection_sock must be the first member */
192 struct inet_connection_sock sk;
193 u64 local_key;
194 u64 remote_key;
195 u64 write_seq;
196 u64 ack_seq;
197 atomic64_t snd_una;
198 unsigned long timer_ival;
199 u32 token;
200 unsigned long flags;
201 bool can_ack;
202 spinlock_t join_list_lock;
203 struct work_struct work;
204 struct list_head conn_list;
205 struct list_head rtx_queue;
206 struct list_head join_list;
207 struct skb_ext *cached_ext; /* for the next sendmsg */
208 struct socket *subflow; /* outgoing connect/listener/!mp_capable */
209 struct sock *first;
210 struct mptcp_pm_data pm;
211};
212
213#define mptcp_for_each_subflow(__msk, __subflow) \
214 list_for_each_entry(__subflow, &((__msk)->conn_list), node)
215
216static inline struct mptcp_sock *mptcp_sk(const struct sock *sk)
217{
218 return (struct mptcp_sock *)sk;
219}
220
221static inline struct mptcp_data_frag *mptcp_rtx_tail(const struct sock *sk)
222{
223 struct mptcp_sock *msk = mptcp_sk(sk);
224
225 if (list_empty(&msk->rtx_queue))
226 return NULL;
227
228 return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list);
229}
230
231static inline struct mptcp_data_frag *mptcp_rtx_head(const struct sock *sk)
232{
233 struct mptcp_sock *msk = mptcp_sk(sk);
234
235 return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list);
236}
237
238struct mptcp_subflow_request_sock {
239 struct tcp_request_sock sk;
240 u16 mp_capable : 1,
241 mp_join : 1,
242 backup : 1;
243 u8 local_id;
244 u8 remote_id;
245 u64 local_key;
246 u64 idsn;
247 u32 token;
248 u32 ssn_offset;
249 u64 thmac;
250 u32 local_nonce;
251 u32 remote_nonce;
252 struct mptcp_sock *msk;
253};
254
255static inline struct mptcp_subflow_request_sock *
256mptcp_subflow_rsk(const struct request_sock *rsk)
257{
258 return (struct mptcp_subflow_request_sock *)rsk;
259}
260
261/* MPTCP subflow context */
262struct mptcp_subflow_context {
263 struct list_head node;/* conn_list of subflows */
264 u64 local_key;
265 u64 remote_key;
266 u64 idsn;
267 u64 map_seq;
268 u32 snd_isn;
269 u32 token;
270 u32 rel_write_seq;
271 u32 map_subflow_seq;
272 u32 ssn_offset;
273 u32 map_data_len;
274 u32 request_mptcp : 1, /* send MP_CAPABLE */
275 request_join : 1, /* send MP_JOIN */
276 request_bkup : 1,
277 mp_capable : 1, /* remote is MPTCP capable */
278 mp_join : 1, /* remote is JOINing */
279 fully_established : 1, /* path validated */
280 pm_notified : 1, /* PM hook called for established status */
281 conn_finished : 1,
282 map_valid : 1,
283 mpc_map : 1,
284 backup : 1,
285 data_avail : 1,
286 rx_eof : 1,
287 data_fin_tx_enable : 1,
288 use_64bit_ack : 1, /* Set when we received a 64-bit DSN */
289 can_ack : 1; /* only after processing the remote a key */
290 u64 data_fin_tx_seq;
291 u32 remote_nonce;
292 u64 thmac;
293 u32 local_nonce;
294 u32 remote_token;
295 u8 hmac[MPTCPOPT_HMAC_LEN];
296 u8 local_id;
297 u8 remote_id;
298
299 struct sock *tcp_sock; /* tcp sk backpointer */
300 struct sock *conn; /* parent mptcp_sock */
301 const struct inet_connection_sock_af_ops *icsk_af_ops;
302 void (*tcp_data_ready)(struct sock *sk);
303 void (*tcp_state_change)(struct sock *sk);
304 void (*tcp_write_space)(struct sock *sk);
305
306 struct rcu_head rcu;
307};
308
309static inline struct mptcp_subflow_context *
310mptcp_subflow_ctx(const struct sock *sk)
311{
312 struct inet_connection_sock *icsk = inet_csk(sk);
313
314 /* Use RCU on icsk_ulp_data only for sock diag code */
315 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data;
316}
317
318static inline struct sock *
319mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow)
320{
321 return subflow->tcp_sock;
322}
323
324static inline u64
325mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow)
326{
327 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq -
328 subflow->ssn_offset -
329 subflow->map_subflow_seq;
330}
331
332static inline u64
333mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow)
334{
335 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow);
336}
337
338int mptcp_is_enabled(struct net *net);
339bool mptcp_subflow_data_available(struct sock *sk);
340void mptcp_subflow_init(void);
341
342/* called with sk socket lock held */
343int __mptcp_subflow_connect(struct sock *sk, int ifindex,
344 const struct mptcp_addr_info *loc,
