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> |
a24d22b2 | 13 | #include <crypto/sha2.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> | |
b19bc294 | 21 | #include <net/transp_v6.h> |
cec37a6e | 22 | #endif |
2303f994 | 23 | #include <net/mptcp.h> |
4596a2c1 | 24 | #include <uapi/linux/mptcp.h> |
2303f994 | 25 | #include "protocol.h" |
fc518953 FW |
26 | #include "mib.h" |
27 | ||
0918e34b GT |
28 | #include <trace/events/mptcp.h> |
29 | ||
b19bc294 PA |
30 | static void mptcp_subflow_ops_undo_override(struct sock *ssk); |
31 | ||
fc518953 FW |
32 | static void SUBFLOW_REQ_INC_STATS(struct request_sock *req, |
33 | enum linux_mptcp_mib_field field) | |
34 | { | |
35 | MPTCP_INC_STATS(sock_net(req_to_sk(req)), field); | |
36 | } | |
2303f994 | 37 | |
79c0949e PK |
38 | static void subflow_req_destructor(struct request_sock *req) |
39 | { | |
40 | struct mptcp_subflow_request_sock *subflow_req = mptcp_subflow_rsk(req); | |
41 | ||
42 | pr_debug("subflow_req=%p", subflow_req); | |
43 | ||
8fd4de12 PA |
44 | if (subflow_req->msk) |
45 | sock_put((struct sock *)subflow_req->msk); | |
46 | ||
2c5ebd00 | 47 | mptcp_token_destroy_request(req); |
79c0949e PK |
48 | tcp_request_sock_ops.destructor(req); |
49 | } | |
50 | ||
f296234c PK |
51 | static void subflow_generate_hmac(u64 key1, u64 key2, u32 nonce1, u32 nonce2, |
52 | void *hmac) | |
53 | { | |
54 | u8 msg[8]; | |
55 | ||
56 | put_unaligned_be32(nonce1, &msg[0]); | |
57 | put_unaligned_be32(nonce2, &msg[4]); | |
58 | ||
59 | mptcp_crypto_hmac_sha(key1, key2, msg, 8, hmac); | |
60 | } | |
61 | ||
4cf8b7e4 PA |
62 | static bool mptcp_can_accept_new_subflow(const struct mptcp_sock *msk) |
63 | { | |
64 | return mptcp_is_fully_established((void *)msk) && | |
4d25247d KM |
65 | ((mptcp_pm_is_userspace(msk) && |
66 | mptcp_userspace_pm_active(msk)) || | |
67 | READ_ONCE(msk->pm.accept_subflow)); | |
4cf8b7e4 PA |
68 | } |
69 | ||
f296234c | 70 | /* validate received token and create truncated hmac and nonce for SYN-ACK */ |
ec20e143 GT |
71 | static void subflow_req_create_thmac(struct mptcp_subflow_request_sock *subflow_req) |
72 | { | |
73 | struct mptcp_sock *msk = subflow_req->msk; | |
74 | u8 hmac[SHA256_DIGEST_SIZE]; | |
75 | ||
76 | get_random_bytes(&subflow_req->local_nonce, sizeof(u32)); | |
77 | ||
78 | subflow_generate_hmac(msk->local_key, msk->remote_key, | |
79 | subflow_req->local_nonce, | |
80 | subflow_req->remote_nonce, hmac); | |
81 | ||
82 | subflow_req->thmac = get_unaligned_be64(hmac); | |
83 | } | |
84 | ||
b5e2e42f | 85 | static struct mptcp_sock *subflow_token_join_request(struct request_sock *req) |
f296234c PK |
86 | { |
87 | struct mptcp_subflow_request_sock *subflow_req = mptcp_subflow_rsk(req); | |
f296234c PK |
88 | struct mptcp_sock *msk; |
89 | int local_id; | |
90 | ||
ea1300b9 | 91 | msk = mptcp_token_get_sock(sock_net(req_to_sk(req)), subflow_req->token); |
f296234c | 92 | if (!msk) { |
fc518953 | 93 | SUBFLOW_REQ_INC_STATS(req, MPTCP_MIB_JOINNOTOKEN); |
8fd4de12 | 94 | return NULL; |
f296234c PK |
95 | } |
96 | ||
97 | local_id = mptcp_pm_get_local_id(msk, (struct sock_common *)req); | |
98 | if (local_id < 0) { | |
99 | sock_put((struct sock *)msk); | |
8fd4de12 | 100 | return NULL; |
f296234c PK |
101 | } |
102 | subflow_req->local_id = local_id; | |
103 | ||
8fd4de12 | 104 | return msk; |
f296234c PK |
105 | } |
106 | ||
d8b59efa | 107 | static void subflow_init_req(struct request_sock *req, const struct sock *sk_listener) |
cec37a6e | 108 | { |
cec37a6e | 109 | struct mptcp_subflow_request_sock *subflow_req = mptcp_subflow_rsk(req); |
cec37a6e PK |
110 | |
111 | subflow_req->mp_capable = 0; | |
f296234c | 112 | subflow_req->mp_join = 0; |
06fe1719 | 113 | subflow_req->csum_reqd = mptcp_is_checksum_enabled(sock_net(sk_listener)); |
bab6b88e | 114 | subflow_req->allow_join_id0 = mptcp_allow_join_id0(sock_net(sk_listener)); |
8fd4de12 | 115 | subflow_req->msk = NULL; |
2c5ebd00 | 116 | mptcp_token_init_request(req); |
78d8b7bc FW |
117 | } |
118 | ||
5bc56388 GT |
119 | static bool subflow_use_different_sport(struct mptcp_sock *msk, const struct sock *sk) |
120 | { | |
121 | return inet_sk(sk)->inet_sport != inet_sk((struct sock *)msk)->inet_sport; | |
122 | } | |
123 | ||
dc87efdb FW |
124 | static void subflow_add_reset_reason(struct sk_buff *skb, u8 reason) |
125 | { | |
126 | struct mptcp_ext *mpext = skb_ext_add(skb, SKB_EXT_MPTCP); | |
127 | ||
128 | if (mpext) { | |
129 | memset(mpext, 0, sizeof(*mpext)); | |
130 | mpext->reset_reason = reason; | |
131 | } | |
132 | } | |
133 | ||
3ecfbe3e FW |
134 | /* Init mptcp request socket. |
135 | * | |
136 | * Returns an error code if a JOIN has failed and a TCP reset | |
137 | * should be sent. | |
138 | */ | |
d8b59efa PA |
139 | static int subflow_check_req(struct request_sock *req, |
140 | const struct sock *sk_listener, | |
141 | struct sk_buff *skb) | |
78d8b7bc FW |
142 | { |
143 | struct mptcp_subflow_context *listener = mptcp_subflow_ctx(sk_listener); | |
144 | struct mptcp_subflow_request_sock *subflow_req = mptcp_subflow_rsk(req); | |
145 | struct mptcp_options_received mp_opt; | |
74c7dfbe | 146 | bool opt_mp_capable, opt_mp_join; |
78d8b7bc FW |
147 | |
148 | pr_debug("subflow_req=%p, listener=%p", subflow_req, listener); | |
149 | ||
d8b59efa PA |
150 | #ifdef CONFIG_TCP_MD5SIG |
151 | /* no MPTCP if MD5SIG is enabled on this socket or we may run out of | |
152 | * TCP option space. | |
153 | */ | |
154 | if (rcu_access_pointer(tcp_sk(sk_listener)->md5sig_info)) | |
155 | return -EINVAL; | |
156 | #endif | |
78d8b7bc | 157 | |
0799e21b | 158 | mptcp_get_options(skb, &mp_opt); |
78d8b7bc | 159 | |
74c7dfbe PA |
160 | opt_mp_capable = !!(mp_opt.suboptions & OPTIONS_MPTCP_MPC); |
161 | opt_mp_join = !!(mp_opt.suboptions & OPTIONS_MPTCP_MPJ); | |
162 | if (opt_mp_capable) { | |
fc518953 FW |
163 | SUBFLOW_REQ_INC_STATS(req, MPTCP_MIB_MPCAPABLEPASSIVE); |
164 | ||
74c7dfbe | 165 | if (opt_mp_join) |
3ecfbe3e | 166 | return 0; |
74c7dfbe | 167 | } else if (opt_mp_join) { |
fc518953 FW |
168 | SUBFLOW_REQ_INC_STATS(req, MPTCP_MIB_JOINSYNRX); |
169 | } | |
f296234c | 170 | |
74c7dfbe | 171 | if (opt_mp_capable && listener->request_mptcp) { |
c68a0cd1 | 172 | int err, retries = MPTCP_TOKEN_MAX_RETRIES; |
535fb815 | 173 | |
c83a47e5 | 174 | subflow_req->ssn_offset = TCP_SKB_CB(skb)->seq; |
535fb815 FW |
175 | again: |
176 | do { | |
177 | get_random_bytes(&subflow_req->local_key, sizeof(subflow_req->local_key)); | |
178 | } while (subflow_req->local_key == 0); | |
79c0949e | 179 | |
c83a47e5 FW |
180 | if (unlikely(req->syncookie)) { |
181 | mptcp_crypto_key_sha(subflow_req->local_key, | |
182 | &subflow_req->token, | |
183 | &subflow_req->idsn); | |
184 | if (mptcp_token_exists(subflow_req->token)) { | |
185 | if (retries-- > 0) | |
186 | goto again; | |
a16195e3 | 187 | SUBFLOW_REQ_INC_STATS(req, MPTCP_MIB_TOKENFALLBACKINIT); |
c83a47e5 FW |
188 | } else { |
189 | subflow_req->mp_capable = 1; | |
190 | } | |
3ecfbe3e | 191 | return 0; |
c83a47e5 FW |
192 | } |
193 | ||
79c0949e PK |
194 | err = mptcp_token_new_request(req); |
195 | if (err == 0) | |
196 | subflow_req->mp_capable = 1; | |
535fb815 FW |
197 | else if (retries-- > 0) |
198 | goto again; | |
a16195e3 PA |
199 | else |
200 | SUBFLOW_REQ_INC_STATS(req, MPTCP_MIB_TOKENFALLBACKINIT); | |
79c0949e | 201 | |
74c7dfbe | 202 | } else if (opt_mp_join && listener->request_mptcp) { |
ec3edaa7 | 203 | subflow_req->ssn_offset = TCP_SKB_CB(skb)->seq; |
f296234c | 204 | subflow_req->mp_join = 1; |
cfde141e PA |
205 | subflow_req->backup = mp_opt.backup; |
206 | subflow_req->remote_id = mp_opt.join_id; | |
207 | subflow_req->token = mp_opt.token; | |
208 | subflow_req->remote_nonce = mp_opt.nonce; | |
b5e2e42f | 209 | subflow_req->msk = subflow_token_join_request(req); |
9466a1cc | 210 | |
3ecfbe3e | 211 | /* Can't fall back to TCP in this case. */ |
dc87efdb FW |
212 | if (!subflow_req->msk) { |
213 | subflow_add_reset_reason(skb, MPTCP_RST_EMPTCP); | |
3ecfbe3e | 214 | return -EPERM; |
dc87efdb | 215 | } |
3ecfbe3e | 216 | |
5bc56388 GT |
217 | if (subflow_use_different_sport(subflow_req->msk, sk_listener)) { |
218 | pr_debug("syn inet_sport=%d %d", | |
219 | ntohs(inet_sk(sk_listener)->inet_sport), | |
220 | ntohs(inet_sk((struct sock *)subflow_req->msk)->inet_sport)); | |
221 | if (!mptcp_pm_sport_in_anno_list(subflow_req->msk, sk_listener)) { | |
2fbdd9ea | 222 | SUBFLOW_REQ_INC_STATS(req, MPTCP_MIB_MISMATCHPORTSYNRX); |
5bc56388 GT |
223 | return -EPERM; |
224 | } | |
2fbdd9ea | 225 | SUBFLOW_REQ_INC_STATS(req, MPTCP_MIB_JOINPORTSYNRX); |
5bc56388 GT |
226 | } |
227 | ||
ec20e143 GT |
228 | subflow_req_create_thmac(subflow_req); |
229 | ||
3ecfbe3e | 230 | if (unlikely(req->syncookie)) { |
9466a1cc FW |
231 | if (mptcp_can_accept_new_subflow(subflow_req->msk)) |
232 | subflow_init_req_cookie_join_save(subflow_req, skb); | |
8547ea5f JW |
233 | else |
234 | return -EPERM; | |
9466a1cc FW |
235 | } |
236 | ||
8fd4de12 PA |
237 | pr_debug("token=%u, remote_nonce=%u msk=%p", subflow_req->token, |
238 | subflow_req->remote_nonce, subflow_req->msk); | |
cec37a6e | 239 | } |
3ecfbe3e FW |
240 | |
241 | return 0; | |
cec37a6e PK |
242 | } |
243 | ||
c83a47e5 FW |
244 | int mptcp_subflow_init_cookie_req(struct request_sock *req, |
245 | const struct sock *sk_listener, | |
246 | struct sk_buff *skb) | |
247 | { | |
248 | struct mptcp_subflow_context *listener = mptcp_subflow_ctx(sk_listener); | |
