]>
Commit | Line | Data |
---|---|---|
2303f994 PK |
1 | // SPDX-License-Identifier: GPL-2.0 |
2 | /* Multipath TCP | |
3 | * | |
4 | * Copyright (c) 2017 - 2019, Intel Corporation. | |
5 | */ | |
6 | ||
79c0949e PK |
7 | #define pr_fmt(fmt) "MPTCP: " fmt |
8 | ||
2303f994 PK |
9 | #include <linux/kernel.h> |
10 | #include <linux/module.h> | |
11 | #include <linux/netdevice.h> | |
f296234c | 12 | #include <crypto/algapi.h> |
bd697222 | 13 | #include <crypto/sha.h> |
2303f994 PK |
14 | #include <net/sock.h> |
15 | #include <net/inet_common.h> | |
16 | #include <net/inet_hashtables.h> | |
17 | #include <net/protocol.h> | |
18 | #include <net/tcp.h> | |
cec37a6e PK |
19 | #if IS_ENABLED(CONFIG_MPTCP_IPV6) |
20 | #include <net/ip6_route.h> | |
21 | #endif | |
2303f994 | 22 | #include <net/mptcp.h> |
4596a2c1 | 23 | #include <uapi/linux/mptcp.h> |
2303f994 | 24 | #include "protocol.h" |
fc518953 FW |
25 | #include "mib.h" |
26 | ||
27 | static void SUBFLOW_REQ_INC_STATS(struct request_sock *req, | |
28 | enum linux_mptcp_mib_field field) | |
29 | { | |
30 | MPTCP_INC_STATS(sock_net(req_to_sk(req)), field); | |
31 | } | |
2303f994 | 32 | |
79c0949e PK |
33 | static void subflow_req_destructor(struct request_sock *req) |
34 | { | |
35 | struct mptcp_subflow_request_sock *subflow_req = mptcp_subflow_rsk(req); | |
36 | ||
37 | pr_debug("subflow_req=%p", subflow_req); | |
38 | ||
8fd4de12 PA |
39 | if (subflow_req->msk) |
40 | sock_put((struct sock *)subflow_req->msk); | |
41 | ||
2c5ebd00 | 42 | mptcp_token_destroy_request(req); |
79c0949e PK |
43 | tcp_request_sock_ops.destructor(req); |
44 | } | |
45 | ||
f296234c PK |
46 | static void subflow_generate_hmac(u64 key1, u64 key2, u32 nonce1, u32 nonce2, |
47 | void *hmac) | |
48 | { | |
49 | u8 msg[8]; | |
50 | ||
51 | put_unaligned_be32(nonce1, &msg[0]); | |
52 | put_unaligned_be32(nonce2, &msg[4]); | |
53 | ||
54 | mptcp_crypto_hmac_sha(key1, key2, msg, 8, hmac); | |
55 | } | |
56 | ||
4cf8b7e4 PA |
57 | static bool mptcp_can_accept_new_subflow(const struct mptcp_sock *msk) |
58 | { | |
59 | return mptcp_is_fully_established((void *)msk) && | |
60 | READ_ONCE(msk->pm.accept_subflow); | |
61 | } | |
62 | ||
f296234c | 63 | /* validate received token and create truncated hmac and nonce for SYN-ACK */ |
8fd4de12 PA |
64 | static struct mptcp_sock *subflow_token_join_request(struct request_sock *req, |
65 | const struct sk_buff *skb) | |
f296234c PK |
66 | { |
67 | struct mptcp_subflow_request_sock *subflow_req = mptcp_subflow_rsk(req); | |
bd697222 | 68 | u8 hmac[SHA256_DIGEST_SIZE]; |
f296234c PK |
69 | struct mptcp_sock *msk; |
70 | int local_id; | |
71 | ||
72 | msk = mptcp_token_get_sock(subflow_req->token); | |
73 | if (!msk) { | |
fc518953 | 74 | SUBFLOW_REQ_INC_STATS(req, MPTCP_MIB_JOINNOTOKEN); |
8fd4de12 | 75 | return NULL; |
f296234c PK |
76 | } |
77 | ||
78 | local_id = mptcp_pm_get_local_id(msk, (struct sock_common *)req); | |
79 | if (local_id < 0) { | |
80 | sock_put((struct sock *)msk); | |
8fd4de12 | 81 | return NULL; |
f296234c PK |
82 | } |
83 | subflow_req->local_id = local_id; | |
84 | ||
85 | get_random_bytes(&subflow_req->local_nonce, sizeof(u32)); | |
86 | ||
87 | subflow_generate_hmac(msk->local_key, msk->remote_key, | |
88 | subflow_req->local_nonce, | |
89 | subflow_req->remote_nonce, hmac); | |
90 | ||
91 | subflow_req->thmac = get_unaligned_be64(hmac); | |
8fd4de12 | 92 | return msk; |
f296234c PK |
93 | } |
94 | ||
78d8b7bc | 95 | static int __subflow_init_req(struct request_sock *req, const struct sock *sk_listener) |
cec37a6e | 96 | { |
cec37a6e | 97 | struct mptcp_subflow_request_sock *subflow_req = mptcp_subflow_rsk(req); |
cec37a6e PK |
98 | |
99 | subflow_req->mp_capable = 0; | |
f296234c | 100 | subflow_req->mp_join = 0; |
8fd4de12 | 101 | subflow_req->msk = NULL; |
2c5ebd00 | 102 | mptcp_token_init_request(req); |
cec37a6e PK |
103 | |
104 | #ifdef CONFIG_TCP_MD5SIG | |
105 | /* no MPTCP if MD5SIG is enabled on this socket or we may run out of | |
106 | * TCP option space. | |
107 | */ | |
108 | if (rcu_access_pointer(tcp_sk(sk_listener)->md5sig_info)) | |
78d8b7bc | 109 | return -EINVAL; |
cec37a6e PK |
110 | #endif |
111 | ||
78d8b7bc FW |
112 | return 0; |
113 | } | |
114 | ||
3ecfbe3e FW |
115 | /* Init mptcp request socket. |
116 | * | |
117 | * Returns an error code if a JOIN has failed and a TCP reset | |
118 | * should be sent. | |
119 | */ | |
120 | static int subflow_init_req(struct request_sock *req, | |
121 | const struct sock *sk_listener, | |
122 | struct sk_buff *skb) | |
78d8b7bc FW |
123 | { |
124 | struct mptcp_subflow_context *listener = mptcp_subflow_ctx(sk_listener); | |
125 | struct mptcp_subflow_request_sock *subflow_req = mptcp_subflow_rsk(req); | |
126 | struct mptcp_options_received mp_opt; | |
127 | int ret; | |
128 | ||
129 | pr_debug("subflow_req=%p, listener=%p", subflow_req, listener); | |
130 | ||
131 | ret = __subflow_init_req(req, sk_listener); | |
132 | if (ret) | |
3ecfbe3e | 133 | return 0; |
78d8b7bc FW |
134 | |
135 | mptcp_get_options(skb, &mp_opt); | |
136 | ||
cfde141e | 137 | if (mp_opt.mp_capable) { |
fc518953 FW |
138 | SUBFLOW_REQ_INC_STATS(req, MPTCP_MIB_MPCAPABLEPASSIVE); |
139 | ||
cfde141e | 140 | if (mp_opt.mp_join) |
3ecfbe3e | 141 | return 0; |
cfde141e | 142 | } else if (mp_opt.mp_join) { |
fc518953 FW |
143 | SUBFLOW_REQ_INC_STATS(req, MPTCP_MIB_JOINSYNRX); |
144 | } | |
f296234c | 145 | |
cfde141e | 146 | if (mp_opt.mp_capable && listener->request_mptcp) { |
535fb815 FW |
147 | int err, retries = 4; |
148 | ||
c83a47e5 | 149 | subflow_req->ssn_offset = TCP_SKB_CB(skb)->seq; |
535fb815 FW |
150 | again: |
151 | do { | |
152 | get_random_bytes(&subflow_req->local_key, sizeof(subflow_req->local_key)); | |
153 | } while (subflow_req->local_key == 0); | |
79c0949e | 154 | |
c83a47e5 FW |
155 | if (unlikely(req->syncookie)) { |
156 | mptcp_crypto_key_sha(subflow_req->local_key, | |
157 | &subflow_req->token, | |
158 | &subflow_req->idsn); | |
159 | if (mptcp_token_exists(subflow_req->token)) { | |
160 | if (retries-- > 0) | |
161 | goto again; | |
162 | } else { | |
163 | subflow_req->mp_capable = 1; | |
164 | } | |
3ecfbe3e | 165 | return 0; |
c83a47e5 FW |
166 | } |
167 | ||
79c0949e PK |
168 | err = mptcp_token_new_request(req); |
169 | if (err == 0) | |
170 | subflow_req->mp_capable = 1; | |
535fb815 FW |
171 | else if (retries-- > 0) |
172 | goto again; | |
79c0949e | 173 | |
cfde141e | 174 | } else if (mp_opt.mp_join && listener->request_mptcp) { |
ec3edaa7 | 175 | subflow_req->ssn_offset = TCP_SKB_CB(skb)->seq; |
f296234c | 176 | subflow_req->mp_join = 1; |
cfde141e PA |
177 | subflow_req->backup = mp_opt.backup; |
178 | subflow_req->remote_id = mp_opt.join_id; | |
179 | subflow_req->token = mp_opt.token; | |
180 | subflow_req->remote_nonce = mp_opt.nonce; | |
8fd4de12 | 181 | subflow_req->msk = subflow_token_join_request(req, skb); |
9466a1cc | 182 | |
3ecfbe3e FW |
183 | /* Can't fall back to TCP in this case. */ |
184 | if (!subflow_req->msk) | |
185 | return -EPERM; | |
186 | ||
187 | if (unlikely(req->syncookie)) { | |
9466a1cc FW |
188 | if (mptcp_can_accept_new_subflow(subflow_req->msk)) |
189 | subflow_init_req_cookie_join_save(subflow_req, skb); | |
