Merge branch 'net-ipa-support-variable-rx-buffer-size'
[linux-2.6-block.git] / net / mptcp / options.c
CommitLineData
eda7acdd
PK
1// SPDX-License-Identifier: GPL-2.0
2/* Multipath TCP
3 *
4 * Copyright (c) 2017 - 2019, Intel Corporation.
5 */
6
c85adced
GT
7#define pr_fmt(fmt) "MPTCP: " fmt
8
eda7acdd 9#include <linux/kernel.h>
a24d22b2 10#include <crypto/sha2.h>
eda7acdd
PK
11#include <net/tcp.h>
12#include <net/mptcp.h>
13#include "protocol.h"
a877de06 14#include "mib.h"
eda7acdd 15
ed66bfb4
GT
16#include <trace/events/mptcp.h>
17
65492c5a
PA
18static bool mptcp_cap_flag_sha256(u8 flags)
19{
20 return (flags & MPTCP_CAP_FLAG_MASK) == MPTCP_CAP_HMAC_SHA256;
21}
22
cfde141e
PA
23static void mptcp_parse_option(const struct sk_buff *skb,
24 const unsigned char *ptr, int opsize,
25 struct mptcp_options_received *mp_opt)
eda7acdd 26{
eda7acdd 27 u8 subtype = *ptr >> 4;
648ef4b8 28 int expected_opsize;
eda7acdd
PK
29 u8 version;
30 u8 flags;
5c4a824d 31 u8 i;
eda7acdd
PK
32
33 switch (subtype) {
34 case MPTCPOPT_MP_CAPABLE:
cc7972ea
CP
35 /* strict size checking */
36 if (!(TCP_SKB_CB(skb)->tcp_flags & TCPHDR_SYN)) {
37 if (skb->len > tcp_hdr(skb)->doff << 2)
38 expected_opsize = TCPOLEN_MPTCP_MPC_ACK_DATA;
39 else
40 expected_opsize = TCPOLEN_MPTCP_MPC_ACK;
41 } else {
42 if (TCP_SKB_CB(skb)->tcp_flags & TCPHDR_ACK)
43 expected_opsize = TCPOLEN_MPTCP_MPC_SYNACK;
44 else
45 expected_opsize = TCPOLEN_MPTCP_MPC_SYN;
46 }
208e8f66
GT
47
48 /* Cfr RFC 8684 Section 3.3.0:
49 * If a checksum is present but its use had
50 * not been negotiated in the MP_CAPABLE handshake, the receiver MUST
51 * close the subflow with a RST, as it is not behaving as negotiated.
52 * If a checksum is not present when its use has been negotiated, the
53 * receiver MUST close the subflow with a RST, as it is considered
54 * broken
55 * We parse even option with mismatching csum presence, so that
56 * later in subflow_data_ready we can trigger the reset.
57 */
58 if (opsize != expected_opsize &&
59 (expected_opsize != TCPOLEN_MPTCP_MPC_ACK_DATA ||
60 opsize != TCPOLEN_MPTCP_MPC_ACK_DATA_CSUM))
eda7acdd
PK
61 break;
62
cc7972ea 63 /* try to be gentle vs future versions on the initial syn */
eda7acdd 64 version = *ptr++ & MPTCP_VERSION_MASK;
cc7972ea
CP
65 if (opsize != TCPOLEN_MPTCP_MPC_SYN) {
66 if (version != MPTCP_SUPPORTED_VERSION)
67 break;
68 } else if (version < MPTCP_SUPPORTED_VERSION) {
eda7acdd 69 break;
cc7972ea 70 }
eda7acdd
PK
71
72 flags = *ptr++;
65492c5a 73 if (!mptcp_cap_flag_sha256(flags) ||
eda7acdd
PK
74 (flags & MPTCP_CAP_EXTENSIBILITY))
75 break;
76
77 /* RFC 6824, Section 3.1:
78 * "For the Checksum Required bit (labeled "A"), if either
79 * host requires the use of checksums, checksums MUST be used.