249 | struct mptcp_subflow_request_sock *subflow_req = mptcp_subflow_rsk(req); | |
250 | struct mptcp_options_received mp_opt; | |
74c7dfbe | 251 | bool opt_mp_capable, opt_mp_join; |
c83a47e5 FW |
252 | int err; |
253 | ||
d8b59efa | 254 | subflow_init_req(req, sk_listener); |
0799e21b | 255 | mptcp_get_options(skb, &mp_opt); |
c83a47e5 | 256 | |
74c7dfbe PA |
257 | opt_mp_capable = !!(mp_opt.suboptions & OPTIONS_MPTCP_MPC); |
258 | opt_mp_join = !!(mp_opt.suboptions & OPTIONS_MPTCP_MPJ); | |
259 | if (opt_mp_capable && opt_mp_join) | |
c83a47e5 FW |
260 | return -EINVAL; |
261 | ||
74c7dfbe | 262 | if (opt_mp_capable && listener->request_mptcp) { |
c83a47e5 FW |
263 | if (mp_opt.sndr_key == 0) |
264 | return -EINVAL; | |
265 | ||
266 | subflow_req->local_key = mp_opt.rcvr_key; | |
267 | err = mptcp_token_new_request(req); | |
268 | if (err) | |
269 | return err; | |
270 | ||
271 | subflow_req->mp_capable = 1; | |
272 | subflow_req->ssn_offset = TCP_SKB_CB(skb)->seq - 1; | |
74c7dfbe | 273 | } else if (opt_mp_join && listener->request_mptcp) { |
9466a1cc FW |
274 | if (!mptcp_token_join_cookie_init_state(subflow_req, skb)) |
275 | return -EINVAL; | |
276 | ||
8547ea5f | 277 | subflow_req->mp_join = 1; |
9466a1cc | 278 | subflow_req->ssn_offset = TCP_SKB_CB(skb)->seq - 1; |
c83a47e5 FW |
279 | } |
280 | ||
281 | return 0; | |
282 | } | |
283 | EXPORT_SYMBOL_GPL(mptcp_subflow_init_cookie_req); | |
284 | ||
7ea851d1 FW |
285 | static struct dst_entry *subflow_v4_route_req(const struct sock *sk, |
286 | struct sk_buff *skb, | |
287 | struct flowi *fl, | |
288 | struct request_sock *req) | |
cec37a6e | 289 | { |
7ea851d1 | 290 | struct dst_entry *dst; |
3ecfbe3e | 291 | int err; |
7ea851d1 | 292 | |
cec37a6e | 293 | tcp_rsk(req)->is_mptcp = 1; |
d8b59efa | 294 | subflow_init_req(req, sk); |
cec37a6e | 295 | |
7ea851d1 FW |
296 | dst = tcp_request_sock_ipv4_ops.route_req(sk, skb, fl, req); |
297 | if (!dst) | |
298 | return NULL; | |
cec37a6e | 299 | |
d8b59efa | 300 | err = subflow_check_req(req, sk, skb); |
3ecfbe3e FW |
301 | if (err == 0) |
302 | return dst; | |
cec37a6e | 303 | |
3ecfbe3e FW |
304 | dst_release(dst); |
305 | if (!req->syncookie) | |
306 | tcp_request_sock_ops.send_reset(sk, skb); | |
307 | return NULL; | |
cec37a6e PK |
308 | } |
309 | ||
310 | #if IS_ENABLED(CONFIG_MPTCP_IPV6) | |
7ea851d1 FW |
311 | static struct dst_entry *subflow_v6_route_req(const struct sock *sk, |
312 | struct sk_buff *skb, | |
313 | struct flowi *fl, | |
314 | struct request_sock *req) | |
cec37a6e | 315 | { |
7ea851d1 | 316 | struct dst_entry *dst; |
3ecfbe3e | 317 | int err; |
7ea851d1 | 318 | |
cec37a6e | 319 | tcp_rsk(req)->is_mptcp = 1; |
d8b59efa | 320 | subflow_init_req(req, sk); |
cec37a6e | 321 | |
7ea851d1 FW |
322 | dst = tcp_request_sock_ipv6_ops.route_req(sk, skb, fl, req); |
323 | if (!dst) | |
324 | return NULL; | |
cec37a6e | 325 | |
d8b59efa | 326 | err = subflow_check_req(req, sk, skb); |
3ecfbe3e FW |
327 | if (err == 0) |
328 | return dst; | |
329 | ||
330 | dst_release(dst); | |
331 | if (!req->syncookie) | |
332 | tcp6_request_sock_ops.send_reset(sk, skb); | |
333 | return NULL; | |
cec37a6e PK |
334 | } |
335 | #endif | |
336 | ||
ec3edaa7 PK |
337 | /* validate received truncated hmac and create hmac for third ACK */ |
338 | static bool subflow_thmac_valid(struct mptcp_subflow_context *subflow) | |
339 | { | |
bd697222 | 340 | u8 hmac[SHA256_DIGEST_SIZE]; |
ec3edaa7 PK |
341 | u64 thmac; |
342 | ||
343 | subflow_generate_hmac(subflow->remote_key, subflow->local_key, | |
344 | subflow->remote_nonce, subflow->local_nonce, | |
345 | hmac); | |
346 | ||
347 | thmac = get_unaligned_be64(hmac); | |
348 | pr_debug("subflow=%p, token=%u, thmac=%llu, subflow->thmac=%llu\n", | |
742e2f36 | 349 | subflow, subflow->token, thmac, subflow->thmac); |
ec3edaa7 PK |
350 | |
351 | return thmac == subflow->thmac; | |
352 | } | |
353 | ||
d5824847 PA |
354 | void mptcp_subflow_reset(struct sock *ssk) |
355 | { | |
0e4f35d7 PA |
356 | struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); |
357 | struct sock *sk = subflow->conn; | |
358 | ||
ab82e996 FW |
359 | /* must hold: tcp_done() could drop last reference on parent */ |
360 | sock_hold(sk); | |
361 | ||
d5824847 PA |
362 | tcp_set_state(ssk, TCP_CLOSE); |
363 | tcp_send_active_reset(ssk, GFP_ATOMIC); | |
364 | tcp_done(ssk); | |
0e4f35d7 PA |
365 | if (!test_and_set_bit(MPTCP_WORK_CLOSE_SUBFLOW, &mptcp_sk(sk)->flags) && |
366 | schedule_work(&mptcp_sk(sk)->work)) | |
ab82e996 FW |
367 | return; /* worker will put sk for us */ |
368 | ||
369 | sock_put(sk); | |
d5824847 PA |
370 | } |
371 | ||
5bc56388 GT |
372 | static bool subflow_use_different_dport(struct mptcp_sock *msk, const struct sock *sk) |
373 | { | |
374 | return inet_sk(sk)->inet_dport != inet_sk((struct sock *)msk)->inet_dport; | |
375 | } | |
376 | ||
490274b4 PA |
377 | void __mptcp_set_connected(struct sock *sk) |
378 | { | |
379 | if (sk->sk_state == TCP_SYN_SENT) { | |
380 | inet_sk_state_store(sk, TCP_ESTABLISHED); | |
381 | sk->sk_state_change(sk); | |
382 | } | |
383 | } | |
384 | ||
385 | static void mptcp_set_connected(struct sock *sk) | |
386 | { | |
387 | mptcp_data_lock(sk); | |
388 | if (!sock_owned_by_user(sk)) | |
389 | __mptcp_set_connected(sk); | |
390 | else | |
e9d09bac | 391 | __set_bit(MPTCP_CONNECTED, &mptcp_sk(sk)->cb_flags); |
490274b4 PA |
392 | mptcp_data_unlock(sk); |
393 | } | |
394 | ||
cec37a6e PK |
395 | static void subflow_finish_connect(struct sock *sk, const struct sk_buff *skb) |
396 | { | |
397 | struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); | |
cfde141e | 398 | struct mptcp_options_received mp_opt; |
c3c123d1 | 399 | struct sock *parent = subflow->conn; |
cec37a6e PK |
400 | |
401 | subflow->icsk_af_ops->sk_rx_dst_set(sk, skb); | |
402 | ||
263e1201 PA |
403 | /* be sure no special action on any packet other than syn-ack */ |
404 | if (subflow->conn_finished) | |
405 | return; | |
406 | ||
5cf92bba | 407 | mptcp_propagate_sndbuf(parent, sk); |
b0977bb2 | 408 | subflow->rel_write_seq = 1; |
263e1201 | 409 | subflow->conn_finished = 1; |
e1ff9e82 DC |
410 | subflow->ssn_offset = TCP_SKB_CB(skb)->seq; |
411 | pr_debug("subflow=%p synack seq=%x", subflow, subflow->ssn_offset); | |
263e1201 | 412 | |
0799e21b | 413 | mptcp_get_options(skb, &mp_opt); |
fa25e815 | 414 | if (subflow->request_mptcp) { |
74c7dfbe | 415 | if (!(mp_opt.suboptions & OPTIONS_MPTCP_MPC)) { |
fa25e815 PA |
416 | MPTCP_INC_STATS(sock_net(sk), |
417 | MPTCP_MIB_MPCAPABLEACTIVEFALLBACK); | |
418 | mptcp_do_fallback(sk); | |
419 | pr_fallback(mptcp_sk(subflow->conn)); | |
420 | goto fallback; | |
421 | } | |
422 | ||
74c7dfbe | 423 | if (mp_opt.suboptions & OPTION_MPTCP_CSUMREQD) |
06251181 | 424 | WRITE_ONCE(mptcp_sk(parent)->csum_enabled, true); |
df377be3 GT |
425 | if (mp_opt.deny_join_id0) |
426 | WRITE_ONCE(mptcp_sk(parent)->pm.remote_deny_join_id0, true); | |
263e1201 PA |
427 | subflow->mp_capable = 1; |
428 | subflow->can_ack = 1; | |
cfde141e | 429 | subflow->remote_key = mp_opt.sndr_key; |
263e1201 PA |
430 | pr_debug("subflow=%p, remote_key=%llu", subflow, |
431 | subflow->remote_key); | |
5695eb88 | 432 | MPTCP_INC_STATS(sock_net(sk), MPTCP_MIB_MPCAPABLEACTIVEACK); |
fa25e815 | 433 | mptcp_finish_connect(sk); |
490274b4 | 434 | mptcp_set_connected(parent); |
fa25e815 PA |
435 | } else if (subflow->request_join) { |
436 | u8 hmac[SHA256_DIGEST_SIZE]; | |
437 | ||
74c7dfbe | 438 | if (!(mp_opt.suboptions & OPTIONS_MPTCP_MPJ)) { |
dc87efdb | 439 | subflow->reset_reason = MPTCP_RST_EMPTCP; |
fa25e815 | 440 | goto do_reset; |
dc87efdb | 441 | } |
fa25e815 | 442 | |
0460ce22 | 443 | subflow->backup = mp_opt.backup; |
cfde141e PA |
444 | subflow->thmac = mp_opt.thmac; |
445 | subflow->remote_nonce = mp_opt.nonce; | |
8a348392 | 446 | subflow->remote_id = mp_opt.join_id; |
0460ce22 PA |
447 | pr_debug("subflow=%p, thmac=%llu, remote_nonce=%u backup=%d", |
448 | subflow, subflow->thmac, subflow->remote_nonce, | |
449 | subflow->backup); | |
263e1201 | 450 | |
ec3edaa7 | 451 | if (!subflow_thmac_valid(subflow)) { |
fc518953 | 452 | MPTCP_INC_STATS(sock_net(sk), MPTCP_MIB_JOINACKMAC); |
dc87efdb | 453 | subflow->reset_reason = MPTCP_RST_EMPTCP; |
ec3edaa7 PK |
454 | goto do_reset; |
455 | } | |
456 | ||
0a4d8e96 JW |
457 | if (!mptcp_finish_join(sk)) |
458 | goto do_reset; | |
459 | ||
ec3edaa7 PK |
460 | subflow_generate_hmac(subflow->local_key, subflow->remote_key, |
461 | subflow->local_nonce, | |
462 | subflow->remote_nonce, | |
bd697222 | 463 | hmac); |
bd697222 | 464 | memcpy(subflow->hmac, hmac, MPTCPOPT_HMAC_LEN); |
ec3edaa7 | 465 | |
fa25e815 | 466 | subflow->mp_join = 1; |
fc518953 | 467 | MPTCP_INC_STATS(sock_net(sk), MPTCP_MIB_JOINSYNACKRX); |
5bc56388 GT |
468 | |
469 | if (subflow_use_different_dport(mptcp_sk(parent), sk)) { | |
470 | pr_debug("synack inet_dport=%d %d", | |
471 | ntohs(inet_sk(sk)->inet_dport), | |
472 | ntohs(inet_sk(parent)->inet_dport)); | |
2fbdd9ea | 473 | MPTCP_INC_STATS(sock_net(sk), MPTCP_MIB_JOINPORTSYNACKRX); |
5bc56388 | 474 | } |
fa25e815 PA |
475 | } else if (mptcp_check_fallback(sk)) { |
476 | fallback: | |
477 | mptcp_rcv_space_init(mptcp_sk(parent), sk); | |
490274b4 | 478 | mptcp_set_connected(parent); |
cec37a6e | 479 | } |
fa25e815 PA |
480 | return; |
481 | ||
482 | do_reset: | |
dc87efdb | 483 | subflow->reset_transient = 0; |
d5824847 | 484 | mptcp_subflow_reset(sk); |
cec37a6e PK |
485 | } |
486 | ||
4cf86ae8 PA |
487 | static void subflow_set_local_id(struct mptcp_subflow_context *subflow, int local_id) |