190 | } | |
191 | ||
8fd4de12 PA |
192 | pr_debug("token=%u, remote_nonce=%u msk=%p", subflow_req->token, |
193 | subflow_req->remote_nonce, subflow_req->msk); | |
cec37a6e | 194 | } |
3ecfbe3e FW |
195 | |
196 | return 0; | |
cec37a6e PK |
197 | } |
198 | ||
c83a47e5 FW |
199 | int mptcp_subflow_init_cookie_req(struct request_sock *req, |
200 | const struct sock *sk_listener, | |
201 | struct sk_buff *skb) | |
202 | { | |
203 | struct mptcp_subflow_context *listener = mptcp_subflow_ctx(sk_listener); | |
204 | struct mptcp_subflow_request_sock *subflow_req = mptcp_subflow_rsk(req); | |
205 | struct mptcp_options_received mp_opt; | |
206 | int err; | |
207 | ||
208 | err = __subflow_init_req(req, sk_listener); | |
209 | if (err) | |
210 | return err; | |
211 | ||
212 | mptcp_get_options(skb, &mp_opt); | |
213 | ||
214 | if (mp_opt.mp_capable && mp_opt.mp_join) | |
215 | return -EINVAL; | |
216 | ||
217 | if (mp_opt.mp_capable && listener->request_mptcp) { | |
218 | if (mp_opt.sndr_key == 0) | |
219 | return -EINVAL; | |
220 | ||
221 | subflow_req->local_key = mp_opt.rcvr_key; | |
222 | err = mptcp_token_new_request(req); | |
223 | if (err) | |
224 | return err; | |
225 | ||
226 | subflow_req->mp_capable = 1; | |
227 | subflow_req->ssn_offset = TCP_SKB_CB(skb)->seq - 1; | |
9466a1cc FW |
228 | } else if (mp_opt.mp_join && listener->request_mptcp) { |
229 | if (!mptcp_token_join_cookie_init_state(subflow_req, skb)) | |
230 | return -EINVAL; | |
231 | ||
232 | if (mptcp_can_accept_new_subflow(subflow_req->msk)) | |
233 | subflow_req->mp_join = 1; | |
234 | ||
235 | subflow_req->ssn_offset = TCP_SKB_CB(skb)->seq - 1; | |
c83a47e5 FW |
236 | } |
237 | ||
238 | return 0; | |
239 | } | |
240 | EXPORT_SYMBOL_GPL(mptcp_subflow_init_cookie_req); | |
241 | ||
7ea851d1 FW |
242 | static struct dst_entry *subflow_v4_route_req(const struct sock *sk, |
243 | struct sk_buff *skb, | |
244 | struct flowi *fl, | |
245 | struct request_sock *req) | |
cec37a6e | 246 | { |
7ea851d1 | 247 | struct dst_entry *dst; |
3ecfbe3e | 248 | int err; |
7ea851d1 | 249 | |
cec37a6e PK |
250 | tcp_rsk(req)->is_mptcp = 1; |
251 | ||
7ea851d1 FW |
252 | dst = tcp_request_sock_ipv4_ops.route_req(sk, skb, fl, req); |
253 | if (!dst) | |
254 | return NULL; | |
cec37a6e | 255 | |
3ecfbe3e FW |
256 | err = subflow_init_req(req, sk, skb); |
257 | if (err == 0) | |
258 | return dst; | |
259 | ||
260 | dst_release(dst); | |
261 | if (!req->syncookie) | |
262 | tcp_request_sock_ops.send_reset(sk, skb); | |
263 | return NULL; | |
cec37a6e PK |
264 | } |
265 | ||
266 | #if IS_ENABLED(CONFIG_MPTCP_IPV6) | |
7ea851d1 FW |
267 | static struct dst_entry *subflow_v6_route_req(const struct sock *sk, |
268 | struct sk_buff *skb, | |
269 | struct flowi *fl, | |
270 | struct request_sock *req) | |
cec37a6e | 271 | { |
7ea851d1 | 272 | struct dst_entry *dst; |
3ecfbe3e | 273 | int err; |
7ea851d1 | 274 | |
cec37a6e PK |
275 | tcp_rsk(req)->is_mptcp = 1; |
276 | ||
7ea851d1 FW |
277 | dst = tcp_request_sock_ipv6_ops.route_req(sk, skb, fl, req); |
278 | if (!dst) | |
279 | return NULL; | |
cec37a6e | 280 | |
3ecfbe3e FW |
281 | err = subflow_init_req(req, sk, skb); |
282 | if (err == 0) | |
283 | return dst; | |
284 | ||
285 | dst_release(dst); | |
286 | if (!req->syncookie) | |
287 | tcp6_request_sock_ops.send_reset(sk, skb); | |
288 | return NULL; | |
cec37a6e PK |
289 | } |
290 | #endif | |
291 | ||
ec3edaa7 PK |
292 | /* validate received truncated hmac and create hmac for third ACK */ |
293 | static bool subflow_thmac_valid(struct mptcp_subflow_context *subflow) | |
294 | { | |
bd697222 | 295 | u8 hmac[SHA256_DIGEST_SIZE]; |
ec3edaa7 PK |
296 | u64 thmac; |
297 | ||
298 | subflow_generate_hmac(subflow->remote_key, subflow->local_key, | |
299 | subflow->remote_nonce, subflow->local_nonce, | |
300 | hmac); | |
301 | ||
302 | thmac = get_unaligned_be64(hmac); | |
303 | pr_debug("subflow=%p, token=%u, thmac=%llu, subflow->thmac=%llu\n", | |
304 | subflow, subflow->token, | |
305 | (unsigned long long)thmac, | |
306 | (unsigned long long)subflow->thmac); | |
307 | ||
308 | return thmac == subflow->thmac; | |
309 | } | |
310 | ||
d5824847 PA |
311 | void mptcp_subflow_reset(struct sock *ssk) |
312 | { | |
0e4f35d7 PA |
313 | struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); |
314 | struct sock *sk = subflow->conn; | |
315 | ||
d5824847 PA |
316 | tcp_set_state(ssk, TCP_CLOSE); |
317 | tcp_send_active_reset(ssk, GFP_ATOMIC); | |
318 | tcp_done(ssk); | |
0e4f35d7 PA |
319 | if (!test_and_set_bit(MPTCP_WORK_CLOSE_SUBFLOW, &mptcp_sk(sk)->flags) && |
320 | schedule_work(&mptcp_sk(sk)->work)) | |
321 | sock_hold(sk); | |
d5824847 PA |
322 | } |
323 | ||
cec37a6e PK |
324 | static void subflow_finish_connect(struct sock *sk, const struct sk_buff *skb) |
325 | { | |
326 | struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); | |
cfde141e | 327 | struct mptcp_options_received mp_opt; |
c3c123d1 | 328 | struct sock *parent = subflow->conn; |
cec37a6e PK |
329 | |
330 | subflow->icsk_af_ops->sk_rx_dst_set(sk, skb); | |
331 | ||
1200832c | 332 | if (inet_sk_state_load(parent) == TCP_SYN_SENT) { |
c3c123d1 DC |
333 | inet_sk_state_store(parent, TCP_ESTABLISHED); |
334 | parent->sk_state_change(parent); | |
335 | } | |
336 | ||
263e1201 PA |
337 | /* be sure no special action on any packet other than syn-ack */ |
338 | if (subflow->conn_finished) | |
339 | return; | |
340 | ||
b0977bb2 | 341 | subflow->rel_write_seq = 1; |
263e1201 | 342 | subflow->conn_finished = 1; |
e1ff9e82 DC |
343 | subflow->ssn_offset = TCP_SKB_CB(skb)->seq; |
344 | pr_debug("subflow=%p synack seq=%x", subflow, subflow->ssn_offset); | |
263e1201 | 345 | |
cfde141e | 346 | mptcp_get_options(skb, &mp_opt); |
fa25e815 PA |
347 | if (subflow->request_mptcp) { |
348 | if (!mp_opt.mp_capable) { | |
349 | MPTCP_INC_STATS(sock_net(sk), | |
350 | MPTCP_MIB_MPCAPABLEACTIVEFALLBACK); | |
351 | mptcp_do_fallback(sk); | |
352 | pr_fallback(mptcp_sk(subflow->conn)); | |
353 | goto fallback; | |
354 | } | |
355 | ||
263e1201 PA |
356 | subflow->mp_capable = 1; |
357 | subflow->can_ack = 1; | |
cfde141e | 358 | subflow->remote_key = mp_opt.sndr_key; |
263e1201 PA |
359 | pr_debug("subflow=%p, remote_key=%llu", subflow, |
360 | subflow->remote_key); | |
fa25e815 PA |
361 | mptcp_finish_connect(sk); |
362 | } else if (subflow->request_join) { | |
363 | u8 hmac[SHA256_DIGEST_SIZE]; | |
364 | ||
365 | if (!mp_opt.mp_join) | |
366 | goto do_reset; | |
367 | ||
cfde141e PA |
368 | subflow->thmac = mp_opt.thmac; |
369 | subflow->remote_nonce = mp_opt.nonce; | |
263e1201 PA |
370 | pr_debug("subflow=%p, thmac=%llu, remote_nonce=%u", subflow, |
371 | subflow->thmac, subflow->remote_nonce); | |
263e1201 | 372 | |
ec3edaa7 | 373 | if (!subflow_thmac_valid(subflow)) { |
fc518953 | 374 | MPTCP_INC_STATS(sock_net(sk), MPTCP_MIB_JOINACKMAC); |
ec3edaa7 PK |
375 | goto do_reset; |
376 | } | |
377 | ||
378 | subflow_generate_hmac(subflow->local_key, subflow->remote_key, | |
379 | subflow->local_nonce, | |
380 | subflow->remote_nonce, | |
bd697222 | 381 | hmac); |
bd697222 | 382 | memcpy(subflow->hmac, hmac, MPTCPOPT_HMAC_LEN); |
ec3edaa7 | 383 | |
ec3edaa7 PK |
384 | if (!mptcp_finish_join(sk)) |