80 * In other words, the only way for checksums not to be used
81 * is if both hosts in their SYNs set A=0."
eda7acdd
PK
82 */
83 if (flags & MPTCP_CAP_CHECKSUM_REQD)
74c7dfbe 84 mp_opt->suboptions |= OPTION_MPTCP_CSUMREQD;
eda7acdd 85
a086aeba 86 mp_opt->deny_join_id0 = !!(flags & MPTCP_CAP_DENY_JOIN_ID0);
df377be3 87
74c7dfbe 88 mp_opt->suboptions |= OPTIONS_MPTCP_MPC;
cc7972ea
CP
89 if (opsize >= TCPOLEN_MPTCP_MPC_SYNACK) {
90 mp_opt->sndr_key = get_unaligned_be64(ptr);
91 ptr += 8;
92 }
93 if (opsize >= TCPOLEN_MPTCP_MPC_ACK) {
eda7acdd
PK
94 mp_opt->rcvr_key = get_unaligned_be64(ptr);
95 ptr += 8;
eda7acdd 96 }
208e8f66 97 if (opsize >= TCPOLEN_MPTCP_MPC_ACK_DATA) {
cc7972ea
CP
98 /* Section 3.1.:
99 * "the data parameters in a MP_CAPABLE are semantically
100 * equivalent to those in a DSS option and can be used
101 * interchangeably."
102 */
74c7dfbe 103 mp_opt->suboptions |= OPTION_MPTCP_DSS;
cc7972ea
CP
104 mp_opt->use_map = 1;
105 mp_opt->mpc_map = 1;
106 mp_opt->data_len = get_unaligned_be16(ptr);
107 ptr += 2;
108 }
208e8f66
GT
109 if (opsize == TCPOLEN_MPTCP_MPC_ACK_DATA_CSUM) {
110 mp_opt->csum = (__force __sum16)get_unaligned_be16(ptr);
74c7dfbe 111 mp_opt->suboptions |= OPTION_MPTCP_CSUMREQD;
208e8f66
GT
112 ptr += 2;
113 }
114 pr_debug("MP_CAPABLE version=%x, flags=%x, optlen=%d sndr=%llu, rcvr=%llu len=%d csum=%u",
cc7972ea 115 version, flags, opsize, mp_opt->sndr_key,
208e8f66 116 mp_opt->rcvr_key, mp_opt->data_len, mp_opt->csum);
eda7acdd
PK
117 break;
118
f296234c 119 case MPTCPOPT_MP_JOIN:
74c7dfbe 120 mp_opt->suboptions |= OPTIONS_MPTCP_MPJ;
f296234c
PK
121 if (opsize == TCPOLEN_MPTCP_MPJ_SYN) {
122 mp_opt->backup = *ptr++ & MPTCPOPT_BACKUP;
123 mp_opt->join_id = *ptr++;
124 mp_opt->token = get_unaligned_be32(ptr);
125 ptr += 4;
126 mp_opt->nonce = get_unaligned_be32(ptr);
127 ptr += 4;
128 pr_debug("MP_JOIN bkup=%u, id=%u, token=%u, nonce=%u",
129 mp_opt->backup, mp_opt->join_id,
130 mp_opt->token, mp_opt->nonce);
131 } else if (opsize == TCPOLEN_MPTCP_MPJ_SYNACK) {
132 mp_opt->backup = *ptr++ & MPTCPOPT_BACKUP;
133 mp_opt->join_id = *ptr++;
134 mp_opt->thmac = get_unaligned_be64(ptr);
135 ptr += 8;
136 mp_opt->nonce = get_unaligned_be32(ptr);
137 ptr += 4;
138 pr_debug("MP_JOIN bkup=%u, id=%u, thmac=%llu, nonce=%u",
139 mp_opt->backup, mp_opt->join_id,
140 mp_opt->thmac, mp_opt->nonce);
141 } else if (opsize == TCPOLEN_MPTCP_MPJ_ACK) {
142 ptr += 2;
143 memcpy(mp_opt->hmac, ptr, MPTCPOPT_HMAC_LEN);
144 pr_debug("MP_JOIN hmac");
145 } else {