488 | { | |
489 | subflow->local_id = local_id; | |
490 | subflow->local_id_valid = 1; | |
491 | } | |
492 | ||
493 | static int subflow_chk_local_id(struct sock *sk) | |
494 | { | |
495 | struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); | |
496 | struct mptcp_sock *msk = mptcp_sk(subflow->conn); | |
497 | int err; | |
498 | ||
499 | if (likely(subflow->local_id_valid)) | |
500 | return 0; | |
501 | ||
502 | err = mptcp_pm_get_local_id(msk, (struct sock_common *)sk); | |
503 | if (err < 0) | |
504 | return err; | |
505 | ||
506 | subflow_set_local_id(subflow, err); | |
507 | return 0; | |
508 | } | |
509 | ||
510 | static int subflow_rebuild_header(struct sock *sk) | |
511 | { | |
512 | int err = subflow_chk_local_id(sk); | |
513 | ||
514 | if (unlikely(err < 0)) | |
515 | return err; | |
516 | ||
517 | return inet_sk_rebuild_header(sk); | |
518 | } | |
519 | ||
520 | #if IS_ENABLED(CONFIG_MPTCP_IPV6) | |
521 | static int subflow_v6_rebuild_header(struct sock *sk) | |
522 | { | |
523 | int err = subflow_chk_local_id(sk); | |
524 | ||
525 | if (unlikely(err < 0)) | |
526 | return err; | |
527 | ||
528 | return inet6_sk_rebuild_header(sk); | |
529 | } | |
530 | #endif | |
531 | ||
08b8d080 | 532 | struct request_sock_ops mptcp_subflow_request_sock_ops; |
51fa7f8e | 533 | static struct tcp_request_sock_ops subflow_request_sock_ipv4_ops __ro_after_init; |
cec37a6e PK |
534 | |
535 | static int subflow_v4_conn_request(struct sock *sk, struct sk_buff *skb) | |
536 | { | |
537 | struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); | |
538 | ||
539 | pr_debug("subflow=%p", subflow); | |
540 | ||
541 | /* Never answer to SYNs sent to broadcast or multicast */ | |
542 | if (skb_rtable(skb)->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) | |
543 | goto drop; | |
544 | ||
08b8d080 | 545 | return tcp_conn_request(&mptcp_subflow_request_sock_ops, |
cec37a6e PK |
546 | &subflow_request_sock_ipv4_ops, |
547 | sk, skb); | |
548 | drop: | |
549 | tcp_listendrop(sk); | |
550 | return 0; | |
551 | } | |
552 | ||
553 | #if IS_ENABLED(CONFIG_MPTCP_IPV6) | |
51fa7f8e FW |
554 | static struct tcp_request_sock_ops subflow_request_sock_ipv6_ops __ro_after_init; |
555 | static struct inet_connection_sock_af_ops subflow_v6_specific __ro_after_init; | |
556 | static struct inet_connection_sock_af_ops subflow_v6m_specific __ro_after_init; | |
b19bc294 | 557 | static struct proto tcpv6_prot_override; |
cec37a6e PK |
558 | |
559 | static int subflow_v6_conn_request(struct sock *sk, struct sk_buff *skb) | |
560 | { | |
561 | struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); | |
562 | ||
563 | pr_debug("subflow=%p", subflow); | |
564 | ||
565 | if (skb->protocol == htons(ETH_P_IP)) | |
566 | return subflow_v4_conn_request(sk, skb); | |
567 | ||
568 | if (!ipv6_unicast_destination(skb)) | |
569 | goto drop; | |
570 | ||
dcc32f4f JK |
571 | if (ipv6_addr_v4mapped(&ipv6_hdr(skb)->saddr)) { |
572 | __IP6_INC_STATS(sock_net(sk), NULL, IPSTATS_MIB_INHDRERRORS); | |
573 | return 0; | |
574 | } | |
575 | ||
08b8d080 | 576 | return tcp_conn_request(&mptcp_subflow_request_sock_ops, |
cec37a6e PK |
577 | &subflow_request_sock_ipv6_ops, sk, skb); |
578 | ||
579 | drop: | |
580 | tcp_listendrop(sk); | |
581 | return 0; /* don't send reset */ | |
582 | } | |
583 | #endif | |
584 | ||
f296234c PK |
585 | /* validate hmac received in third ACK */ |
586 | static bool subflow_hmac_valid(const struct request_sock *req, | |
cfde141e | 587 | const struct mptcp_options_received *mp_opt) |
f296234c PK |
588 | { |
589 | const struct mptcp_subflow_request_sock *subflow_req; | |
bd697222 | 590 | u8 hmac[SHA256_DIGEST_SIZE]; |
f296234c | 591 | struct mptcp_sock *msk; |
f296234c PK |
592 | |
593 | subflow_req = mptcp_subflow_rsk(req); | |
8fd4de12 | 594 | msk = subflow_req->msk; |
f296234c PK |
595 | if (!msk) |
596 | return false; | |
597 | ||
598 | subflow_generate_hmac(msk->remote_key, msk->local_key, | |
599 | subflow_req->remote_nonce, | |
600 | subflow_req->local_nonce, hmac); | |
601 | ||
8fd4de12 | 602 | return !crypto_memneq(hmac, mp_opt->hmac, MPTCPOPT_HMAC_LEN); |
f296234c PK |
603 | } |
604 | ||
df1036da FW |
605 | static void mptcp_sock_destruct(struct sock *sk) |
606 | { | |
607 | /* if new mptcp socket isn't accepted, it is free'd | |
608 | * from the tcp listener sockets request queue, linked | |
609 | * from req->sk. The tcp socket is released. | |
610 | * This calls the ULP release function which will | |
611 | * also remove the mptcp socket, via | |
612 | * sock_put(ctx->conn). | |
613 | * | |
7ee24926 PA |
614 | * Problem is that the mptcp socket will be in |
615 | * ESTABLISHED state and will not have the SOCK_DEAD flag. | |
df1036da FW |
616 | * Both result in warnings from inet_sock_destruct. |
617 | */ | |
578c18ef | 618 | if ((1 << sk->sk_state) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT)) { |
df1036da FW |
619 | sk->sk_state = TCP_CLOSE; |
620 | WARN_ON_ONCE(sk->sk_socket); | |
621 | sock_orphan(sk); | |
622 | } | |
623 | ||
5c8c1640 | 624 | mptcp_destroy_common(mptcp_sk(sk)); |
df1036da FW |
625 | inet_sock_destruct(sk); |
626 | } | |
627 | ||
9f5ca6a5 FW |
628 | static void mptcp_force_close(struct sock *sk) |
629 | { | |
490274b4 | 630 | /* the msk is not yet exposed to user-space */ |
9f5ca6a5 FW |
631 | inet_sk_state_store(sk, TCP_CLOSE); |
632 | sk_common_release(sk); | |
633 | } | |
634 | ||
4c8941de PA |
635 | static void subflow_ulp_fallback(struct sock *sk, |
636 | struct mptcp_subflow_context *old_ctx) | |
637 | { | |
638 | struct inet_connection_sock *icsk = inet_csk(sk); | |
639 | ||
640 | mptcp_subflow_tcp_fallback(sk, old_ctx); | |
641 | icsk->icsk_ulp_ops = NULL; | |
642 | rcu_assign_pointer(icsk->icsk_ulp_data, NULL); | |
643 | tcp_sk(sk)->is_mptcp = 0; | |
b19bc294 PA |
644 | |
645 | mptcp_subflow_ops_undo_override(sk); | |
4c8941de PA |
646 | } |
647 | ||
39884604 PA |
648 | static void subflow_drop_ctx(struct sock *ssk) |
649 | { | |
650 | struct mptcp_subflow_context *ctx = mptcp_subflow_ctx(ssk); | |
651 | ||
652 | if (!ctx) | |
653 | return; | |
654 | ||
655 | subflow_ulp_fallback(ssk, ctx); | |
656 | if (ctx->conn) | |
657 | sock_put(ctx->conn); | |
658 | ||
659 | kfree_rcu(ctx, rcu); | |
660 | } | |
661 | ||
b93df08c PA |
662 | void mptcp_subflow_fully_established(struct mptcp_subflow_context *subflow, |
663 | struct mptcp_options_received *mp_opt) | |
664 | { | |
665 | struct mptcp_sock *msk = mptcp_sk(subflow->conn); | |
666 | ||
667 | subflow->remote_key = mp_opt->sndr_key; | |
668 | subflow->fully_established = 1; | |
669 | subflow->can_ack = 1; | |
670 | WRITE_ONCE(msk->fully_established, true); | |
671 | } | |
672 | ||
cec37a6e PK |
673 | static struct sock *subflow_syn_recv_sock(const struct sock *sk, |
674 | struct sk_buff *skb, | |
675 | struct request_sock *req, | |
676 | struct dst_entry *dst, | |
677 | struct request_sock *req_unhash, | |
678 | bool *own_req) | |
679 | { | |
680 | struct mptcp_subflow_context *listener = mptcp_subflow_ctx(sk); | |
cc7972ea | 681 | struct mptcp_subflow_request_sock *subflow_req; |
cfde141e | 682 | struct mptcp_options_received mp_opt; |
9e365ff5 | 683 | bool fallback, fallback_is_fatal; |
58b09919 | 684 | struct sock *new_msk = NULL; |
cec37a6e PK |
685 | struct sock *child; |
686 | ||
687 | pr_debug("listener=%p, req=%p, conn=%p", listener, req, listener->conn); | |
688 | ||
74c7dfbe | 689 | /* After child creation we must look for MPC even when options |
9e365ff5 | 690 | * are not parsed |
cfde141e | 691 | */ |
74c7dfbe | 692 | mp_opt.suboptions = 0; |
9e365ff5 PA |
693 | |
694 | /* hopefully temporary handling for MP_JOIN+syncookie */ | |
695 | subflow_req = mptcp_subflow_rsk(req); | |
b7514694 | 696 | fallback_is_fatal = tcp_rsk(req)->is_mptcp && subflow_req->mp_join; |
9e365ff5 PA |
697 | fallback = !tcp_rsk(req)->is_mptcp; |
698 | if (fallback) | |
ae2dd716 FW |
699 | goto create_child; |
700 | ||
d22f4988 | 701 | /* if the sk is MP_CAPABLE, we try to fetch the client key */ |
cc7972ea | 702 | if (subflow_req->mp_capable) { |
06f9a435 PA |
703 | /* we can receive and accept an in-window, out-of-order pkt, |
704 | * which may not carry the MP_CAPABLE opt even on mptcp enabled | |
705 | * paths: always try to extract the peer key, and fallback | |
706 | * for packets missing it. | |
707 | * Even OoO DSS packets coming legitly after dropped or | |
708 | * reordered MPC will cause fallback, but we don't have other | |
709 | * options. | |
710 | */ | |
0799e21b | 711 | mptcp_get_options(skb, &mp_opt); |
74c7dfbe | 712 | if (!(mp_opt.suboptions & OPTIONS_MPTCP_MPC)) { |
4c8941de | 713 | fallback = true; |
58b09919 | 714 | goto create_child; |
d22f4988 | 715 | } |
58b09919 | 716 | |
cfde141e | 717 | new_msk = mptcp_sk_clone(listener->conn, &mp_opt, req); |
58b09919 | 718 | if (!new_msk) |
4c8941de | 719 | fallback = true; |
f296234c | 720 | } else if (subflow_req->mp_join) { |
0799e21b | 721 | mptcp_get_options(skb, &mp_opt); |
74c7dfbe PA |
722 | if (!(mp_opt.suboptions & OPTIONS_MPTCP_MPJ) || |
723 | !subflow_hmac_valid(req, &mp_opt) || | |
d3ab7885 | 724 | !mptcp_can_accept_new_subflow(subflow_req->msk)) { |
fc518953 | 725 | SUBFLOW_REQ_INC_STATS(req, MPTCP_MIB_JOINACKMAC); |
9e365ff5 | 726 | fallback = true; |
fc518953 | 727 | } |
cc7972ea | 728 | } |
cec37a6e | 729 | |
d22f4988 | 730 | create_child: |
cec37a6e PK |
731 | child = listener->icsk_af_ops->syn_recv_sock(sk, skb, req, dst, |