385 | goto do_reset; | |
386 | ||
fa25e815 | 387 | subflow->mp_join = 1; |
fc518953 | 388 | MPTCP_INC_STATS(sock_net(sk), MPTCP_MIB_JOINSYNACKRX); |
fa25e815 PA |
389 | } else if (mptcp_check_fallback(sk)) { |
390 | fallback: | |
391 | mptcp_rcv_space_init(mptcp_sk(parent), sk); | |
cec37a6e | 392 | } |
fa25e815 PA |
393 | return; |
394 | ||
395 | do_reset: | |
d5824847 | 396 | mptcp_subflow_reset(sk); |
cec37a6e PK |
397 | } |
398 | ||
08b8d080 FW |
399 | struct request_sock_ops mptcp_subflow_request_sock_ops; |
400 | EXPORT_SYMBOL_GPL(mptcp_subflow_request_sock_ops); | |
cec37a6e PK |
401 | static struct tcp_request_sock_ops subflow_request_sock_ipv4_ops; |
402 | ||
403 | static int subflow_v4_conn_request(struct sock *sk, struct sk_buff *skb) | |
404 | { | |
405 | struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); | |
406 | ||
407 | pr_debug("subflow=%p", subflow); | |
408 | ||
409 | /* Never answer to SYNs sent to broadcast or multicast */ | |
410 | if (skb_rtable(skb)->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) | |
411 | goto drop; | |
412 | ||
08b8d080 | 413 | return tcp_conn_request(&mptcp_subflow_request_sock_ops, |
cec37a6e PK |
414 | &subflow_request_sock_ipv4_ops, |
415 | sk, skb); | |
416 | drop: | |
417 | tcp_listendrop(sk); | |
418 | return 0; | |
419 | } | |
420 | ||
421 | #if IS_ENABLED(CONFIG_MPTCP_IPV6) | |
422 | static struct tcp_request_sock_ops subflow_request_sock_ipv6_ops; | |
423 | static struct inet_connection_sock_af_ops subflow_v6_specific; | |
424 | static struct inet_connection_sock_af_ops subflow_v6m_specific; | |
425 | ||
426 | static int subflow_v6_conn_request(struct sock *sk, struct sk_buff *skb) | |
427 | { | |
428 | struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); | |
429 | ||
430 | pr_debug("subflow=%p", subflow); | |
431 | ||
432 | if (skb->protocol == htons(ETH_P_IP)) | |
433 | return subflow_v4_conn_request(sk, skb); | |
434 | ||
435 | if (!ipv6_unicast_destination(skb)) | |
436 | goto drop; | |
437 | ||
08b8d080 | 438 | return tcp_conn_request(&mptcp_subflow_request_sock_ops, |
cec37a6e PK |
439 | &subflow_request_sock_ipv6_ops, sk, skb); |
440 | ||
441 | drop: | |
442 | tcp_listendrop(sk); | |
443 | return 0; /* don't send reset */ | |
444 | } | |
445 | #endif | |
446 | ||
f296234c PK |
447 | /* validate hmac received in third ACK */ |
448 | static bool subflow_hmac_valid(const struct request_sock *req, | |
cfde141e | 449 | const struct mptcp_options_received *mp_opt) |
f296234c PK |
450 | { |
451 | const struct mptcp_subflow_request_sock *subflow_req; | |
bd697222 | 452 | u8 hmac[SHA256_DIGEST_SIZE]; |
f296234c | 453 | struct mptcp_sock *msk; |
f296234c PK |
454 | |
455 | subflow_req = mptcp_subflow_rsk(req); | |
8fd4de12 | 456 | msk = subflow_req->msk; |
f296234c PK |
457 | if (!msk) |
458 | return false; | |
459 | ||
460 | subflow_generate_hmac(msk->remote_key, msk->local_key, | |
461 | subflow_req->remote_nonce, | |
462 | subflow_req->local_nonce, hmac); | |
463 | ||
8fd4de12 | 464 | return !crypto_memneq(hmac, mp_opt->hmac, MPTCPOPT_HMAC_LEN); |
f296234c PK |
465 | } |
466 | ||
df1036da FW |
467 | static void mptcp_sock_destruct(struct sock *sk) |
468 | { | |
469 | /* if new mptcp socket isn't accepted, it is free'd | |
470 | * from the tcp listener sockets request queue, linked | |
471 | * from req->sk. The tcp socket is released. | |
472 | * This calls the ULP release function which will | |
473 | * also remove the mptcp socket, via | |
474 | * sock_put(ctx->conn). | |
475 | * | |
7ee24926 PA |
476 | * Problem is that the mptcp socket will be in |
477 | * ESTABLISHED state and will not have the SOCK_DEAD flag. | |
df1036da FW |
478 | * Both result in warnings from inet_sock_destruct. |
479 | */ | |
480 | ||
7ee24926 | 481 | if (sk->sk_state == TCP_ESTABLISHED) { |
df1036da FW |
482 | sk->sk_state = TCP_CLOSE; |
483 | WARN_ON_ONCE(sk->sk_socket); | |
484 | sock_orphan(sk); | |
485 | } | |
486 | ||
5c8c1640 | 487 | mptcp_destroy_common(mptcp_sk(sk)); |
df1036da FW |
488 | inet_sock_destruct(sk); |
489 | } | |
490 | ||
9f5ca6a5 FW |
491 | static void mptcp_force_close(struct sock *sk) |
492 | { | |
493 | inet_sk_state_store(sk, TCP_CLOSE); | |
494 | sk_common_release(sk); | |
495 | } | |
496 | ||
4c8941de PA |
497 | static void subflow_ulp_fallback(struct sock *sk, |
498 | struct mptcp_subflow_context *old_ctx) | |
499 | { | |
500 | struct inet_connection_sock *icsk = inet_csk(sk); | |
501 | ||
502 | mptcp_subflow_tcp_fallback(sk, old_ctx); | |
503 | icsk->icsk_ulp_ops = NULL; | |
504 | rcu_assign_pointer(icsk->icsk_ulp_data, NULL); | |
505 | tcp_sk(sk)->is_mptcp = 0; | |
506 | } | |
507 | ||
39884604 PA |
508 | static void subflow_drop_ctx(struct sock *ssk) |
509 | { | |
510 | struct mptcp_subflow_context *ctx = mptcp_subflow_ctx(ssk); | |
511 | ||
512 | if (!ctx) | |
513 | return; | |
514 | ||
515 | subflow_ulp_fallback(ssk, ctx); | |
516 | if (ctx->conn) | |
517 | sock_put(ctx->conn); | |
518 | ||
519 | kfree_rcu(ctx, rcu); | |
520 | } | |
521 | ||
b93df08c PA |
522 | void mptcp_subflow_fully_established(struct mptcp_subflow_context *subflow, |
523 | struct mptcp_options_received *mp_opt) | |
524 | { | |
525 | struct mptcp_sock *msk = mptcp_sk(subflow->conn); | |
526 | ||
527 | subflow->remote_key = mp_opt->sndr_key; | |
528 | subflow->fully_established = 1; | |
529 | subflow->can_ack = 1; | |
530 | WRITE_ONCE(msk->fully_established, true); | |
531 | } | |
532 | ||
cec37a6e PK |
533 | static struct sock *subflow_syn_recv_sock(const struct sock *sk, |
534 | struct sk_buff *skb, | |
535 | struct request_sock *req, | |
536 | struct dst_entry *dst, | |
537 | struct request_sock *req_unhash, | |
538 | bool *own_req) | |
539 | { | |
540 | struct mptcp_subflow_context *listener = mptcp_subflow_ctx(sk); | |
cc7972ea | 541 | struct mptcp_subflow_request_sock *subflow_req; |
cfde141e | 542 | struct mptcp_options_received mp_opt; |
9e365ff5 | 543 | bool fallback, fallback_is_fatal; |
58b09919 | 544 | struct sock *new_msk = NULL; |
cec37a6e PK |
545 | struct sock *child; |
546 | ||
547 | pr_debug("listener=%p, req=%p, conn=%p", listener, req, listener->conn); | |
548 | ||
9e365ff5 PA |
549 | /* After child creation we must look for 'mp_capable' even when options |
550 | * are not parsed | |
cfde141e PA |
551 | */ |
552 | mp_opt.mp_capable = 0; | |
9e365ff5 PA |
553 | |
554 | /* hopefully temporary handling for MP_JOIN+syncookie */ | |
555 | subflow_req = mptcp_subflow_rsk(req); | |
b7514694 | 556 | fallback_is_fatal = tcp_rsk(req)->is_mptcp && subflow_req->mp_join; |
9e365ff5 PA |
557 | fallback = !tcp_rsk(req)->is_mptcp; |
558 | if (fallback) | |
ae2dd716 FW |
559 | goto create_child; |
560 | ||
d22f4988 | 561 | /* if the sk is MP_CAPABLE, we try to fetch the client key */ |
cc7972ea | 562 | if (subflow_req->mp_capable) { |
d22f4988 CP |
563 | if (TCP_SKB_CB(skb)->seq != subflow_req->ssn_offset + 1) { |
564 | /* here we can receive and accept an in-window, | |
565 | * out-of-order pkt, which will not carry the MP_CAPABLE | |
566 | * opt even on mptcp enabled paths | |
567 | */ | |
58b09919 | 568 | goto create_msk; |
d22f4988 CP |
569 | } |
570 | ||
cfde141e PA |
571 | mptcp_get_options(skb, &mp_opt); |
572 | if (!mp_opt.mp_capable) { | |