345 const struct mptcp_addr_info *remote);
346int mptcp_subflow_create_socket(struct sock *sk, struct socket **new_sock);
347
348static inline void mptcp_subflow_tcp_fallback(struct sock *sk,
349 struct mptcp_subflow_context *ctx)
350{
351 sk->sk_data_ready = ctx->tcp_data_ready;
352 sk->sk_state_change = ctx->tcp_state_change;
353 sk->sk_write_space = ctx->tcp_write_space;
354
355 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops;
356}
357
358extern const struct inet_connection_sock_af_ops ipv4_specific;
359#if IS_ENABLED(CONFIG_MPTCP_IPV6)
360extern const struct inet_connection_sock_af_ops ipv6_specific;
361#endif
362
363void mptcp_proto_init(void);
364#if IS_ENABLED(CONFIG_MPTCP_IPV6)
365int mptcp_proto_v6_init(void);
366#endif
367
368struct sock *mptcp_sk_clone(const struct sock *sk,
369 const struct mptcp_options_received *mp_opt,
370 struct request_sock *req);
371void mptcp_get_options(const struct sk_buff *skb,
372 struct mptcp_options_received *mp_opt);
373
374void mptcp_finish_connect(struct sock *sk);
375void mptcp_data_ready(struct sock *sk, struct sock *ssk);
376bool mptcp_finish_join(struct sock *sk);
377void mptcp_data_acked(struct sock *sk);
378void mptcp_subflow_eof(struct sock *sk);
379
380int mptcp_token_new_request(struct request_sock *req);
381void mptcp_token_destroy_request(u32 token);
382int mptcp_token_new_connect(struct sock *sk);
383int mptcp_token_new_accept(u32 token, struct sock *conn);
384struct mptcp_sock *mptcp_token_get_sock(u32 token);
385void mptcp_token_destroy(u32 token);
386
387void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn);
388static inline void mptcp_crypto_key_gen_sha(u64 *key, u32 *token, u64 *idsn)
389{
390 /* we might consider a faster version that computes the key as a
391 * hash of some information available in the MPTCP socket. Use
392 * random data at the moment, as it's probably the safest option
393 * in case multiple sockets are opened in different namespaces at
394 * the same time.
395 */
396 get_random_bytes(key, sizeof(u64));
397 mptcp_crypto_key_sha(*key, token, idsn);
398}
399
400void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac);
401
402void mptcp_pm_init(void);
403void mptcp_pm_data_init(struct mptcp_sock *msk);
404void mptcp_pm_close(struct mptcp_sock *msk);
405void mptcp_pm_new_connection(struct mptcp_sock *msk, int server_side);
406void mptcp_pm_fully_established(struct mptcp_sock *msk);
407bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk);
408void mptcp_pm_connection_closed(struct mptcp_sock *msk);
409void mptcp_pm_subflow_established(struct mptcp_sock *msk,
410 struct mptcp_subflow_context *subflow);
411void mptcp_pm_subflow_closed(struct mptcp_sock *msk, u8 id);
412void mptcp_pm_add_addr_received(struct mptcp_sock *msk,
413 const struct mptcp_addr_info *addr);
414
415int mptcp_pm_announce_addr(struct mptcp_sock *msk,
416 const struct mptcp_addr_info *addr);
417int mptcp_pm_remove_addr(struct mptcp_sock *msk, u8 local_id);
418int mptcp_pm_remove_subflow(struct mptcp_sock *msk, u8 remote_id);
419
420static inline bool mptcp_pm_should_signal(struct mptcp_sock *msk)
421{
422 return READ_ONCE(msk->pm.addr_signal);
423}
424
425static inline unsigned int mptcp_add_addr_len(int family)
426{
427 if (family == AF_INET)
428 return TCPOLEN_MPTCP_ADD_ADDR;
429 return TCPOLEN_MPTCP_ADD_ADDR6;
430}
431
432bool mptcp_pm_addr_signal(struct mptcp_sock *msk, unsigned int remaining,
433 struct mptcp_addr_info *saddr);
434int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
435
436void mptcp_pm_nl_init(void);
437void mptcp_pm_nl_data_init(struct mptcp_sock *msk);
438void mptcp_pm_nl_fully_established(struct mptcp_sock *msk);
439void mptcp_pm_nl_subflow_established(struct mptcp_sock *msk);
440void mptcp_pm_nl_add_addr_received(struct mptcp_sock *msk);
441int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
442
443static inline struct mptcp_ext *mptcp_get_ext(struct sk_buff *skb)
444{
445 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP);
446}
447
448static inline bool before64(__u64 seq1, __u64 seq2)
449{
450 return (__s64)(seq1 - seq2) < 0;
451}
452
453#define after64(seq2, seq1) before64(seq1, seq2)
454
455void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops);
456
457#endif /* __MPTCP_PROTOCOL_H */