732 | req_unhash, own_req); | |
733 | ||
734 | if (child && *own_req) { | |
79c0949e PK |
735 | struct mptcp_subflow_context *ctx = mptcp_subflow_ctx(child); |
736 | ||
90bf4513 PA |
737 | tcp_rsk(req)->drop_req = false; |
738 | ||
4c8941de PA |
739 | /* we need to fallback on ctx allocation failure and on pre-reqs |
740 | * checking above. In the latter scenario we additionally need | |
741 | * to reset the context to non MPTCP status. | |
79c0949e | 742 | */ |
4c8941de | 743 | if (!ctx || fallback) { |
dc87efdb FW |
744 | if (fallback_is_fatal) { |
745 | subflow_add_reset_reason(skb, MPTCP_RST_EMPTCP); | |
729cd643 | 746 | goto dispose_child; |
dc87efdb | 747 | } |
4c8941de | 748 | |
39884604 | 749 | subflow_drop_ctx(child); |
58b09919 | 750 | goto out; |
f296234c | 751 | } |
79c0949e | 752 | |
df00b087 FW |
753 | /* ssk inherits options of listener sk */ |
754 | ctx->setsockopt_seq = listener->setsockopt_seq; | |
755 | ||
79c0949e | 756 | if (ctx->mp_capable) { |
b93df08c PA |
757 | /* this can't race with mptcp_close(), as the msk is |
758 | * not yet exposted to user-space | |
759 | */ | |
760 | inet_sk_state_store((void *)new_msk, TCP_ESTABLISHED); | |
761 | ||
5b950ff4 PA |
762 | /* record the newly created socket as the first msk |
763 | * subflow, but don't link it yet into conn_list | |
764 | */ | |
0397c6d8 PA |
765 | WRITE_ONCE(mptcp_sk(new_msk)->first, child); |
766 | ||
58b09919 PA |
767 | /* new mpc subflow takes ownership of the newly |
768 | * created mptcp socket | |
769 | */ | |
df1036da | 770 | new_msk->sk_destruct = mptcp_sock_destruct; |
df00b087 | 771 | mptcp_sk(new_msk)->setsockopt_seq = ctx->setsockopt_seq; |
6c714f1b | 772 | mptcp_pm_new_connection(mptcp_sk(new_msk), child, 1); |
2c5ebd00 | 773 | mptcp_token_accept(subflow_req, mptcp_sk(new_msk)); |
58b09919 PA |
774 | ctx->conn = new_msk; |
775 | new_msk = NULL; | |
fca5c82c PA |
776 | |
777 | /* with OoO packets we can reach here without ingress | |
778 | * mpc option | |
779 | */ | |
74c7dfbe | 780 | if (mp_opt.suboptions & OPTIONS_MPTCP_MPC) |
b93df08c | 781 | mptcp_subflow_fully_established(ctx, &mp_opt); |
f296234c PK |
782 | } else if (ctx->mp_join) { |
783 | struct mptcp_sock *owner; | |
784 | ||
8fd4de12 | 785 | owner = subflow_req->msk; |
dc87efdb FW |
786 | if (!owner) { |
787 | subflow_add_reset_reason(skb, MPTCP_RST_EPROHIBIT); | |
729cd643 | 788 | goto dispose_child; |
dc87efdb | 789 | } |
f296234c | 790 | |
8fd4de12 PA |
791 | /* move the msk reference ownership to the subflow */ |
792 | subflow_req->msk = NULL; | |
f296234c | 793 | ctx->conn = (struct sock *)owner; |
5bc56388 GT |
794 | |
795 | if (subflow_use_different_sport(owner, sk)) { | |
796 | pr_debug("ack inet_sport=%d %d", | |
797 | ntohs(inet_sk(sk)->inet_sport), | |
798 | ntohs(inet_sk((struct sock *)owner)->inet_sport)); | |
2fbdd9ea GT |
799 | if (!mptcp_pm_sport_in_anno_list(owner, sk)) { |
800 | SUBFLOW_REQ_INC_STATS(req, MPTCP_MIB_MISMATCHPORTACKRX); | |
9238e900 | 801 | goto dispose_child; |
2fbdd9ea GT |
802 | } |
803 | SUBFLOW_REQ_INC_STATS(req, MPTCP_MIB_JOINPORTACKRX); | |
5bc56388 | 804 | } |
9238e900 GT |
805 | |
806 | if (!mptcp_finish_join(child)) | |
807 | goto dispose_child; | |
808 | ||
809 | SUBFLOW_REQ_INC_STATS(req, MPTCP_MIB_JOINACKRX); | |
810 | tcp_rsk(req)->drop_req = true; | |
cec37a6e PK |
811 | } |
812 | } | |
813 | ||
58b09919 PA |
814 | out: |
815 | /* dispose of the left over mptcp master, if any */ | |
816 | if (unlikely(new_msk)) | |
9f5ca6a5 | 817 | mptcp_force_close(new_msk); |
4c8941de PA |
818 | |
819 | /* check for expected invariant - should never trigger, just help | |
820 | * catching eariler subtle bugs | |
821 | */ | |
ac2b47fb | 822 | WARN_ON_ONCE(child && *own_req && tcp_sk(child)->is_mptcp && |
4c8941de PA |
823 | (!mptcp_subflow_ctx(child) || |
824 | !mptcp_subflow_ctx(child)->conn)); | |
cec37a6e | 825 | return child; |
f296234c | 826 | |
729cd643 | 827 | dispose_child: |
39884604 | 828 | subflow_drop_ctx(child); |
729cd643 | 829 | tcp_rsk(req)->drop_req = true; |
729cd643 | 830 | inet_csk_prepare_for_destroy_sock(child); |
f296234c | 831 | tcp_done(child); |
97e61751 | 832 | req->rsk_ops->send_reset(sk, skb); |
729cd643 PA |
833 | |
834 | /* The last child reference will be released by the caller */ | |
835 | return child; | |
cec37a6e PK |
836 | } |
837 | ||
51fa7f8e | 838 | static struct inet_connection_sock_af_ops subflow_specific __ro_after_init; |
b19bc294 | 839 | static struct proto tcp_prot_override; |
cec37a6e | 840 | |
648ef4b8 MM |
841 | enum mapping_status { |
842 | MAPPING_OK, | |
843 | MAPPING_INVALID, | |
844 | MAPPING_EMPTY, | |
e1ff9e82 | 845 | MAPPING_DATA_FIN, |
31bf11de PA |
846 | MAPPING_DUMMY, |
847 | MAPPING_BAD_CSUM | |
648ef4b8 MM |
848 | }; |
849 | ||
61e71022 | 850 | static void dbg_bad_map(struct mptcp_subflow_context *subflow, u32 ssn) |
648ef4b8 | 851 | { |
61e71022 PA |
852 | pr_debug("Bad mapping: ssn=%d map_seq=%d map_data_len=%d", |
853 | ssn, subflow->map_subflow_seq, subflow->map_data_len); | |
648ef4b8 MM |
854 | } |
855 | ||
856 | static bool skb_is_fully_mapped(struct sock *ssk, struct sk_buff *skb) | |
857 | { | |
858 | struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); | |
859 | unsigned int skb_consumed; | |
860 | ||
861 | skb_consumed = tcp_sk(ssk)->copied_seq - TCP_SKB_CB(skb)->seq; | |
862 | if (WARN_ON_ONCE(skb_consumed >= skb->len)) | |
863 | return true; | |
864 | ||
865 | return skb->len - skb_consumed <= subflow->map_data_len - | |
866 | mptcp_subflow_get_map_offset(subflow); | |
867 | } | |
868 | ||
869 | static bool validate_mapping(struct sock *ssk, struct sk_buff *skb) | |
870 | { | |
871 | struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); | |
872 | u32 ssn = tcp_sk(ssk)->copied_seq - subflow->ssn_offset; | |
873 | ||
874 | if (unlikely(before(ssn, subflow->map_subflow_seq))) { | |
875 | /* Mapping covers data later in the subflow stream, | |
876 | * currently unsupported. | |
877 | */ | |
61e71022 | 878 | dbg_bad_map(subflow, ssn); |
648ef4b8 MM |
879 | return false; |
880 | } | |
881 | if (unlikely(!before(ssn, subflow->map_subflow_seq + | |
882 | subflow->map_data_len))) { | |
883 | /* Mapping does covers past subflow data, invalid */ | |
61e71022 | 884 | dbg_bad_map(subflow, ssn); |
648ef4b8 MM |
885 | return false; |
886 | } | |
887 | return true; | |
888 | } | |
889 | ||
dd8bcd17 PA |
890 | static enum mapping_status validate_data_csum(struct sock *ssk, struct sk_buff *skb, |
891 | bool csum_reqd) | |
892 | { | |
893 | struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); | |
dd8bcd17 | 894 | u32 offset, seq, delta; |
ba2c89e0 | 895 | __sum16 csum; |
dd8bcd17 PA |
896 | int len; |
897 | ||
898 | if (!csum_reqd) | |
899 | return MAPPING_OK; | |
900 | ||
901 | /* mapping already validated on previous traversal */ | |
902 | if (subflow->map_csum_len == subflow->map_data_len) | |
903 | return MAPPING_OK; | |
904 | ||
905 | /* traverse the receive queue, ensuring it contains a full | |
906 | * DSS mapping and accumulating the related csum. | |
907 | * Preserve the accoumlate csum across multiple calls, to compute | |
908 | * the csum only once | |
909 | */ | |
910 | delta = subflow->map_data_len - subflow->map_csum_len; | |
911 | for (;;) { | |
912 | seq = tcp_sk(ssk)->copied_seq + subflow->map_csum_len; | |
913 | offset = seq - TCP_SKB_CB(skb)->seq; | |
914 | ||
915 | /* if the current skb has not been accounted yet, csum its contents | |
916 | * up to the amount covered by the current DSS | |
917 | */ | |
918 | if (offset < skb->len) { | |
919 | __wsum csum; | |
920 | ||
921 | len = min(skb->len - offset, delta); | |
922 | csum = skb_checksum(skb, offset, len, 0); | |
923 | subflow->map_data_csum = csum_block_add(subflow->map_data_csum, csum, | |
924 | subflow->map_csum_len); | |
925 | ||
926 | delta -= len; | |
927 | subflow->map_csum_len += len; | |
928 | } | |
929 | if (delta == 0) | |
930 | break; | |
931 | ||
932 | if (skb_queue_is_last(&ssk->sk_receive_queue, skb)) { | |
933 | /* if this subflow is closed, the partial mapping | |
934 | * will be never completed; flush the pending skbs, so | |
935 | * that subflow_sched_work_if_closed() can kick in | |
936 | */ | |
937 | if (unlikely(ssk->sk_state == TCP_CLOSE)) | |
938 | while ((skb = skb_peek(&ssk->sk_receive_queue))) | |
939 | sk_eat_skb(ssk, skb); | |
940 | ||
941 | /* not enough data to validate the csum */ | |
942 | return MAPPING_EMPTY; | |
943 | } | |
944 | ||
945 | /* the DSS mapping for next skbs will be validated later, | |
946 | * when a get_mapping_status call will process such skb | |
947 | */ | |
948 | skb = skb->next; | |
949 | } | |
950 | ||
951 | /* note that 'map_data_len' accounts only for the carried data, does | |
952 | * not include the eventual seq increment due to the data fin, | |
953 | * while the pseudo header requires the original DSS data len, | |
954 | * including that | |
955 | */ | |
8401e87f GT |
956 | csum = __mptcp_make_csum(subflow->map_seq, |
957 | subflow->map_subflow_seq, | |
958 | subflow->map_data_len + subflow->map_data_fin, | |
959 | subflow->map_data_csum); | |
960 | if (unlikely(csum)) { | |
fe3ab1cb | 961 | MPTCP_INC_STATS(sock_net(ssk), MPTCP_MIB_DATACSUMERR); |
31bf11de | 962 | return MAPPING_BAD_CSUM; |
fe3ab1cb | 963 | } |
dd8bcd17 | 964 | |
ae66fb2b | 965 | subflow->valid_csum_seen = 1; |
dd8bcd17 PA |
966 | return MAPPING_OK; |
967 | } | |
968 | ||
43b54c6e MM |
969 | static enum mapping_status get_mapping_status(struct sock *ssk, |