4c8941de | 573 | fallback = true; |
58b09919 | 574 | goto create_child; |
d22f4988 | 575 | } |
58b09919 PA |
576 | |
577 | create_msk: | |
cfde141e | 578 | new_msk = mptcp_sk_clone(listener->conn, &mp_opt, req); |
58b09919 | 579 | if (!new_msk) |
4c8941de | 580 | fallback = true; |
f296234c | 581 | } else if (subflow_req->mp_join) { |
cfde141e | 582 | mptcp_get_options(skb, &mp_opt); |
d3ab7885 PA |
583 | if (!mp_opt.mp_join || !subflow_hmac_valid(req, &mp_opt) || |
584 | !mptcp_can_accept_new_subflow(subflow_req->msk)) { | |
fc518953 | 585 | SUBFLOW_REQ_INC_STATS(req, MPTCP_MIB_JOINACKMAC); |
9e365ff5 | 586 | fallback = true; |
fc518953 | 587 | } |
cc7972ea | 588 | } |
cec37a6e | 589 | |
d22f4988 | 590 | create_child: |
cec37a6e PK |
591 | child = listener->icsk_af_ops->syn_recv_sock(sk, skb, req, dst, |
592 | req_unhash, own_req); | |
593 | ||
594 | if (child && *own_req) { | |
79c0949e PK |
595 | struct mptcp_subflow_context *ctx = mptcp_subflow_ctx(child); |
596 | ||
90bf4513 PA |
597 | tcp_rsk(req)->drop_req = false; |
598 | ||
4c8941de PA |
599 | /* we need to fallback on ctx allocation failure and on pre-reqs |
600 | * checking above. In the latter scenario we additionally need | |
601 | * to reset the context to non MPTCP status. | |
79c0949e | 602 | */ |
4c8941de | 603 | if (!ctx || fallback) { |
f296234c | 604 | if (fallback_is_fatal) |
729cd643 | 605 | goto dispose_child; |
4c8941de | 606 | |
39884604 | 607 | subflow_drop_ctx(child); |
58b09919 | 608 | goto out; |
f296234c | 609 | } |
79c0949e PK |
610 | |
611 | if (ctx->mp_capable) { | |
b93df08c PA |
612 | /* this can't race with mptcp_close(), as the msk is |
613 | * not yet exposted to user-space | |
614 | */ | |
615 | inet_sk_state_store((void *)new_msk, TCP_ESTABLISHED); | |
616 | ||
5b950ff4 PA |
617 | /* record the newly created socket as the first msk |
618 | * subflow, but don't link it yet into conn_list | |
619 | */ | |
0397c6d8 PA |
620 | WRITE_ONCE(mptcp_sk(new_msk)->first, child); |
621 | ||
58b09919 PA |
622 | /* new mpc subflow takes ownership of the newly |
623 | * created mptcp socket | |
624 | */ | |
df1036da | 625 | new_msk->sk_destruct = mptcp_sock_destruct; |
1b1c7a0e | 626 | mptcp_pm_new_connection(mptcp_sk(new_msk), 1); |
2c5ebd00 | 627 | mptcp_token_accept(subflow_req, mptcp_sk(new_msk)); |
58b09919 PA |
628 | ctx->conn = new_msk; |
629 | new_msk = NULL; | |
fca5c82c PA |
630 | |
631 | /* with OoO packets we can reach here without ingress | |
632 | * mpc option | |
633 | */ | |
b93df08c PA |
634 | if (mp_opt.mp_capable) |
635 | mptcp_subflow_fully_established(ctx, &mp_opt); | |
f296234c PK |
636 | } else if (ctx->mp_join) { |
637 | struct mptcp_sock *owner; | |
638 | ||
8fd4de12 | 639 | owner = subflow_req->msk; |
f296234c | 640 | if (!owner) |
729cd643 | 641 | goto dispose_child; |
f296234c | 642 | |
8fd4de12 PA |
643 | /* move the msk reference ownership to the subflow */ |
644 | subflow_req->msk = NULL; | |
f296234c PK |
645 | ctx->conn = (struct sock *)owner; |
646 | if (!mptcp_finish_join(child)) | |
729cd643 | 647 | goto dispose_child; |
fc518953 FW |
648 | |
649 | SUBFLOW_REQ_INC_STATS(req, MPTCP_MIB_JOINACKRX); | |
90bf4513 | 650 | tcp_rsk(req)->drop_req = true; |
cec37a6e PK |
651 | } |
652 | } | |
653 | ||
58b09919 PA |
654 | out: |
655 | /* dispose of the left over mptcp master, if any */ | |
656 | if (unlikely(new_msk)) | |
9f5ca6a5 | 657 | mptcp_force_close(new_msk); |
4c8941de PA |
658 | |
659 | /* check for expected invariant - should never trigger, just help | |
660 | * catching eariler subtle bugs | |
661 | */ | |
ac2b47fb | 662 | WARN_ON_ONCE(child && *own_req && tcp_sk(child)->is_mptcp && |
4c8941de PA |
663 | (!mptcp_subflow_ctx(child) || |
664 | !mptcp_subflow_ctx(child)->conn)); | |
cec37a6e | 665 | return child; |
f296234c | 666 | |
729cd643 | 667 | dispose_child: |
39884604 | 668 | subflow_drop_ctx(child); |
729cd643 | 669 | tcp_rsk(req)->drop_req = true; |
729cd643 | 670 | inet_csk_prepare_for_destroy_sock(child); |
f296234c | 671 | tcp_done(child); |
97e61751 | 672 | req->rsk_ops->send_reset(sk, skb); |
729cd643 PA |
673 | |
674 | /* The last child reference will be released by the caller */ | |
675 | return child; | |
cec37a6e PK |
676 | } |
677 | ||
678 | static struct inet_connection_sock_af_ops subflow_specific; | |
679 | ||
648ef4b8 MM |
680 | enum mapping_status { |
681 | MAPPING_OK, | |
682 | MAPPING_INVALID, | |
683 | MAPPING_EMPTY, | |
e1ff9e82 DC |
684 | MAPPING_DATA_FIN, |
685 | MAPPING_DUMMY | |
648ef4b8 MM |
686 | }; |
687 | ||
688 | static u64 expand_seq(u64 old_seq, u16 old_data_len, u64 seq) | |
689 | { | |
690 | if ((u32)seq == (u32)old_seq) | |
691 | return old_seq; | |
692 | ||
693 | /* Assume map covers data not mapped yet. */ | |
694 | return seq | ((old_seq + old_data_len + 1) & GENMASK_ULL(63, 32)); | |
695 | } | |
696 | ||
697 | static void warn_bad_map(struct mptcp_subflow_context *subflow, u32 ssn) | |
698 | { | |
699 | WARN_ONCE(1, "Bad mapping: ssn=%d map_seq=%d map_data_len=%d", | |
700 | ssn, subflow->map_subflow_seq, subflow->map_data_len); | |
701 | } | |
702 | ||
703 | static bool skb_is_fully_mapped(struct sock *ssk, struct sk_buff *skb) | |
704 | { | |
705 | struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); | |
706 | unsigned int skb_consumed; | |
707 | ||
708 | skb_consumed = tcp_sk(ssk)->copied_seq - TCP_SKB_CB(skb)->seq; | |
709 | if (WARN_ON_ONCE(skb_consumed >= skb->len)) | |
710 | return true; | |
711 | ||
712 | return skb->len - skb_consumed <= subflow->map_data_len - | |
713 | mptcp_subflow_get_map_offset(subflow); | |
714 | } | |
715 | ||
716 | static bool validate_mapping(struct sock *ssk, struct sk_buff *skb) | |
717 | { | |
718 | struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); | |
719 | u32 ssn = tcp_sk(ssk)->copied_seq - subflow->ssn_offset; | |
720 | ||
721 | if (unlikely(before(ssn, subflow->map_subflow_seq))) { | |
722 | /* Mapping covers data later in the subflow stream, | |
723 | * currently unsupported. | |
724 | */ | |
725 | warn_bad_map(subflow, ssn); | |
726 | return false; | |
727 | } | |
728 | if (unlikely(!before(ssn, subflow->map_subflow_seq + | |
729 | subflow->map_data_len))) { | |
730 | /* Mapping does covers past subflow data, invalid */ | |
731 | warn_bad_map(subflow, ssn + skb->len); | |
732 | return false; | |
733 | } | |
734 | return true; | |
735 | } | |
736 | ||
43b54c6e MM |
737 | static enum mapping_status get_mapping_status(struct sock *ssk, |
738 | struct mptcp_sock *msk) | |
648ef4b8 MM |
739 | { |
740 | struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); | |
741 | struct mptcp_ext *mpext; | |
742 | struct sk_buff *skb; | |
743 | u16 data_len; | |
744 | u64 map_seq; | |
745 | ||
746 | skb = skb_peek(&ssk->sk_receive_queue); | |
747 | if (!skb) | |
748 | return MAPPING_EMPTY; | |
749 | ||
e1ff9e82 DC |
750 | if (mptcp_check_fallback(ssk)) |
751 | return MAPPING_DUMMY; | |
752 | ||
648ef4b8 MM |
753 | mpext = mptcp_get_ext(skb); |
754 | if (!mpext || !mpext->use_map) { | |
755 | if (!subflow->map_valid && !skb->len) { | |
756 | /* the TCP stack deliver 0 len FIN pkt to the receive | |
757 | * queue, that is the only 0len pkts ever expected here, | |
758 | * and we can admit no mapping only for 0 len pkts | |
759 | */ | |
760 | if (!(TCP_SKB_CB(skb)->tcp_flags & TCPHDR_FIN)) | |
761 | WARN_ONCE(1, "0len seq %d:%d flags %x", | |
762 | TCP_SKB_CB(skb)->seq, | |
763 | TCP_SKB_CB(skb)->end_seq, | |
764 | TCP_SKB_CB(skb)->tcp_flags); | |
765 | sk_eat_skb(ssk, skb); | |
766 | return MAPPING_EMPTY; | |
767 | } | |
768 | ||
769 | if (!subflow->map_valid) | |
770 | return MAPPING_INVALID; | |
771 | ||
772 | goto validate_seq; | |
773 | } | |
774 | ||
775 | pr_debug("seq=%llu is64=%d ssn=%u data_len=%u data_fin=%d", | |
776 | mpext->data_seq, mpext->dsn64, mpext->subflow_seq, | |
777 | mpext->data_len, mpext->data_fin); | |
778 | ||
779 | data_len = mpext->data_len; | |
780 | if (data_len == 0) { | |
781 | pr_err("Infinite mapping not handled"); | |
fc518953 | 782 | MPTCP_INC_STATS(sock_net(ssk), MPTCP_MIB_INFINITEMAPRX); |
648ef4b8 MM |
783 | return MAPPING_INVALID; |
784 | } | |
785 | ||
786 | if (mpext->data_fin == 1) { | |
787 | if (data_len == 1) { | |
1a49b2c2 MM |
788 | bool updated = mptcp_update_rcv_data_fin(msk, mpext->data_seq, |
789 | mpext->dsn64); | |
43b54c6e | 790 | pr_debug("DATA_FIN with no payload seq=%llu", mpext->data_seq); |
648ef4b8 MM |
791 | if (subflow->map_valid) { |
792 | /* A DATA_FIN might arrive in a DSS | |
793 | * option before the previous mapping | |
794 | * has been fully consumed. Continue | |
795 | * handling the existing mapping. | |
796 | */ | |
797 | skb_ext_del(skb, SKB_EXT_MPTCP); | |
798 | return MAPPING_OK; | |
799 | } else { | |
ef59b195 MM |
800 | if (updated && schedule_work(&msk->work)) |
801 | sock_hold((struct sock *)msk); | |
802 | ||
648ef4b8 MM |
803 | return MAPPING_DATA_FIN; |
804 | } | |
43b54c6e | 805 | } else { |
017512a0 | 806 | u64 data_fin_seq = mpext->data_seq + data_len - 1; |
1a49b2c2 MM |
807 | |
808 | /* If mpext->data_seq is a 32-bit value, data_fin_seq | |
809 | * must also be limited to 32 bits. | |
810 | */ | |
811 | if (!mpext->dsn64) | |
812 | data_fin_seq &= GENMASK_ULL(31, 0); | |
813 | ||
814 | mptcp_update_rcv_data_fin(msk, data_fin_seq, mpext->dsn64); | |
815 | pr_debug("DATA_FIN with mapping seq=%llu dsn64=%d", | |
816 | data_fin_seq, mpext->dsn64); | |
648ef4b8 MM |
817 | } |
818 | ||
819 | /* Adjust for DATA_FIN using 1 byte of sequence space */ | |
820 | data_len--; | |
821 | } | |
822 | ||
823 | if (!mpext->dsn64) { | |
824 | map_seq = expand_seq(subflow->map_seq, subflow->map_data_len, | |
825 | mpext->data_seq); | |
826 | pr_debug("expanded seq=%llu", subflow->map_seq); | |
827 | } else { | |
828 | map_seq = mpext->data_seq; | |
829 | } | |
37198e93 | 830 | WRITE_ONCE(mptcp_sk(subflow->conn)->use_64bit_ack, !!mpext->dsn64); |
648ef4b8 MM |
831 | |
832 | if (subflow->map_valid) { | |
833 | /* Allow replacing only with an identical map */ | |
834 | if (subflow->map_seq == map_seq && | |
835 | subflow->map_subflow_seq == mpext->subflow_seq && | |
836 | subflow->map_data_len == data_len) { | |
837 | skb_ext_del(skb, SKB_EXT_MPTCP); | |
838 | return MAPPING_OK; | |
839 | } | |
840 | ||
841 | /* If this skb data are fully covered by the current mapping, | |
842 | * the new map would need caching, which is not supported | |
843 | */ | |
fc518953 FW |
844 | if (skb_is_fully_mapped(ssk, skb)) { |
845 | MPTCP_INC_STATS(sock_net(ssk), MPTCP_MIB_DSSNOMATCH); | |
648ef4b8 | 846 | return MAPPING_INVALID; |
fc518953 | 847 | } |
648ef4b8 MM |
848 | |
849 | /* will validate the next map after consuming the current one */ | |
850 | return MAPPING_OK; | |
851 | } | |
852 | ||
853 | subflow->map_seq = map_seq; | |
854 | subflow->map_subflow_seq = mpext->subflow_seq; | |
855 | subflow->map_data_len = data_len; | |
856 | subflow->map_valid = 1; | |
d22f4988 | 857 | subflow->mpc_map = mpext->mpc_map; |
648ef4b8 MM |
858 | pr_debug("new map seq=%llu subflow_seq=%u data_len=%u", |
859 | subflow->map_seq, subflow->map_subflow_seq, | |
860 | subflow->map_data_len); | |
861 | ||
862 | validate_seq: | |
863 | /* we revalidate valid mapping on new skb, because we must ensure | |
864 | * the current skb is completely covered by the available mapping | |
865 | */ | |
866 | if (!validate_mapping(ssk, skb)) | |
867 | return MAPPING_INVALID; | |
868 | ||
869 | skb_ext_del(skb, SKB_EXT_MPTCP); | |
870 | return MAPPING_OK; | |
871 | } | |
872 | ||
04e4cd4f | 873 | static void mptcp_subflow_discard_data(struct sock *ssk, struct sk_buff *skb, |
1d39cd8c | 874 | u64 limit) |
6719331c PA |
875 | { |
876 | struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); | |
04e4cd4f PA |
877 | bool fin = TCP_SKB_CB(skb)->tcp_flags & TCPHDR_FIN; |
878 | u32 incr; | |
879 | ||
880 | incr = limit >= skb->len ? skb->len + fin : limit; | |
881 | ||
882 | pr_debug("discarding=%d len=%d seq=%d", incr, skb->len, | |
883 | subflow->map_subflow_seq); | |
06242e44 | 884 | MPTCP_INC_STATS(sock_net(ssk), MPTCP_MIB_DUPDATA); |
04e4cd4f PA |
885 | tcp_sk(ssk)->copied_seq += incr; |
886 | if (!before(tcp_sk(ssk)->copied_seq, TCP_SKB_CB(skb)->end_seq)) | |
887 | sk_eat_skb(ssk, skb); | |
888 | if (mptcp_subflow_get_map_offset(subflow) >= subflow->map_data_len) | |
889 | subflow->map_valid = 0; | |
6719331c PA |
890 | } |
891 | ||
648ef4b8 MM |
892 | static bool subflow_check_data_avail(struct sock *ssk) |
893 | { | |
894 | struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); | |
895 | enum mapping_status status; | |
896 | struct mptcp_sock *msk; | |
897 | struct sk_buff *skb; | |
898 | ||
899 | pr_debug("msk=%p ssk=%p data_avail=%d skb=%p", subflow->conn, ssk, | |
900 | subflow->data_avail, skb_peek(&ssk->sk_receive_queue)); | |
47bebdf3 PA |
901 | if (!skb_peek(&ssk->sk_receive_queue)) |
902 | subflow->data_avail = 0; | |
648ef4b8 MM |
903 | if (subflow->data_avail) |
904 | return true; | |
905 | ||
648ef4b8 MM |
906 | msk = mptcp_sk(subflow->conn); |
907 | for (;;) { | |
648ef4b8 MM |
908 | u64 ack_seq; |
909 | u64 old_ack; | |
910 | ||
43b54c6e | 911 | status = get_mapping_status(ssk, msk); |
648ef4b8 MM |
912 | pr_debug("msk=%p ssk=%p status=%d", msk, ssk, status); |
913 | if (status == MAPPING_INVALID) { | |
914 | ssk->sk_err = EBADMSG; | |
915 | goto fatal; | |
916 | } | |
e1ff9e82 DC |
917 | if (status == MAPPING_DUMMY) { |
918 | __mptcp_do_fallback(msk); | |
919 | skb = skb_peek(&ssk->sk_receive_queue); | |
920 | subflow->map_valid = 1; | |
921 | subflow->map_seq = READ_ONCE(msk->ack_seq); | |
922 | subflow->map_data_len = skb->len; | |
923 | subflow->map_subflow_seq = tcp_sk(ssk)->copied_seq - | |
924 | subflow->ssn_offset; | |
6719331c | 925 | subflow->data_avail = MPTCP_SUBFLOW_DATA_AVAIL; |
e1ff9e82 DC |
926 | return true; |
927 | } | |
648ef4b8 MM |
928 | |
929 | if (status != MAPPING_OK) | |
930 | return false; | |
931 | ||
932 | skb = skb_peek(&ssk->sk_receive_queue); | |
933 | if (WARN_ON_ONCE(!skb)) | |
934 | return false; | |
935 | ||
d22f4988 CP |
936 | /* if msk lacks the remote key, this subflow must provide an |
937 | * MP_CAPABLE-based mapping | |
938 | */ | |
939 | if (unlikely(!READ_ONCE(msk->can_ack))) { | |
940 | if (!subflow->mpc_map) { | |
941 | ssk->sk_err = EBADMSG; | |
942 | goto fatal; | |
943 | } | |
944 | WRITE_ONCE(msk->remote_key, subflow->remote_key); | |
945 | WRITE_ONCE(msk->ack_seq, subflow->map_seq); | |