970 | struct mptcp_sock *msk) | |
648ef4b8 MM |
971 | { |
972 | struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); | |
dd8bcd17 | 973 | bool csum_reqd = READ_ONCE(msk->csum_enabled); |
49fa1919 | 974 | struct sock *sk = (struct sock *)msk; |
648ef4b8 MM |
975 | struct mptcp_ext *mpext; |
976 | struct sk_buff *skb; | |
977 | u16 data_len; | |
978 | u64 map_seq; | |
979 | ||
980 | skb = skb_peek(&ssk->sk_receive_queue); | |
981 | if (!skb) | |
982 | return MAPPING_EMPTY; | |
983 | ||
e1ff9e82 DC |
984 | if (mptcp_check_fallback(ssk)) |
985 | return MAPPING_DUMMY; | |
986 | ||
648ef4b8 MM |
987 | mpext = mptcp_get_ext(skb); |
988 | if (!mpext || !mpext->use_map) { | |
989 | if (!subflow->map_valid && !skb->len) { | |
990 | /* the TCP stack deliver 0 len FIN pkt to the receive | |
991 | * queue, that is the only 0len pkts ever expected here, | |
992 | * and we can admit no mapping only for 0 len pkts | |
993 | */ | |
994 | if (!(TCP_SKB_CB(skb)->tcp_flags & TCPHDR_FIN)) | |
995 | WARN_ONCE(1, "0len seq %d:%d flags %x", | |
996 | TCP_SKB_CB(skb)->seq, | |
997 | TCP_SKB_CB(skb)->end_seq, | |
998 | TCP_SKB_CB(skb)->tcp_flags); | |
999 | sk_eat_skb(ssk, skb); | |
1000 | return MAPPING_EMPTY; | |
1001 | } | |
1002 | ||
1003 | if (!subflow->map_valid) | |
1004 | return MAPPING_INVALID; | |
1005 | ||
1006 | goto validate_seq; | |
1007 | } | |
1008 | ||
0918e34b | 1009 | trace_get_mapping_status(mpext); |
648ef4b8 MM |
1010 | |
1011 | data_len = mpext->data_len; | |
1012 | if (data_len == 0) { | |
f8d4bcac | 1013 | pr_debug("infinite mapping received"); |
fc518953 | 1014 | MPTCP_INC_STATS(sock_net(ssk), MPTCP_MIB_INFINITEMAPRX); |
f8d4bcac | 1015 | subflow->map_data_len = 0; |
d42f9e4e MM |
1016 | if (!sock_flag(ssk, SOCK_DEAD)) |
1017 | sk_stop_timer(sk, &sk->sk_timer); | |
1018 | ||
648ef4b8 MM |
1019 | return MAPPING_INVALID; |
1020 | } | |
1021 | ||
1022 | if (mpext->data_fin == 1) { | |
1023 | if (data_len == 1) { | |
1a49b2c2 MM |
1024 | bool updated = mptcp_update_rcv_data_fin(msk, mpext->data_seq, |
1025 | mpext->dsn64); | |
43b54c6e | 1026 | pr_debug("DATA_FIN with no payload seq=%llu", mpext->data_seq); |
648ef4b8 MM |
1027 | if (subflow->map_valid) { |
1028 | /* A DATA_FIN might arrive in a DSS | |
1029 | * option before the previous mapping | |
1030 | * has been fully consumed. Continue | |
1031 | * handling the existing mapping. | |
1032 | */ | |
1033 | skb_ext_del(skb, SKB_EXT_MPTCP); | |
1034 | return MAPPING_OK; | |
1035 | } else { | |
ef59b195 MM |
1036 | if (updated && schedule_work(&msk->work)) |
1037 | sock_hold((struct sock *)msk); | |
1038 | ||
648ef4b8 MM |
1039 | return MAPPING_DATA_FIN; |
1040 | } | |
43b54c6e | 1041 | } else { |
017512a0 | 1042 | u64 data_fin_seq = mpext->data_seq + data_len - 1; |
1a49b2c2 MM |
1043 | |
1044 | /* If mpext->data_seq is a 32-bit value, data_fin_seq | |
1045 | * must also be limited to 32 bits. | |
1046 | */ | |
1047 | if (!mpext->dsn64) | |
1048 | data_fin_seq &= GENMASK_ULL(31, 0); | |
1049 | ||
1050 | mptcp_update_rcv_data_fin(msk, data_fin_seq, mpext->dsn64); | |
1051 | pr_debug("DATA_FIN with mapping seq=%llu dsn64=%d", | |
1052 | data_fin_seq, mpext->dsn64); | |
648ef4b8 MM |
1053 | } |
1054 | ||
1055 | /* Adjust for DATA_FIN using 1 byte of sequence space */ | |
1056 | data_len--; | |
1057 | } | |
1058 | ||
5957a890 | 1059 | map_seq = mptcp_expand_seq(READ_ONCE(msk->ack_seq), mpext->data_seq, mpext->dsn64); |
37198e93 | 1060 | WRITE_ONCE(mptcp_sk(subflow->conn)->use_64bit_ack, !!mpext->dsn64); |
648ef4b8 MM |
1061 | |
1062 | if (subflow->map_valid) { | |
1063 | /* Allow replacing only with an identical map */ | |
1064 | if (subflow->map_seq == map_seq && | |
1065 | subflow->map_subflow_seq == mpext->subflow_seq && | |
dd8bcd17 PA |
1066 | subflow->map_data_len == data_len && |
1067 | subflow->map_csum_reqd == mpext->csum_reqd) { | |
648ef4b8 | 1068 | skb_ext_del(skb, SKB_EXT_MPTCP); |
dd8bcd17 | 1069 | goto validate_csum; |
648ef4b8 MM |
1070 | } |
1071 | ||
1072 | /* If this skb data are fully covered by the current mapping, | |
1073 | * the new map would need caching, which is not supported | |
1074 | */ | |
fc518953 FW |
1075 | if (skb_is_fully_mapped(ssk, skb)) { |
1076 | MPTCP_INC_STATS(sock_net(ssk), MPTCP_MIB_DSSNOMATCH); | |
648ef4b8 | 1077 | return MAPPING_INVALID; |
fc518953 | 1078 | } |
648ef4b8 MM |
1079 | |
1080 | /* will validate the next map after consuming the current one */ | |
dd8bcd17 | 1081 | goto validate_csum; |
648ef4b8 MM |
1082 | } |
1083 | ||
1084 | subflow->map_seq = map_seq; | |
1085 | subflow->map_subflow_seq = mpext->subflow_seq; | |
1086 | subflow->map_data_len = data_len; | |
1087 | subflow->map_valid = 1; | |
dd8bcd17 | 1088 | subflow->map_data_fin = mpext->data_fin; |
d22f4988 | 1089 | subflow->mpc_map = mpext->mpc_map; |
dd8bcd17 PA |
1090 | subflow->map_csum_reqd = mpext->csum_reqd; |
1091 | subflow->map_csum_len = 0; | |
1092 | subflow->map_data_csum = csum_unfold(mpext->csum); | |
1093 | ||
1094 | /* Cfr RFC 8684 Section 3.3.0 */ | |
1095 | if (unlikely(subflow->map_csum_reqd != csum_reqd)) | |
1096 | return MAPPING_INVALID; | |
1097 | ||
1098 | pr_debug("new map seq=%llu subflow_seq=%u data_len=%u csum=%d:%u", | |
648ef4b8 | 1099 | subflow->map_seq, subflow->map_subflow_seq, |
dd8bcd17 PA |
1100 | subflow->map_data_len, subflow->map_csum_reqd, |
1101 | subflow->map_data_csum); | |
648ef4b8 MM |
1102 | |
1103 | validate_seq: | |
1104 | /* we revalidate valid mapping on new skb, because we must ensure | |
1105 | * the current skb is completely covered by the available mapping | |
1106 | */ | |
06285da9 PA |
1107 | if (!validate_mapping(ssk, skb)) { |
1108 | MPTCP_INC_STATS(sock_net(ssk), MPTCP_MIB_DSSTCPMISMATCH); | |
648ef4b8 | 1109 | return MAPPING_INVALID; |
06285da9 | 1110 | } |
648ef4b8 MM |
1111 | |
1112 | skb_ext_del(skb, SKB_EXT_MPTCP); | |
dd8bcd17 PA |
1113 | |
1114 | validate_csum: | |
1115 | return validate_data_csum(ssk, skb, csum_reqd); | |
648ef4b8 MM |
1116 | } |
1117 | ||
04e4cd4f | 1118 | static void mptcp_subflow_discard_data(struct sock *ssk, struct sk_buff *skb, |
1d39cd8c | 1119 | u64 limit) |
6719331c PA |
1120 | { |
1121 | struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); | |
04e4cd4f PA |
1122 | bool fin = TCP_SKB_CB(skb)->tcp_flags & TCPHDR_FIN; |
1123 | u32 incr; | |
1124 | ||
1125 | incr = limit >= skb->len ? skb->len + fin : limit; | |
1126 | ||
1127 | pr_debug("discarding=%d len=%d seq=%d", incr, skb->len, | |
1128 | subflow->map_subflow_seq); | |
06242e44 | 1129 | MPTCP_INC_STATS(sock_net(ssk), MPTCP_MIB_DUPDATA); |
04e4cd4f PA |
1130 | tcp_sk(ssk)->copied_seq += incr; |
1131 | if (!before(tcp_sk(ssk)->copied_seq, TCP_SKB_CB(skb)->end_seq)) | |
1132 | sk_eat_skb(ssk, skb); | |
1133 | if (mptcp_subflow_get_map_offset(subflow) >= subflow->map_data_len) | |
1134 | subflow->map_valid = 0; | |
6719331c PA |
1135 | } |
1136 | ||
40947e13 FW |
1137 | /* sched mptcp worker to remove the subflow if no more data is pending */ |
1138 | static void subflow_sched_work_if_closed(struct mptcp_sock *msk, struct sock *ssk) | |
1139 | { | |
1140 | struct sock *sk = (struct sock *)msk; | |
1141 | ||
1142 | if (likely(ssk->sk_state != TCP_CLOSE)) | |
1143 | return; | |
1144 | ||
1145 | if (skb_queue_empty(&ssk->sk_receive_queue) && | |
1146 | !test_and_set_bit(MPTCP_WORK_CLOSE_SUBFLOW, &msk->flags)) { | |
1147 | sock_hold(sk); | |
1148 | if (!schedule_work(&msk->work)) | |
1149 | sock_put(sk); | |
1150 | } | |
1151 | } | |
1152 | ||
ae66fb2b MM |
1153 | static bool subflow_can_fallback(struct mptcp_subflow_context *subflow) |
1154 | { | |
1155 | struct mptcp_sock *msk = mptcp_sk(subflow->conn); | |
1156 | ||
1157 | if (subflow->mp_join) | |
1158 | return false; | |
1159 | else if (READ_ONCE(msk->csum_enabled)) | |
1160 | return !subflow->valid_csum_seen; | |
1161 | else | |
1162 | return !subflow->fully_established; | |
1163 | } | |
1164 | ||
648ef4b8 MM |
1165 | static bool subflow_check_data_avail(struct sock *ssk) |
1166 | { | |
1167 | struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); | |
1168 | enum mapping_status status; | |
1169 | struct mptcp_sock *msk; | |
1170 | struct sk_buff *skb; | |
1171 | ||
47bebdf3 | 1172 | if (!skb_peek(&ssk->sk_receive_queue)) |
ea56dcb4 | 1173 | WRITE_ONCE(subflow->data_avail, MPTCP_SUBFLOW_NODATA); |
648ef4b8 MM |
1174 | if (subflow->data_avail) |
1175 | return true; | |
1176 | ||
648ef4b8 MM |
1177 | msk = mptcp_sk(subflow->conn); |
1178 | for (;;) { | |
648ef4b8 MM |
1179 | u64 ack_seq; |
1180 | u64 old_ack; | |
1181 | ||
43b54c6e | 1182 | status = get_mapping_status(ssk, msk); |
d96a838a | 1183 | trace_subflow_check_data_avail(status, skb_peek(&ssk->sk_receive_queue)); |
31bf11de PA |
1184 | if (unlikely(status == MAPPING_INVALID || status == MAPPING_DUMMY || |
1185 | status == MAPPING_BAD_CSUM)) | |
dea2b1ea | 1186 | goto fallback; |
648ef4b8 MM |
1187 | |
1188 | if (status != MAPPING_OK) | |
40947e13 | 1189 | goto no_data; |
648ef4b8 MM |
1190 | |
1191 | skb = skb_peek(&ssk->sk_receive_queue); | |
1192 | if (WARN_ON_ONCE(!skb)) | |
40947e13 | 1193 | goto no_data; |
648ef4b8 | 1194 | |
d22f4988 CP |
1195 | /* if msk lacks the remote key, this subflow must provide an |
1196 | * MP_CAPABLE-based mapping | |
1197 | */ | |
1198 | if (unlikely(!READ_ONCE(msk->can_ack))) { | |
dea2b1ea PA |
1199 | if (!subflow->mpc_map) |
1200 | goto fallback; | |
d22f4988 CP |
1201 | WRITE_ONCE(msk->remote_key, subflow->remote_key); |
1202 | WRITE_ONCE(msk->ack_seq, subflow->map_seq); | |