946 | WRITE_ONCE(msk->can_ack, true); | |
947 | } | |
948 | ||
648ef4b8 MM |
949 | old_ack = READ_ONCE(msk->ack_seq); |
950 | ack_seq = mptcp_subflow_get_mapped_dsn(subflow); | |
951 | pr_debug("msk ack_seq=%llx subflow ack_seq=%llx", old_ack, | |
952 | ack_seq); | |
47bebdf3 | 953 | if (ack_seq == old_ack) { |
6719331c PA |
954 | subflow->data_avail = MPTCP_SUBFLOW_DATA_AVAIL; |
955 | break; | |
956 | } else if (after64(ack_seq, old_ack)) { | |
957 | subflow->data_avail = MPTCP_SUBFLOW_OOO_DATA; | |
648ef4b8 | 958 | break; |
47bebdf3 | 959 | } |
648ef4b8 MM |
960 | |
961 | /* only accept in-sequence mapping. Old values are spurious | |
6719331c | 962 | * retransmission |
648ef4b8 | 963 | */ |
04e4cd4f | 964 | mptcp_subflow_discard_data(ssk, skb, old_ack - ack_seq); |
648ef4b8 MM |
965 | } |
966 | return true; | |
967 | ||
968 | fatal: | |
969 | /* fatal protocol error, close the socket */ | |
970 | /* This barrier is coupled with smp_rmb() in tcp_poll() */ | |
971 | smp_wmb(); | |
972 | ssk->sk_error_report(ssk); | |
973 | tcp_set_state(ssk, TCP_CLOSE); | |
974 | tcp_send_active_reset(ssk, GFP_ATOMIC); | |
47bebdf3 | 975 | subflow->data_avail = 0; |
648ef4b8 MM |
976 | return false; |
977 | } | |
978 | ||
979 | bool mptcp_subflow_data_available(struct sock *sk) | |
980 | { | |
981 | struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); | |
648ef4b8 MM |
982 | |
983 | /* check if current mapping is still valid */ | |
984 | if (subflow->map_valid && | |
985 | mptcp_subflow_get_map_offset(subflow) >= subflow->map_data_len) { | |
986 | subflow->map_valid = 0; | |
987 | subflow->data_avail = 0; | |
988 | ||
989 | pr_debug("Done with mapping: seq=%u data_len=%u", | |
990 | subflow->map_subflow_seq, | |
991 | subflow->map_data_len); | |
992 | } | |
993 | ||
47bebdf3 | 994 | return subflow_check_data_avail(sk); |
648ef4b8 MM |
995 | } |
996 | ||
071c8ed6 FW |
997 | /* If ssk has an mptcp parent socket, use the mptcp rcvbuf occupancy, |
998 | * not the ssk one. | |
999 | * | |
1000 | * In mptcp, rwin is about the mptcp-level connection data. | |
1001 | * | |
1002 | * Data that is still on the ssk rx queue can thus be ignored, | |
1003 | * as far as mptcp peer is concerened that data is still inflight. | |
1004 | * DSS ACK is updated when skb is moved to the mptcp rx queue. | |
1005 | */ | |
1006 | void mptcp_space(const struct sock *ssk, int *space, int *full_space) | |
1007 | { | |
1008 | const struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); | |
1009 | const struct sock *sk = subflow->conn; | |
1010 | ||
ea4ca586 | 1011 | *space = __mptcp_space(sk); |
071c8ed6 FW |
1012 | *full_space = tcp_full_space(sk); |
1013 | } | |
1014 | ||
648ef4b8 MM |
1015 | static void subflow_data_ready(struct sock *sk) |
1016 | { | |
1017 | struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); | |
8c728940 | 1018 | u16 state = 1 << inet_sk_state_load(sk); |
648ef4b8 | 1019 | struct sock *parent = subflow->conn; |
e1ff9e82 | 1020 | struct mptcp_sock *msk; |
648ef4b8 | 1021 | |
e1ff9e82 | 1022 | msk = mptcp_sk(parent); |
8c728940 | 1023 | if (state & TCPF_LISTEN) { |
e1ff9e82 | 1024 | set_bit(MPTCP_DATA_READY, &msk->flags); |
dc093db5 | 1025 | parent->sk_data_ready(parent); |
648ef4b8 MM |
1026 | return; |
1027 | } | |
1028 | ||
e1ff9e82 | 1029 | WARN_ON_ONCE(!__mptcp_check_fallback(msk) && !subflow->mp_capable && |
8c728940 | 1030 | !subflow->mp_join && !(state & TCPF_CLOSE)); |
e1ff9e82 | 1031 | |
101f6f85 | 1032 | if (mptcp_subflow_data_available(sk)) |
2e52213c | 1033 | mptcp_data_ready(parent, sk); |
648ef4b8 MM |
1034 | } |
1035 | ||
6e628cd3 | 1036 | static void subflow_write_space(struct sock *ssk) |
648ef4b8 | 1037 | { |
6e628cd3 | 1038 | /* we take action in __mptcp_clean_una() */ |
648ef4b8 MM |
1039 | } |
1040 | ||
cec37a6e PK |
1041 | static struct inet_connection_sock_af_ops * |
1042 | subflow_default_af_ops(struct sock *sk) | |
1043 | { | |
1044 | #if IS_ENABLED(CONFIG_MPTCP_IPV6) | |
1045 | if (sk->sk_family == AF_INET6) | |
1046 | return &subflow_v6_specific; | |
1047 | #endif | |
1048 | return &subflow_specific; | |
1049 | } | |
1050 | ||
cec37a6e | 1051 | #if IS_ENABLED(CONFIG_MPTCP_IPV6) |
31484d56 GU |
1052 | void mptcpv6_handle_mapped(struct sock *sk, bool mapped) |
1053 | { | |
cec37a6e PK |
1054 | struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); |
1055 | struct inet_connection_sock *icsk = inet_csk(sk); | |
1056 | struct inet_connection_sock_af_ops *target; | |
1057 | ||
1058 | target = mapped ? &subflow_v6m_specific : subflow_default_af_ops(sk); | |
1059 | ||
1060 | pr_debug("subflow=%p family=%d ops=%p target=%p mapped=%d", | |
edc7e489 | 1061 | subflow, sk->sk_family, icsk->icsk_af_ops, target, mapped); |
cec37a6e PK |
1062 | |
1063 | if (likely(icsk->icsk_af_ops == target)) | |
1064 | return; | |
1065 | ||
1066 | subflow->icsk_af_ops = icsk->icsk_af_ops; | |
1067 | icsk->icsk_af_ops = target; | |
cec37a6e | 1068 | } |
31484d56 | 1069 | #endif |
cec37a6e | 1070 | |
ec3edaa7 PK |
1071 | static void mptcp_info2sockaddr(const struct mptcp_addr_info *info, |
1072 | struct sockaddr_storage *addr) | |
1073 | { | |
1074 | memset(addr, 0, sizeof(*addr)); | |
1075 | addr->ss_family = info->family; | |
1076 | if (addr->ss_family == AF_INET) { | |
1077 | struct sockaddr_in *in_addr = (struct sockaddr_in *)addr; | |
1078 | ||
1079 | in_addr->sin_addr = info->addr; | |
1080 | in_addr->sin_port = info->port; | |
1081 | } | |
1082 | #if IS_ENABLED(CONFIG_MPTCP_IPV6) | |
1083 | else if (addr->ss_family == AF_INET6) { | |
1084 | struct sockaddr_in6 *in6_addr = (struct sockaddr_in6 *)addr; | |
1085 | ||
1086 | in6_addr->sin6_addr = info->addr6; | |
1087 | in6_addr->sin6_port = info->port; | |
1088 | } | |
1089 | #endif | |
1090 | } | |
1091 | ||
ef0da3b8 | 1092 | int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc, |
ec3edaa7 PK |
1093 | const struct mptcp_addr_info *remote) |
1094 | { | |
1095 | struct mptcp_sock *msk = mptcp_sk(sk); | |
1096 | struct mptcp_subflow_context *subflow; | |
1097 | struct sockaddr_storage addr; | |
2ff0e566 | 1098 | int remote_id = remote->id; |
6bad912b | 1099 | int local_id = loc->id; |
ec3edaa7 | 1100 | struct socket *sf; |
6bad912b | 1101 | struct sock *ssk; |
ec3edaa7 PK |
1102 | u32 remote_token; |
1103 | int addrlen; | |
1104 | int err; | |
1105 | ||
b93df08c | 1106 | if (!mptcp_is_fully_established(sk)) |
ec3edaa7 PK |
1107 | return -ENOTCONN; |
1108 | ||
1109 | err = mptcp_subflow_create_socket(sk, &sf); | |
1110 | if (err) | |
1111 | return err; | |
1112 | ||
6bad912b PA |
1113 | ssk = sf->sk; |
1114 | subflow = mptcp_subflow_ctx(ssk); | |
1115 | do { | |
1116 | get_random_bytes(&subflow->local_nonce, sizeof(u32)); | |
1117 | } while (!subflow->local_nonce); | |
1118 | ||
1119 | if (!local_id) { | |
1120 | err = mptcp_pm_get_local_id(msk, (struct sock_common *)ssk); | |
1121 | if (err < 0) | |
1122 | goto failed; | |
1123 | ||
1124 | local_id = err; | |
1125 | } | |
1126 | ||
ec3edaa7 PK |
1127 | subflow->remote_key = msk->remote_key; |
1128 | subflow->local_key = msk->local_key; | |
1129 | subflow->token = msk->token; | |
1130 | mptcp_info2sockaddr(loc, &addr); | |
1131 | ||
1132 | addrlen = sizeof(struct sockaddr_in); | |
1133 | #if IS_ENABLED(CONFIG_MPTCP_IPV6) | |
1134 | if (loc->family == AF_INET6) | |
1135 | addrlen = sizeof(struct sockaddr_in6); | |
1136 | #endif | |
ef0da3b8 | 1137 | ssk->sk_bound_dev_if = loc->ifindex; |
ec3edaa7 PK |
1138 | err = kernel_bind(sf, (struct sockaddr *)&addr, addrlen); |
1139 | if (err) | |
1140 | goto failed; | |
1141 | ||
1142 | mptcp_crypto_key_sha(subflow->remote_key, &remote_token, NULL); | |
2ff0e566 GT |
1143 | pr_debug("msk=%p remote_token=%u local_id=%d remote_id=%d", msk, |
1144 | remote_token, local_id, remote_id); | |
ec3edaa7 | 1145 | subflow->remote_token = remote_token; |
6bad912b | 1146 | subflow->local_id = local_id; |
2ff0e566 | 1147 | subflow->remote_id = remote_id; |
ec3edaa7 | 1148 | subflow->request_join = 1; |
4596a2c1 | 1149 | subflow->request_bkup = !!(loc->flags & MPTCP_PM_ADDR_FLAG_BACKUP); |
ec3edaa7 PK |
1150 | mptcp_info2sockaddr(remote, &addr); |
1151 | ||
5b950ff4 | 1152 | mptcp_add_pending_subflow(msk, subflow); |
ec3edaa7 PK |
1153 | err = kernel_connect(sf, (struct sockaddr *)&addr, addrlen, O_NONBLOCK); |
1154 | if (err && err != -EINPROGRESS) | |
5b950ff4 | 1155 | goto failed_unlink; |
ec3edaa7 | 1156 | |
ec3edaa7 PK |
1157 | return err; |
1158 | ||
5b950ff4 PA |
1159 | failed_unlink: |
1160 | spin_lock_bh(&msk->join_list_lock); | |
1161 | list_del(&subflow->node); | |
1162 | spin_unlock_bh(&msk->join_list_lock); | |
1163 | ||
ec3edaa7 | 1164 | failed: |
e16163b6 | 1165 | subflow->disposable = 1; |
ec3edaa7 PK |
1166 | sock_release(sf); |
1167 | return err; | |
1168 | } | |
1169 | ||
3764b0c5 NR |
1170 | static void mptcp_attach_cgroup(struct sock *parent, struct sock *child) |
1171 | { | |
1172 | #ifdef CONFIG_SOCK_CGROUP_DATA | |
1173 | struct sock_cgroup_data *parent_skcd = &parent->sk_cgrp_data, | |
1174 | *child_skcd = &child->sk_cgrp_data; | |
1175 | ||
1176 | /* only the additional subflows created by kworkers have to be modified */ | |
1177 | if (cgroup_id(sock_cgroup_ptr(parent_skcd)) != | |
1178 | cgroup_id(sock_cgroup_ptr(child_skcd))) { | |
1179 | #ifdef CONFIG_MEMCG | |
1180 | struct mem_cgroup *memcg = parent->sk_memcg; | |
1181 | ||
1182 | mem_cgroup_sk_free(child); | |
1183 | if (memcg && css_tryget(&memcg->css)) | |
1184 | child->sk_memcg = memcg; | |
1185 | #endif /* CONFIG_MEMCG */ | |
1186 | ||
1187 | cgroup_sk_free(child_skcd); | |
1188 | *child_skcd = *parent_skcd; | |
1189 | cgroup_sk_clone(child_skcd); | |
1190 | } | |
1191 | #endif /* CONFIG_SOCK_CGROUP_DATA */ | |
1192 | } | |
1193 | ||
2303f994 PK |
1194 | int mptcp_subflow_create_socket(struct sock *sk, struct socket **new_sock) |
1195 | { | |
1196 | struct mptcp_subflow_context *subflow; | |
1197 | struct net *net = sock_net(sk); | |
1198 | struct socket *sf; | |
1199 | int err; | |
1200 | ||
adf73410 PA |
1201 | /* un-accepted server sockets can reach here - on bad configuration |
1202 | * bail early to avoid greater trouble later | |
1203 | */ | |
1204 | if (unlikely(!sk->sk_socket)) | |
1205 | return -EINVAL; | |
1206 | ||
cec37a6e PK |
1207 | err = sock_create_kern(net, sk->sk_family, SOCK_STREAM, IPPROTO_TCP, |
1208 | &sf); | |
2303f994 PK |
1209 | if (err) |
1210 | return err; | |
1211 | ||
1212 | lock_sock(sf->sk); | |
1213 | ||
3764b0c5 NR |
1214 | /* the newly created socket has to be in the same cgroup as its parent */ |
1215 | mptcp_attach_cgroup(sk, sf->sk); | |
1216 | ||
2303f994 PK |
1217 | /* kernel sockets do not by default acquire net ref, but TCP timer |
1218 | * needs it. | |
1219 | */ | |
1220 | sf->sk->sk_net_refcnt = 1; | |
1221 | get_net(net); | |
f6f7d8cf | 1222 | #ifdef CONFIG_PROC_FS |
2303f994 | 1223 | this_cpu_add(*net->core.sock_inuse, 1); |
f6f7d8cf | 1224 | #endif |
2303f994 PK |
1225 | err = tcp_set_ulp(sf->sk, "mptcp"); |
1226 | release_sock(sf->sk); | |
1227 | ||
b8ad540d WY |
1228 | if (err) { |
1229 | sock_release(sf); | |
2303f994 | 1230 | return err; |
b8ad540d | 1231 | } |
2303f994 | 1232 | |
7d14b0d2 PA |
1233 | /* the newly created socket really belongs to the owning MPTCP master |
1234 | * socket, even if for additional subflows the allocation is performed | |
1235 | * by a kernel workqueue. Adjust inode references, so that the | |
1236 | * procfs/diag interaces really show this one belonging to the correct | |
1237 | * user. | |
1238 | */ | |
1239 | SOCK_INODE(sf)->i_ino = SOCK_INODE(sk->sk_socket)->i_ino; | |
1240 | SOCK_INODE(sf)->i_uid = SOCK_INODE(sk->sk_socket)->i_uid; | |
1241 | SOCK_INODE(sf)->i_gid = SOCK_INODE(sk->sk_socket)->i_gid; | |
1242 | ||
2303f994 PK |
1243 | subflow = mptcp_subflow_ctx(sf->sk); |
1244 | pr_debug("subflow=%p", subflow); | |
1245 | ||
1246 | *new_sock = sf; | |
79c0949e | 1247 | sock_hold(sk); |
2303f994 PK |
1248 | subflow->conn = sk; |
1249 | ||
1250 | return 0; | |
1251 | } | |
1252 | ||
1253 | static struct mptcp_subflow_context *subflow_create_ctx(struct sock *sk, | |
1254 | gfp_t priority) | |
1255 | { | |
1256 | struct inet_connection_sock *icsk = inet_csk(sk); | |
1257 | struct mptcp_subflow_context *ctx; | |
1258 | ||
1259 | ctx = kzalloc(sizeof(*ctx), priority); | |
1260 | if (!ctx) | |
1261 | return NULL; | |
1262 | ||
1263 | rcu_assign_pointer(icsk->icsk_ulp_data, ctx); | |
cec37a6e | 1264 | INIT_LIST_HEAD(&ctx->node); |
2303f994 PK |
1265 | |
1266 | pr_debug("subflow=%p", ctx); | |
1267 | ||
1268 | ctx->tcp_sock = sk; | |
1269 | ||
1270 | return ctx; | |
1271 | } | |
1272 | ||
648ef4b8 MM |
1273 | static void __subflow_state_change(struct sock *sk) |
1274 | { | |
1275 | struct socket_wq *wq; | |
1276 | ||
1277 | rcu_read_lock(); | |
1278 | wq = rcu_dereference(sk->sk_wq); | |
1279 | if (skwq_has_sleeper(wq)) | |
1280 | wake_up_interruptible_all(&wq->wait); | |
1281 | rcu_read_unlock(); | |
1282 | } | |
1283 | ||
1284 | static bool subflow_is_done(const struct sock *sk) | |
1285 | { | |
1286 | return sk->sk_shutdown & RCV_SHUTDOWN || sk->sk_state == TCP_CLOSE; | |
1287 | } | |
1288 | ||
1289 | static void subflow_state_change(struct sock *sk) | |
1290 | { | |
1291 | struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); | |
dc093db5 | 1292 | struct sock *parent = subflow->conn; |
648ef4b8 MM |
1293 | |
1294 | __subflow_state_change(sk); | |
1295 | ||
8fd73804 DC |
1296 | if (subflow_simultaneous_connect(sk)) { |
1297 | mptcp_do_fallback(sk); | |
a6b118fe | 1298 | mptcp_rcv_space_init(mptcp_sk(parent), sk); |
8fd73804 DC |
1299 | pr_fallback(mptcp_sk(parent)); |
1300 | subflow->conn_finished = 1; | |
1301 | if (inet_sk_state_load(parent) == TCP_SYN_SENT) { | |
1302 | inet_sk_state_store(parent, TCP_ESTABLISHED); | |
1303 | parent->sk_state_change(parent); | |
1304 | } | |
1305 | } | |
1306 | ||
648ef4b8 MM |
1307 | /* as recvmsg() does not acquire the subflow socket for ssk selection |
1308 | * a fin packet carrying a DSS can be unnoticed if we don't trigger | |
1309 | * the data available machinery here. | |
1310 | */ | |
e1ff9e82 | 1311 | if (mptcp_subflow_data_available(sk)) |
2e52213c | 1312 | mptcp_data_ready(parent, sk); |
648ef4b8 | 1313 | |
067a0b3d | 1314 | if (__mptcp_check_fallback(mptcp_sk(parent)) && |
648ef4b8 MM |
1315 | !subflow->rx_eof && subflow_is_done(sk)) { |
1316 | subflow->rx_eof = 1; | |
59832e24 | 1317 | mptcp_subflow_eof(parent); |