1203 | WRITE_ONCE(msk->can_ack, true); | |
1204 | } | |
1205 | ||
648ef4b8 MM |
1206 | old_ack = READ_ONCE(msk->ack_seq); |
1207 | ack_seq = mptcp_subflow_get_mapped_dsn(subflow); | |
1208 | pr_debug("msk ack_seq=%llx subflow ack_seq=%llx", old_ack, | |
1209 | ack_seq); | |
99d1055c PA |
1210 | if (unlikely(before64(ack_seq, old_ack))) { |
1211 | mptcp_subflow_discard_data(ssk, skb, old_ack - ack_seq); | |
1212 | continue; | |
47bebdf3 | 1213 | } |
648ef4b8 | 1214 | |
99d1055c PA |
1215 | WRITE_ONCE(subflow->data_avail, MPTCP_SUBFLOW_DATA_AVAIL); |
1216 | break; | |
648ef4b8 MM |
1217 | } |
1218 | return true; | |
1219 | ||
40947e13 FW |
1220 | no_data: |
1221 | subflow_sched_work_if_closed(msk, ssk); | |
1222 | return false; | |
dea2b1ea PA |
1223 | |
1224 | fallback: | |
0348c690 GT |
1225 | if (!__mptcp_check_fallback(msk)) { |
1226 | /* RFC 8684 section 3.7. */ | |
31bf11de PA |
1227 | if (status == MAPPING_BAD_CSUM && |
1228 | (subflow->mp_join || subflow->valid_csum_seen)) { | |
1229 | subflow->send_mp_fail = 1; | |
1230 | ||
7b16871f | 1231 | if (!READ_ONCE(msk->allow_infinite_fallback)) { |
0348c690 GT |
1232 | ssk->sk_err = EBADMSG; |
1233 | tcp_set_state(ssk, TCP_CLOSE); | |
1234 | subflow->reset_transient = 0; | |
1235 | subflow->reset_reason = MPTCP_RST_EMIDDLEBOX; | |
1236 | tcp_send_active_reset(ssk, GFP_ATOMIC); | |
1237 | while ((skb = skb_peek(&ssk->sk_receive_queue))) | |
1238 | sk_eat_skb(ssk, skb); | |
d42f9e4e | 1239 | } else if (!sock_flag(ssk, SOCK_DEAD)) { |
9c81be0d | 1240 | WRITE_ONCE(subflow->mp_fail_response_expect, true); |
49fa1919 GT |
1241 | sk_reset_timer((struct sock *)msk, |
1242 | &((struct sock *)msk)->sk_timer, | |
1243 | jiffies + TCP_RTO_MAX); | |
0348c690 GT |
1244 | } |
1245 | WRITE_ONCE(subflow->data_avail, MPTCP_SUBFLOW_NODATA); | |
1246 | return true; | |
1247 | } | |
1248 | ||
d7e6f583 | 1249 | if (!subflow_can_fallback(subflow) && subflow->map_data_len) { |
0348c690 GT |
1250 | /* fatal protocol error, close the socket. |
1251 | * subflow_error_report() will introduce the appropriate barriers | |
1252 | */ | |
1761fed2 GT |
1253 | ssk->sk_err = EBADMSG; |
1254 | tcp_set_state(ssk, TCP_CLOSE); | |
1255 | subflow->reset_transient = 0; | |
0348c690 | 1256 | subflow->reset_reason = MPTCP_RST_EMPTCP; |
1761fed2 | 1257 | tcp_send_active_reset(ssk, GFP_ATOMIC); |
0348c690 GT |
1258 | WRITE_ONCE(subflow->data_avail, MPTCP_SUBFLOW_NODATA); |
1259 | return false; | |
478d7700 | 1260 | } |
478d7700 | 1261 | |
0348c690 | 1262 | __mptcp_do_fallback(msk); |
dea2b1ea PA |
1263 | } |
1264 | ||
dea2b1ea PA |
1265 | skb = skb_peek(&ssk->sk_receive_queue); |
1266 | subflow->map_valid = 1; | |
1267 | subflow->map_seq = READ_ONCE(msk->ack_seq); | |
1268 | subflow->map_data_len = skb->len; | |
1269 | subflow->map_subflow_seq = tcp_sk(ssk)->copied_seq - subflow->ssn_offset; | |
99d1055c | 1270 | WRITE_ONCE(subflow->data_avail, MPTCP_SUBFLOW_DATA_AVAIL); |
dea2b1ea | 1271 | return true; |
648ef4b8 MM |
1272 | } |
1273 | ||
1274 | bool mptcp_subflow_data_available(struct sock *sk) | |
1275 | { | |
1276 | struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); | |
648ef4b8 MM |
1277 | |
1278 | /* check if current mapping is still valid */ | |
1279 | if (subflow->map_valid && | |
1280 | mptcp_subflow_get_map_offset(subflow) >= subflow->map_data_len) { | |
1281 | subflow->map_valid = 0; | |
ea56dcb4 | 1282 | WRITE_ONCE(subflow->data_avail, MPTCP_SUBFLOW_NODATA); |
648ef4b8 MM |
1283 | |
1284 | pr_debug("Done with mapping: seq=%u data_len=%u", | |
1285 | subflow->map_subflow_seq, | |
1286 | subflow->map_data_len); | |
1287 | } | |
1288 | ||
47bebdf3 | 1289 | return subflow_check_data_avail(sk); |
648ef4b8 MM |
1290 | } |
1291 | ||
071c8ed6 FW |
1292 | /* If ssk has an mptcp parent socket, use the mptcp rcvbuf occupancy, |
1293 | * not the ssk one. | |
1294 | * | |
1295 | * In mptcp, rwin is about the mptcp-level connection data. | |
1296 | * | |
1297 | * Data that is still on the ssk rx queue can thus be ignored, | |
55320b82 | 1298 | * as far as mptcp peer is concerned that data is still inflight. |
071c8ed6 FW |
1299 | * DSS ACK is updated when skb is moved to the mptcp rx queue. |
1300 | */ | |
1301 | void mptcp_space(const struct sock *ssk, int *space, int *full_space) | |
1302 | { | |
1303 | const struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); | |
1304 | const struct sock *sk = subflow->conn; | |
1305 | ||
ea4ca586 | 1306 | *space = __mptcp_space(sk); |
071c8ed6 FW |
1307 | *full_space = tcp_full_space(sk); |
1308 | } | |
1309 | ||
15cc1045 PA |
1310 | void __mptcp_error_report(struct sock *sk) |
1311 | { | |
1312 | struct mptcp_subflow_context *subflow; | |
1313 | struct mptcp_sock *msk = mptcp_sk(sk); | |
1314 | ||
1315 | mptcp_for_each_subflow(msk, subflow) { | |
1316 | struct sock *ssk = mptcp_subflow_tcp_sock(subflow); | |
1317 | int err = sock_error(ssk); | |
1318 | ||
1319 | if (!err) | |
1320 | continue; | |
1321 | ||
1322 | /* only propagate errors on fallen-back sockets or | |
1323 | * on MPC connect | |
1324 | */ | |
1325 | if (sk->sk_state != TCP_SYN_SENT && !__mptcp_check_fallback(msk)) | |
1326 | continue; | |
1327 | ||
1328 | inet_sk_state_store(sk, inet_sk_state_load(ssk)); | |
1329 | sk->sk_err = -err; | |
1330 | ||
1331 | /* This barrier is coupled with smp_rmb() in mptcp_poll() */ | |
1332 | smp_wmb(); | |
e3ae2365 | 1333 | sk_error_report(sk); |
15cc1045 PA |
1334 | break; |
1335 | } | |
1336 | } | |
1337 | ||
1338 | static void subflow_error_report(struct sock *ssk) | |
1339 | { | |
1340 | struct sock *sk = mptcp_subflow_ctx(ssk)->conn; | |
1341 | ||
1342 | mptcp_data_lock(sk); | |
1343 | if (!sock_owned_by_user(sk)) | |
1344 | __mptcp_error_report(sk); | |
1345 | else | |
e9d09bac | 1346 | __set_bit(MPTCP_ERROR_REPORT, &mptcp_sk(sk)->cb_flags); |
15cc1045 PA |
1347 | mptcp_data_unlock(sk); |
1348 | } | |
1349 | ||
499ada50 PA |
1350 | static void subflow_data_ready(struct sock *sk) |
1351 | { | |
1352 | struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); | |
1353 | u16 state = 1 << inet_sk_state_load(sk); | |
1354 | struct sock *parent = subflow->conn; | |
1355 | struct mptcp_sock *msk; | |
1356 | ||
1357 | msk = mptcp_sk(parent); | |
1358 | if (state & TCPF_LISTEN) { | |
1359 | /* MPJ subflow are removed from accept queue before reaching here, | |
1360 | * avoid stray wakeups | |
1361 | */ | |
1362 | if (reqsk_queue_empty(&inet_csk(sk)->icsk_accept_queue)) | |
1363 | return; | |
1364 | ||
499ada50 PA |
1365 | parent->sk_data_ready(parent); |
1366 | return; | |
1367 | } | |
1368 | ||
1369 | WARN_ON_ONCE(!__mptcp_check_fallback(msk) && !subflow->mp_capable && | |
1370 | !subflow->mp_join && !(state & TCPF_CLOSE)); | |
1371 | ||
1372 | if (mptcp_subflow_data_available(sk)) | |
1373 | mptcp_data_ready(parent, sk); | |
1374 | else if (unlikely(sk->sk_err)) | |
1375 | subflow_error_report(sk); | |
1376 | } | |
1377 | ||
1378 | static void subflow_write_space(struct sock *ssk) | |
1379 | { | |
1380 | struct sock *sk = mptcp_subflow_ctx(ssk)->conn; | |
1381 | ||
1382 | mptcp_propagate_sndbuf(sk, ssk); | |
1383 | mptcp_write_space(sk); | |
1384 | } | |
1385 | ||
51fa7f8e | 1386 | static const struct inet_connection_sock_af_ops * |
cec37a6e PK |
1387 | subflow_default_af_ops(struct sock *sk) |
1388 | { | |
1389 | #if IS_ENABLED(CONFIG_MPTCP_IPV6) | |
1390 | if (sk->sk_family == AF_INET6) | |
1391 | return &subflow_v6_specific; | |
1392 | #endif | |
1393 | return &subflow_specific; | |
1394 | } | |
1395 | ||
cec37a6e | 1396 | #if IS_ENABLED(CONFIG_MPTCP_IPV6) |
31484d56 GU |
1397 | void mptcpv6_handle_mapped(struct sock *sk, bool mapped) |
1398 | { | |
cec37a6e PK |
1399 | struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); |
1400 | struct inet_connection_sock *icsk = inet_csk(sk); | |
51fa7f8e | 1401 | const struct inet_connection_sock_af_ops *target; |
cec37a6e PK |
1402 | |
1403 | target = mapped ? &subflow_v6m_specific : subflow_default_af_ops(sk); | |
1404 | ||
1405 | pr_debug("subflow=%p family=%d ops=%p target=%p mapped=%d", | |
edc7e489 | 1406 | subflow, sk->sk_family, icsk->icsk_af_ops, target, mapped); |
cec37a6e PK |
1407 | |
1408 | if (likely(icsk->icsk_af_ops == target)) | |
1409 | return; | |
1410 | ||
1411 | subflow->icsk_af_ops = icsk->icsk_af_ops; | |
1412 | icsk->icsk_af_ops = target; | |
cec37a6e | 1413 | } |
31484d56 | 1414 | #endif |
cec37a6e | 1415 | |
1729cf18 GT |
1416 | void mptcp_info2sockaddr(const struct mptcp_addr_info *info, |
1417 | struct sockaddr_storage *addr, | |
1418 | unsigned short family) | |
ec3edaa7 PK |
1419 | { |
1420 | memset(addr, 0, sizeof(*addr)); | |
50a13bc3 | 1421 | addr->ss_family = family; |
ec3edaa7 PK |
1422 | if (addr->ss_family == AF_INET) { |
1423 | struct sockaddr_in *in_addr = (struct sockaddr_in *)addr; | |
1424 | ||
50a13bc3 MB |
1425 | if (info->family == AF_INET) |
1426 | in_addr->sin_addr = info->addr; | |
1427 | #if IS_ENABLED(CONFIG_MPTCP_IPV6) | |
1428 | else if (ipv6_addr_v4mapped(&info->addr6)) | |
1429 | in_addr->sin_addr.s_addr = info->addr6.s6_addr32[3]; | |
1430 | #endif | |
ec3edaa7 PK |
1431 | in_addr->sin_port = info->port; |
1432 | } | |
1433 | #if IS_ENABLED(CONFIG_MPTCP_IPV6) | |
1434 | else if (addr->ss_family == AF_INET6) { | |
1435 | struct sockaddr_in6 *in6_addr = (struct sockaddr_in6 *)addr; | |
1436 | ||
50a13bc3 MB |
1437 | if (info->family == AF_INET) |
1438 | ipv6_addr_set_v4mapped(info->addr.s_addr, | |
1439 | &in6_addr->sin6_addr); | |
1440 | else | |