648ef4b8 MM |
1318 | } |
1319 | } | |
1320 | ||
2303f994 PK |
1321 | static int subflow_ulp_init(struct sock *sk) |
1322 | { | |
cec37a6e | 1323 | struct inet_connection_sock *icsk = inet_csk(sk); |
2303f994 PK |
1324 | struct mptcp_subflow_context *ctx; |
1325 | struct tcp_sock *tp = tcp_sk(sk); | |
1326 | int err = 0; | |
1327 | ||
1328 | /* disallow attaching ULP to a socket unless it has been | |
1329 | * created with sock_create_kern() | |
1330 | */ | |
1331 | if (!sk->sk_kern_sock) { | |
1332 | err = -EOPNOTSUPP; | |
1333 | goto out; | |
1334 | } | |
1335 | ||
1336 | ctx = subflow_create_ctx(sk, GFP_KERNEL); | |
1337 | if (!ctx) { | |
1338 | err = -ENOMEM; | |
1339 | goto out; | |
1340 | } | |
1341 | ||
1342 | pr_debug("subflow=%p, family=%d", ctx, sk->sk_family); | |
1343 | ||
1344 | tp->is_mptcp = 1; | |
cec37a6e PK |
1345 | ctx->icsk_af_ops = icsk->icsk_af_ops; |
1346 | icsk->icsk_af_ops = subflow_default_af_ops(sk); | |
648ef4b8 MM |
1347 | ctx->tcp_data_ready = sk->sk_data_ready; |
1348 | ctx->tcp_state_change = sk->sk_state_change; | |
1349 | ctx->tcp_write_space = sk->sk_write_space; | |
1350 | sk->sk_data_ready = subflow_data_ready; | |
1351 | sk->sk_write_space = subflow_write_space; | |
1352 | sk->sk_state_change = subflow_state_change; | |
2303f994 PK |
1353 | out: |
1354 | return err; | |
1355 | } | |
1356 | ||
e16163b6 | 1357 | static void subflow_ulp_release(struct sock *ssk) |
2303f994 | 1358 | { |
e16163b6 PA |
1359 | struct mptcp_subflow_context *ctx = mptcp_subflow_ctx(ssk); |
1360 | bool release = true; | |
1361 | struct sock *sk; | |
2303f994 PK |
1362 | |
1363 | if (!ctx) | |
1364 | return; | |
1365 | ||
e16163b6 PA |
1366 | sk = ctx->conn; |
1367 | if (sk) { | |
1368 | /* if the msk has been orphaned, keep the ctx | |
0597d0f8 PA |
1369 | * alive, will be freed by __mptcp_close_ssk(), |
1370 | * when the subflow is still unaccepted | |
e16163b6 | 1371 | */ |
0597d0f8 | 1372 | release = ctx->disposable || list_empty(&ctx->node); |
e16163b6 PA |
1373 | sock_put(sk); |
1374 | } | |
79c0949e | 1375 | |
e16163b6 PA |
1376 | if (release) |
1377 | kfree_rcu(ctx, rcu); | |
2303f994 PK |
1378 | } |
1379 | ||
cec37a6e PK |
1380 | static void subflow_ulp_clone(const struct request_sock *req, |
1381 | struct sock *newsk, | |
1382 | const gfp_t priority) | |
1383 | { | |
1384 | struct mptcp_subflow_request_sock *subflow_req = mptcp_subflow_rsk(req); | |
1385 | struct mptcp_subflow_context *old_ctx = mptcp_subflow_ctx(newsk); | |
1386 | struct mptcp_subflow_context *new_ctx; | |
1387 | ||
f296234c PK |
1388 | if (!tcp_rsk(req)->is_mptcp || |
1389 | (!subflow_req->mp_capable && !subflow_req->mp_join)) { | |
648ef4b8 | 1390 | subflow_ulp_fallback(newsk, old_ctx); |
cec37a6e PK |
1391 | return; |
1392 | } | |
1393 | ||
1394 | new_ctx = subflow_create_ctx(newsk, priority); | |
edc7e489 | 1395 | if (!new_ctx) { |
648ef4b8 | 1396 | subflow_ulp_fallback(newsk, old_ctx); |
cec37a6e PK |
1397 | return; |
1398 | } | |
1399 | ||
1400 | new_ctx->conn_finished = 1; | |
1401 | new_ctx->icsk_af_ops = old_ctx->icsk_af_ops; | |
648ef4b8 MM |
1402 | new_ctx->tcp_data_ready = old_ctx->tcp_data_ready; |
1403 | new_ctx->tcp_state_change = old_ctx->tcp_state_change; | |
1404 | new_ctx->tcp_write_space = old_ctx->tcp_write_space; | |
58b09919 PA |
1405 | new_ctx->rel_write_seq = 1; |
1406 | new_ctx->tcp_sock = newsk; | |
1407 | ||
f296234c PK |
1408 | if (subflow_req->mp_capable) { |
1409 | /* see comments in subflow_syn_recv_sock(), MPTCP connection | |
1410 | * is fully established only after we receive the remote key | |
1411 | */ | |
1412 | new_ctx->mp_capable = 1; | |
f296234c PK |
1413 | new_ctx->local_key = subflow_req->local_key; |
1414 | new_ctx->token = subflow_req->token; | |
1415 | new_ctx->ssn_offset = subflow_req->ssn_offset; | |
1416 | new_ctx->idsn = subflow_req->idsn; | |
1417 | } else if (subflow_req->mp_join) { | |
ec3edaa7 | 1418 | new_ctx->ssn_offset = subflow_req->ssn_offset; |
f296234c PK |
1419 | new_ctx->mp_join = 1; |
1420 | new_ctx->fully_established = 1; | |
1421 | new_ctx->backup = subflow_req->backup; | |
1422 | new_ctx->local_id = subflow_req->local_id; | |
2ff0e566 | 1423 | new_ctx->remote_id = subflow_req->remote_id; |
f296234c PK |
1424 | new_ctx->token = subflow_req->token; |
1425 | new_ctx->thmac = subflow_req->thmac; | |
1426 | } | |
cec37a6e PK |
1427 | } |
1428 | ||
2303f994 PK |
1429 | static struct tcp_ulp_ops subflow_ulp_ops __read_mostly = { |
1430 | .name = "mptcp", | |
1431 | .owner = THIS_MODULE, | |
1432 | .init = subflow_ulp_init, | |
1433 | .release = subflow_ulp_release, | |
cec37a6e | 1434 | .clone = subflow_ulp_clone, |
2303f994 PK |
1435 | }; |
1436 | ||
cec37a6e PK |
1437 | static int subflow_ops_init(struct request_sock_ops *subflow_ops) |
1438 | { | |
1439 | subflow_ops->obj_size = sizeof(struct mptcp_subflow_request_sock); | |
1440 | subflow_ops->slab_name = "request_sock_subflow"; | |
1441 | ||
1442 | subflow_ops->slab = kmem_cache_create(subflow_ops->slab_name, | |
1443 | subflow_ops->obj_size, 0, | |
1444 | SLAB_ACCOUNT | | |
1445 | SLAB_TYPESAFE_BY_RCU, | |
1446 | NULL); | |
1447 | if (!subflow_ops->slab) | |
1448 | return -ENOMEM; | |
1449 | ||
79c0949e PK |
1450 | subflow_ops->destructor = subflow_req_destructor; |
1451 | ||
cec37a6e PK |
1452 | return 0; |
1453 | } | |
1454 | ||
d39dceca | 1455 | void __init mptcp_subflow_init(void) |
2303f994 | 1456 | { |
08b8d080 FW |
1457 | mptcp_subflow_request_sock_ops = tcp_request_sock_ops; |
1458 | if (subflow_ops_init(&mptcp_subflow_request_sock_ops) != 0) | |
cec37a6e PK |
1459 | panic("MPTCP: failed to init subflow request sock ops\n"); |
1460 | ||
1461 | subflow_request_sock_ipv4_ops = tcp_request_sock_ipv4_ops; | |
7ea851d1 | 1462 | subflow_request_sock_ipv4_ops.route_req = subflow_v4_route_req; |
cec37a6e PK |
1463 | |
1464 | subflow_specific = ipv4_specific; | |
1465 | subflow_specific.conn_request = subflow_v4_conn_request; | |
1466 | subflow_specific.syn_recv_sock = subflow_syn_recv_sock; | |
1467 | subflow_specific.sk_rx_dst_set = subflow_finish_connect; | |
1468 | ||
1469 | #if IS_ENABLED(CONFIG_MPTCP_IPV6) | |
1470 | subflow_request_sock_ipv6_ops = tcp_request_sock_ipv6_ops; | |
7ea851d1 | 1471 | subflow_request_sock_ipv6_ops.route_req = subflow_v6_route_req; |
cec37a6e PK |
1472 | |
1473 | subflow_v6_specific = ipv6_specific; | |
1474 | subflow_v6_specific.conn_request = subflow_v6_conn_request; | |
1475 | subflow_v6_specific.syn_recv_sock = subflow_syn_recv_sock; | |
1476 | subflow_v6_specific.sk_rx_dst_set = subflow_finish_connect; | |
1477 | ||
1478 | subflow_v6m_specific = subflow_v6_specific; | |
1479 | subflow_v6m_specific.queue_xmit = ipv4_specific.queue_xmit; | |
1480 | subflow_v6m_specific.send_check = ipv4_specific.send_check; | |
1481 | subflow_v6m_specific.net_header_len = ipv4_specific.net_header_len; | |
1482 | subflow_v6m_specific.mtu_reduced = ipv4_specific.mtu_reduced; | |
1483 | subflow_v6m_specific.net_frag_header_len = 0; | |
1484 | #endif | |
1485 | ||
5147dfb5 DC |
1486 | mptcp_diag_subflow_init(&subflow_ulp_ops); |
1487 | ||
2303f994 PK |
1488 | if (tcp_register_ulp(&subflow_ulp_ops) != 0) |
1489 | panic("MPTCP: failed to register subflows to ULP\n"); | |
1490 | } |