1441 | in6_addr->sin6_addr = info->addr6; | |
ec3edaa7 PK |
1442 | in6_addr->sin6_port = info->port; |
1443 | } | |
1444 | #endif | |
1445 | } | |
1446 | ||
ef0da3b8 | 1447 | int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc, |
ee285257 | 1448 | const struct mptcp_addr_info *remote) |
ec3edaa7 PK |
1449 | { |
1450 | struct mptcp_sock *msk = mptcp_sk(sk); | |
1451 | struct mptcp_subflow_context *subflow; | |
1452 | struct sockaddr_storage addr; | |
2ff0e566 | 1453 | int remote_id = remote->id; |
6bad912b | 1454 | int local_id = loc->id; |
95d68651 | 1455 | int err = -ENOTCONN; |
ec3edaa7 | 1456 | struct socket *sf; |
6bad912b | 1457 | struct sock *ssk; |
ec3edaa7 PK |
1458 | u32 remote_token; |
1459 | int addrlen; | |
ee285257 GT |
1460 | int ifindex; |
1461 | u8 flags; | |
ec3edaa7 | 1462 | |
b93df08c | 1463 | if (!mptcp_is_fully_established(sk)) |
95d68651 | 1464 | goto err_out; |
ec3edaa7 PK |
1465 | |
1466 | err = mptcp_subflow_create_socket(sk, &sf); | |
1467 | if (err) | |
95d68651 | 1468 | goto err_out; |
ec3edaa7 | 1469 | |
6bad912b PA |
1470 | ssk = sf->sk; |
1471 | subflow = mptcp_subflow_ctx(ssk); | |
1472 | do { | |
1473 | get_random_bytes(&subflow->local_nonce, sizeof(u32)); | |
1474 | } while (!subflow->local_nonce); | |
1475 | ||
4cf86ae8 PA |
1476 | if (local_id) |
1477 | subflow_set_local_id(subflow, local_id); | |
6bad912b | 1478 | |
8b201370 | 1479 | mptcp_pm_get_flags_and_ifindex_by_id(msk, local_id, |
ee285257 | 1480 | &flags, &ifindex); |
ec3edaa7 PK |
1481 | subflow->remote_key = msk->remote_key; |
1482 | subflow->local_key = msk->local_key; | |
1483 | subflow->token = msk->token; | |
50a13bc3 | 1484 | mptcp_info2sockaddr(loc, &addr, ssk->sk_family); |
ec3edaa7 PK |
1485 | |
1486 | addrlen = sizeof(struct sockaddr_in); | |
1487 | #if IS_ENABLED(CONFIG_MPTCP_IPV6) | |
50a13bc3 | 1488 | if (addr.ss_family == AF_INET6) |
ec3edaa7 PK |
1489 | addrlen = sizeof(struct sockaddr_in6); |
1490 | #endif | |
c9406a23 FW |
1491 | mptcp_sockopt_sync(msk, ssk); |
1492 | ||
daa83ab0 | 1493 | ssk->sk_bound_dev_if = ifindex; |
ec3edaa7 PK |
1494 | err = kernel_bind(sf, (struct sockaddr *)&addr, addrlen); |
1495 | if (err) | |
1496 | goto failed; | |
1497 | ||
1498 | mptcp_crypto_key_sha(subflow->remote_key, &remote_token, NULL); | |
2ff0e566 GT |
1499 | pr_debug("msk=%p remote_token=%u local_id=%d remote_id=%d", msk, |
1500 | remote_token, local_id, remote_id); | |
ec3edaa7 | 1501 | subflow->remote_token = remote_token; |
2ff0e566 | 1502 | subflow->remote_id = remote_id; |
ec3edaa7 | 1503 | subflow->request_join = 1; |
daa83ab0 | 1504 | subflow->request_bkup = !!(flags & MPTCP_PM_ADDR_FLAG_BACKUP); |
50a13bc3 | 1505 | mptcp_info2sockaddr(remote, &addr, ssk->sk_family); |
ec3edaa7 | 1506 | |
3e501490 PA |
1507 | sock_hold(ssk); |
1508 | list_add_tail(&subflow->node, &msk->conn_list); | |
ec3edaa7 PK |
1509 | err = kernel_connect(sf, (struct sockaddr *)&addr, addrlen, O_NONBLOCK); |
1510 | if (err && err != -EINPROGRESS) | |
5b950ff4 | 1511 | goto failed_unlink; |
ec3edaa7 | 1512 | |
866f26f2 PA |
1513 | /* discard the subflow socket */ |
1514 | mptcp_sock_graft(ssk, sk->sk_socket); | |
1515 | iput(SOCK_INODE(sf)); | |
0530020a | 1516 | WRITE_ONCE(msk->allow_infinite_fallback, false); |
ec3edaa7 PK |
1517 | return err; |
1518 | ||
5b950ff4 | 1519 | failed_unlink: |
5b950ff4 | 1520 | list_del(&subflow->node); |
f0715779 | 1521 | sock_put(mptcp_subflow_tcp_sock(subflow)); |
ec3edaa7 | 1522 | |
ec3edaa7 | 1523 | failed: |
e16163b6 | 1524 | subflow->disposable = 1; |
ec3edaa7 | 1525 | sock_release(sf); |
95d68651 PA |
1526 | |
1527 | err_out: | |
1528 | /* we account subflows before the creation, and this failures will not | |
1529 | * be caught by sk_state_change() | |
1530 | */ | |
1531 | mptcp_pm_close_subflow(msk); | |
ec3edaa7 PK |
1532 | return err; |
1533 | } | |
1534 | ||
3764b0c5 NR |
1535 | static void mptcp_attach_cgroup(struct sock *parent, struct sock *child) |
1536 | { | |
1537 | #ifdef CONFIG_SOCK_CGROUP_DATA | |
1538 | struct sock_cgroup_data *parent_skcd = &parent->sk_cgrp_data, | |
1539 | *child_skcd = &child->sk_cgrp_data; | |
1540 | ||
1541 | /* only the additional subflows created by kworkers have to be modified */ | |
1542 | if (cgroup_id(sock_cgroup_ptr(parent_skcd)) != | |
1543 | cgroup_id(sock_cgroup_ptr(child_skcd))) { | |
1544 | #ifdef CONFIG_MEMCG | |
1545 | struct mem_cgroup *memcg = parent->sk_memcg; | |
1546 | ||
1547 | mem_cgroup_sk_free(child); | |
1548 | if (memcg && css_tryget(&memcg->css)) | |
1549 | child->sk_memcg = memcg; | |
1550 | #endif /* CONFIG_MEMCG */ | |
1551 | ||
1552 | cgroup_sk_free(child_skcd); | |
1553 | *child_skcd = *parent_skcd; | |
1554 | cgroup_sk_clone(child_skcd); | |
1555 | } | |
1556 | #endif /* CONFIG_SOCK_CGROUP_DATA */ | |
1557 | } | |
1558 | ||
b19bc294 PA |
1559 | static void mptcp_subflow_ops_override(struct sock *ssk) |
1560 | { | |
1561 | #if IS_ENABLED(CONFIG_MPTCP_IPV6) | |
1562 | if (ssk->sk_prot == &tcpv6_prot) | |
1563 | ssk->sk_prot = &tcpv6_prot_override; | |
1564 | else | |
1565 | #endif | |
1566 | ssk->sk_prot = &tcp_prot_override; | |
1567 | } | |
1568 | ||
1569 | static void mptcp_subflow_ops_undo_override(struct sock *ssk) | |
1570 | { | |
1571 | #if IS_ENABLED(CONFIG_MPTCP_IPV6) | |
1572 | if (ssk->sk_prot == &tcpv6_prot_override) | |
1573 | ssk->sk_prot = &tcpv6_prot; | |
1574 | else | |
1575 | #endif | |
1576 | ssk->sk_prot = &tcp_prot; | |
1577 | } | |
2303f994 PK |
1578 | int mptcp_subflow_create_socket(struct sock *sk, struct socket **new_sock) |
1579 | { | |
1580 | struct mptcp_subflow_context *subflow; | |
1581 | struct net *net = sock_net(sk); | |
1582 | struct socket *sf; | |
1583 | int err; | |
1584 | ||
adf73410 PA |
1585 | /* un-accepted server sockets can reach here - on bad configuration |
1586 | * bail early to avoid greater trouble later | |
1587 | */ | |
1588 | if (unlikely(!sk->sk_socket)) | |
1589 | return -EINVAL; | |
1590 | ||
cec37a6e PK |
1591 | err = sock_create_kern(net, sk->sk_family, SOCK_STREAM, IPPROTO_TCP, |
1592 | &sf); | |
2303f994 PK |
1593 | if (err) |
1594 | return err; | |
1595 | ||
1596 | lock_sock(sf->sk); | |
1597 | ||
3764b0c5 NR |
1598 | /* the newly created socket has to be in the same cgroup as its parent */ |
1599 | mptcp_attach_cgroup(sk, sf->sk); | |
1600 | ||
2303f994 PK |
1601 | /* kernel sockets do not by default acquire net ref, but TCP timer |
1602 | * needs it. | |
1603 | */ | |
1604 | sf->sk->sk_net_refcnt = 1; | |
1d2f3d3c | 1605 | get_net_track(net, &sf->sk->ns_tracker, GFP_KERNEL); |
d477eb90 | 1606 | sock_inuse_add(net, 1); |
2303f994 PK |
1607 | err = tcp_set_ulp(sf->sk, "mptcp"); |
1608 | release_sock(sf->sk); | |
1609 | ||
b8ad540d WY |
1610 | if (err) { |
1611 | sock_release(sf); | |
2303f994 | 1612 | return err; |
b8ad540d | 1613 | } |
2303f994 | 1614 | |
7d14b0d2 PA |
1615 | /* the newly created socket really belongs to the owning MPTCP master |
1616 | * socket, even if for additional subflows the allocation is performed | |
1617 | * by a kernel workqueue. Adjust inode references, so that the | |
1618 | * procfs/diag interaces really show this one belonging to the correct | |
1619 | * user. | |
1620 | */ | |
1621 | SOCK_INODE(sf)->i_ino = SOCK_INODE(sk->sk_socket)->i_ino; | |
1622 | SOCK_INODE(sf)->i_uid = SOCK_INODE(sk->sk_socket)->i_uid; | |
1623 | SOCK_INODE(sf)->i_gid = SOCK_INODE(sk->sk_socket)->i_gid; | |
1624 | ||
2303f994 PK |
1625 | subflow = mptcp_subflow_ctx(sf->sk); |
1626 | pr_debug("subflow=%p", subflow); | |
1627 | ||
1628 | *new_sock = sf; | |
79c0949e | 1629 | sock_hold(sk); |
2303f994 | 1630 | subflow->conn = sk; |
b19bc294 | 1631 | mptcp_subflow_ops_override(sf->sk); |
2303f994 PK |
1632 | |
1633 | return 0; | |
1634 | } | |
1635 | ||
1636 | static struct mptcp_subflow_context *subflow_create_ctx(struct sock *sk, | |
1637 | gfp_t priority) | |
1638 | { | |
1639 | struct inet_connection_sock *icsk = inet_csk(sk); | |
1640 | struct mptcp_subflow_context *ctx; | |
1641 | ||
1642 | ctx = kzalloc(sizeof(*ctx), priority); | |
1643 | if (!ctx) | |
1644 | return NULL; | |
1645 | ||
1646 | rcu_assign_pointer(icsk->icsk_ulp_data, ctx); | |
cec37a6e | 1647 | INIT_LIST_HEAD(&ctx->node); |
b19bc294 | 1648 | INIT_LIST_HEAD(&ctx->delegated_node); |
2303f994 PK |
1649 | |
1650 | pr_debug("subflow=%p", ctx); | |
1651 | ||
1652 | ctx->tcp_sock = sk; | |
1653 | ||
1654 | return ctx; | |
1655 | } | |
1656 | ||
648ef4b8 MM |
1657 | static void __subflow_state_change(struct sock *sk) |
1658 | { | |
1659 | struct socket_wq *wq; | |
1660 | ||
1661 | rcu_read_lock(); | |
1662 | wq = rcu_dereference(sk->sk_wq); | |
1663 | if (skwq_has_sleeper(wq)) | |
1664 | wake_up_interruptible_all(&wq->wait); | |
1665 | rcu_read_unlock(); | |
1666 | } | |
1667 | ||
1668 | static bool subflow_is_done(const struct sock *sk) | |
1669 | { | |
1670 | return sk->sk_shutdown & RCV_SHUTDOWN || sk->sk_state == TCP_CLOSE; | |
1671 | } | |
1672 | ||
1673 | static void subflow_state_change(struct sock *sk) | |
1674 | { | |
1675 | struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); | |
dc093db5 | 1676 | struct sock *parent = subflow->conn; |
648ef4b8 MM |
1677 | |
1678 | __subflow_state_change(sk); | |
1679 | ||
8fd73804 | 1680 | if (subflow_simultaneous_connect(sk)) { |
5cf92bba | 1681 | mptcp_propagate_sndbuf(parent, sk); |
8fd73804 | 1682 | mptcp_do_fallback(sk); |
a6b118fe | 1683 | mptcp_rcv_space_init(mptcp_sk(parent), sk); |
8fd73804 DC |
1684 | pr_fallback(mptcp_sk(parent)); |
1685 | subflow->conn_finished = 1; | |
490274b4 | 1686 | mptcp_set_connected(parent); |
8fd73804 DC |
1687 | } |
1688 | ||
648ef4b8 MM |
1689 | /* as recvmsg() does not acquire the subflow socket for ssk selection |
1690 | * a fin packet carrying a DSS can be unnoticed if we don't trigger | |
1691 | * the data available machinery here. | |
1692 | */ | |
e1ff9e82 | 1693 | if (mptcp_subflow_data_available(sk)) |
2e52213c | 1694 | mptcp_data_ready(parent, sk); |
499ada50 PA |
1695 | else if (unlikely(sk->sk_err)) |
1696 | subflow_error_report(sk); | |
648ef4b8 | 1697 | |
40947e13 FW |
1698 | subflow_sched_work_if_closed(mptcp_sk(parent), sk); |
1699 | ||
067a0b3d | 1700 | if (__mptcp_check_fallback(mptcp_sk(parent)) && |
648ef4b8 MM |
1701 | !subflow->rx_eof && subflow_is_done(sk)) { |
1702 | subflow->rx_eof = 1; | |
59832e24 | 1703 | mptcp_subflow_eof(parent); |
648ef4b8 MM |
1704 | } |
1705 | } | |
1706 | ||
2303f994 PK |
1707 | static int subflow_ulp_init(struct sock *sk) |
1708 | { | |
cec37a6e | 1709 | struct inet_connection_sock *icsk = inet_csk(sk); |
2303f994 PK |
1710 | struct mptcp_subflow_context *ctx; |
1711 | struct tcp_sock *tp = tcp_sk(sk); | |
1712 | int err = 0; | |
1713 | ||
1714 | /* disallow attaching ULP to a socket unless it has been | |
1715 | * created with sock_create_kern() | |
1716 | */ | |
1717 | if (!sk->sk_kern_sock) { | |
1718 | err = -EOPNOTSUPP; | |
1719 | goto out; | |
1720 | } | |
1721 | ||
1722 | ctx = subflow_create_ctx(sk, GFP_KERNEL); | |
1723 | if (!ctx) { | |
1724 | err = -ENOMEM; | |
1725 | goto out; | |
1726 | } | |
1727 | ||
1728 | pr_debug("subflow=%p, family=%d", ctx, sk->sk_family); | |
1729 | ||
1730 | tp->is_mptcp = 1; | |
cec37a6e PK |
1731 | ctx->icsk_af_ops = icsk->icsk_af_ops; |
1732 | icsk->icsk_af_ops = subflow_default_af_ops(sk); | |
648ef4b8 | 1733 | ctx->tcp_state_change = sk->sk_state_change; |
15cc1045 | 1734 | ctx->tcp_error_report = sk->sk_error_report; |
952382c6 FW |
1735 | |
1736 | WARN_ON_ONCE(sk->sk_data_ready != sock_def_readable); | |
1737 | WARN_ON_ONCE(sk->sk_write_space != sk_stream_write_space); | |
1738 | ||
648ef4b8 MM |
1739 | sk->sk_data_ready = subflow_data_ready; |
1740 | sk->sk_write_space = subflow_write_space; | |
1741 | sk->sk_state_change = subflow_state_change; | |
15cc1045 | 1742 | sk->sk_error_report = subflow_error_report; |
2303f994 PK |
1743 | out: |
1744 | return err; | |
1745 | } | |
1746 | ||
e16163b6 | 1747 | static void subflow_ulp_release(struct sock *ssk) |
2303f994 | 1748 | { |
e16163b6 PA |
1749 | struct mptcp_subflow_context *ctx = mptcp_subflow_ctx(ssk); |
1750 | bool release = true; | |
1751 | struct sock *sk; | |
2303f994 PK |
1752 | |
1753 | if (!ctx) | |
1754 | return; | |
1755 | ||
e16163b6 PA |
1756 | sk = ctx->conn; |
1757 | if (sk) { | |
1758 | /* if the msk has been orphaned, keep the ctx | |
0597d0f8 PA |
1759 | * alive, will be freed by __mptcp_close_ssk(), |
1760 | * when the subflow is still unaccepted | |
e16163b6 | 1761 | */ |
0597d0f8 | 1762 | release = ctx->disposable || list_empty(&ctx->node); |
e16163b6 PA |
1763 | sock_put(sk); |
1764 | } | |
79c0949e | 1765 | |
b19bc294 | 1766 | mptcp_subflow_ops_undo_override(ssk); |
e16163b6 PA |
1767 | if (release) |
1768 | kfree_rcu(ctx, rcu); | |
2303f994 PK |
1769 | } |
1770 | ||
cec37a6e PK |
1771 | static void subflow_ulp_clone(const struct request_sock *req, |
1772 | struct sock *newsk, | |
1773 | const gfp_t priority) | |
1774 | { | |
1775 | struct mptcp_subflow_request_sock *subflow_req = mptcp_subflow_rsk(req); | |
1776 | struct mptcp_subflow_context *old_ctx = mptcp_subflow_ctx(newsk); | |
1777 | struct mptcp_subflow_context *new_ctx; | |
1778 | ||
f296234c PK |
1779 | if (!tcp_rsk(req)->is_mptcp || |
1780 | (!subflow_req->mp_capable && !subflow_req->mp_join)) { | |
648ef4b8 | 1781 | subflow_ulp_fallback(newsk, old_ctx); |
cec37a6e PK |
1782 | return; |
1783 | } | |
1784 | ||
1785 | new_ctx = subflow_create_ctx(newsk, priority); | |
edc7e489 | 1786 | if (!new_ctx) { |
648ef4b8 | 1787 | subflow_ulp_fallback(newsk, old_ctx); |
cec37a6e PK |
1788 | return; |
1789 | } | |
1790 | ||
1791 | new_ctx->conn_finished = 1; | |
1792 | new_ctx->icsk_af_ops = old_ctx->icsk_af_ops; | |
648ef4b8 | 1793 | new_ctx->tcp_state_change = old_ctx->tcp_state_change; |
15cc1045 | 1794 | new_ctx->tcp_error_report = old_ctx->tcp_error_report; |
58b09919 PA |
1795 | new_ctx->rel_write_seq = 1; |
1796 | new_ctx->tcp_sock = newsk; | |
1797 | ||
f296234c PK |
1798 | if (subflow_req->mp_capable) { |
1799 | /* see comments in subflow_syn_recv_sock(), MPTCP connection | |
1800 | * is fully established only after we receive the remote key | |
1801 | */ | |
1802 | new_ctx->mp_capable = 1; | |
f296234c PK |
1803 | new_ctx->local_key = subflow_req->local_key; |
1804 | new_ctx->token = subflow_req->token; | |
1805 | new_ctx->ssn_offset = subflow_req->ssn_offset; | |
1806 | new_ctx->idsn = subflow_req->idsn; | |
4cf86ae8 PA |
1807 | |
1808 | /* this is the first subflow, id is always 0 */ | |
1809 | new_ctx->local_id_valid = 1; | |
f296234c | 1810 | } else if (subflow_req->mp_join) { |
ec3edaa7 | 1811 | new_ctx->ssn_offset = subflow_req->ssn_offset; |
f296234c PK |
1812 | new_ctx->mp_join = 1; |
1813 | new_ctx->fully_established = 1; | |
1814 | new_ctx->backup = subflow_req->backup; | |
2ff0e566 | 1815 | new_ctx->remote_id = subflow_req->remote_id; |
f296234c PK |
1816 | new_ctx->token = subflow_req->token; |
1817 | new_ctx->thmac = subflow_req->thmac; | |
4cf86ae8 PA |
1818 | |
1819 | /* the subflow req id is valid, fetched via subflow_check_req() | |
1820 | * and subflow_token_join_request() | |
1821 | */ | |
1822 | subflow_set_local_id(new_ctx, subflow_req->local_id); | |
f296234c | 1823 | } |
cec37a6e PK |
1824 | } |
1825 | ||
b19bc294 PA |
1826 | static void tcp_release_cb_override(struct sock *ssk) |
1827 | { | |
1828 | struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk); | |
1829 | ||
1830 | if (mptcp_subflow_has_delegated_action(subflow)) | |
1831 | mptcp_subflow_process_delegated(ssk); | |
1832 | ||
1833 | tcp_release_cb(ssk); | |
1834 | } | |
1835 | ||
2303f994 PK |
1836 | static struct tcp_ulp_ops subflow_ulp_ops __read_mostly = { |
1837 | .name = "mptcp", | |
1838 | .owner = THIS_MODULE, | |
1839 | .init = subflow_ulp_init, | |
1840 | .release = subflow_ulp_release, | |
cec37a6e | 1841 | .clone = subflow_ulp_clone, |
2303f994 PK |
1842 | }; |
1843 | ||
cec37a6e PK |
1844 | static int subflow_ops_init(struct request_sock_ops *subflow_ops) |
1845 | { | |
1846 | subflow_ops->obj_size = sizeof(struct mptcp_subflow_request_sock); | |
1847 | subflow_ops->slab_name = "request_sock_subflow"; | |
1848 | ||
1849 | subflow_ops->slab = kmem_cache_create(subflow_ops->slab_name, | |
1850 | subflow_ops->obj_size, 0, | |
1851 | SLAB_ACCOUNT | | |
1852 | SLAB_TYPESAFE_BY_RCU, | |
1853 | NULL); | |
1854 | if (!subflow_ops->slab) | |
1855 | return -ENOMEM; | |
1856 | ||
79c0949e PK |
1857 | subflow_ops->destructor = subflow_req_destructor; |
1858 | ||
cec37a6e PK |
1859 | return 0; |
1860 | } | |
1861 | ||
d39dceca | 1862 | void __init mptcp_subflow_init(void) |
2303f994 | 1863 | { |
08b8d080 FW |
1864 | mptcp_subflow_request_sock_ops = tcp_request_sock_ops; |
1865 | if (subflow_ops_init(&mptcp_subflow_request_sock_ops) != 0) | |
cec37a6e PK |
1866 | panic("MPTCP: failed to init subflow request sock ops\n"); |
1867 | ||
1868 | subflow_request_sock_ipv4_ops = tcp_request_sock_ipv4_ops; | |
7ea851d1 | 1869 | subflow_request_sock_ipv4_ops.route_req = subflow_v4_route_req; |
cec37a6e PK |
1870 | |
1871 | subflow_specific = ipv4_specific; | |
1872 | subflow_specific.conn_request = subflow_v4_conn_request; | |
1873 | subflow_specific.syn_recv_sock = subflow_syn_recv_sock; | |
1874 | subflow_specific.sk_rx_dst_set = subflow_finish_connect; | |
4cf86ae8 | 1875 | subflow_specific.rebuild_header = subflow_rebuild_header; |
cec37a6e | 1876 | |
b19bc294 PA |
1877 | tcp_prot_override = tcp_prot; |
1878 | tcp_prot_override.release_cb = tcp_release_cb_override; | |
1879 | ||
cec37a6e PK |
1880 | #if IS_ENABLED(CONFIG_MPTCP_IPV6) |
1881 | subflow_request_sock_ipv6_ops = tcp_request_sock_ipv6_ops; | |
7ea851d1 | 1882 | subflow_request_sock_ipv6_ops.route_req = subflow_v6_route_req; |
cec37a6e PK |
1883 | |
1884 | subflow_v6_specific = ipv6_specific; | |
1885 | subflow_v6_specific.conn_request = subflow_v6_conn_request; | |
1886 | subflow_v6_specific.syn_recv_sock = subflow_syn_recv_sock; | |
1887 | subflow_v6_specific.sk_rx_dst_set = subflow_finish_connect; | |
4cf86ae8 | 1888 | subflow_v6_specific.rebuild_header = subflow_v6_rebuild_header; |
cec37a6e PK |
1889 | |
1890 | subflow_v6m_specific = subflow_v6_specific; | |
1891 | subflow_v6m_specific.queue_xmit = ipv4_specific.queue_xmit; | |
1892 | subflow_v6m_specific.send_check = ipv4_specific.send_check; | |
1893 | subflow_v6m_specific.net_header_len = ipv4_specific.net_header_len; | |
1894 | subflow_v6m_specific.mtu_reduced = ipv4_specific.mtu_reduced; | |
1895 | subflow_v6m_specific.net_frag_header_len = 0; | |
4cf86ae8 | 1896 | subflow_v6m_specific.rebuild_header = subflow_rebuild_header; |
b19bc294 PA |
1897 | |
1898 | tcpv6_prot_override = tcpv6_prot; | |
1899 | tcpv6_prot_override.release_cb = tcp_release_cb_override; | |
cec37a6e PK |
1900 | #endif |
1901 | ||
5147dfb5 DC |
1902 | mptcp_diag_subflow_init(&subflow_ulp_ops); |
1903 | ||
2303f994 PK |
1904 | if (tcp_register_ulp(&subflow_ulp_ops) != 0) |
1905 | panic("MPTCP: failed to register subflows to ULP\n"); | |
1906 | } |