74c7dfbe 146 mp_opt->suboptions &= ~OPTIONS_MPTCP_MPJ;
f296234c
PK
147 }
148 break;
149
eda7acdd
PK
150 case MPTCPOPT_DSS:
151 pr_debug("DSS");
648ef4b8
MM
152 ptr++;
153
cc7972ea
CP
154 /* we must clear 'mpc_map' be able to detect MP_CAPABLE
155 * map vs DSS map in mptcp_incoming_options(), and reconstruct
156 * map info accordingly
157 */
158 mp_opt->mpc_map = 0;
648ef4b8
MM
159 flags = (*ptr++) & MPTCP_DSS_FLAG_MASK;
160 mp_opt->data_fin = (flags & MPTCP_DSS_DATA_FIN) != 0;
161 mp_opt->dsn64 = (flags & MPTCP_DSS_DSN64) != 0;
162 mp_opt->use_map = (flags & MPTCP_DSS_HAS_MAP) != 0;
163 mp_opt->ack64 = (flags & MPTCP_DSS_ACK64) != 0;
164 mp_opt->use_ack = (flags & MPTCP_DSS_HAS_ACK);
165
166 pr_debug("data_fin=%d dsn64=%d use_map=%d ack64=%d use_ack=%d",
167 mp_opt->data_fin, mp_opt->dsn64,
168 mp_opt->use_map, mp_opt->ack64,
169 mp_opt->use_ack);
170
171 expected_opsize = TCPOLEN_MPTCP_DSS_BASE;
172
173 if (mp_opt->use_ack) {
174 if (mp_opt->ack64)
175 expected_opsize += TCPOLEN_MPTCP_DSS_ACK64;
176 else
177 expected_opsize += TCPOLEN_MPTCP_DSS_ACK32;
178 }
179
180 if (mp_opt->use_map) {
181 if (mp_opt->dsn64)
182 expected_opsize += TCPOLEN_MPTCP_DSS_MAP64;
183 else
184 expected_opsize += TCPOLEN_MPTCP_DSS_MAP32;
185 }
186
390b95a5
GT
187 /* Always parse any csum presence combination, we will enforce
188 * RFC 8684 Section 3.3.0 checks later in subflow_data_ready
648ef4b8
MM
189 */
190 if (opsize != expected_opsize &&
191 opsize != expected_opsize + TCPOLEN_MPTCP_DSS_CHECKSUM)
192 break;
193
74c7dfbe 194 mp_opt->suboptions |= OPTION_MPTCP_DSS;
648ef4b8
MM
195 if (mp_opt->use_ack) {
196 if (mp_opt->ack64) {
197 mp_opt->data_ack = get_unaligned_be64(ptr);
198 ptr += 8;
199 } else {
200 mp_opt->data_ack = get_unaligned_be32(ptr);
201 ptr += 4;
202 }
203
204 pr_debug("data_ack=%llu", mp_opt->data_ack);
205 }
206
207 if (mp_opt->use_map) {
208 if (mp_opt->dsn64) {
209 mp_opt->data_seq = get_unaligned_be64(ptr);
210 ptr += 8;
211 } else {
212 mp_opt->data_seq = get_unaligned_be32(ptr);
213 ptr += 4;
214 }
215
216 mp_opt->subflow_seq = get_unaligned_be32(ptr);
217 ptr += 4;
218
219 mp_opt->data_len = get_unaligned_be16(ptr);
220 ptr += 2;
221
390b95a5 222 if (opsize == expected_opsize + TCPOLEN_MPTCP_DSS_CHECKSUM) {
74c7dfbe 223 mp_opt->suboptions |= OPTION_MPTCP_CSUMREQD;
390b95a5
GT
224 mp_opt->csum = (__force __sum16)get_unaligned_be16(ptr);
225 ptr += 2;
226 }
227
228 pr_debug("data_seq=%llu subflow_seq=%u data_len=%u csum=%d:%u",
648ef4b8 229 mp_opt->data_seq, mp_opt->subflow_seq,
74c7dfbe
PA
230 mp_opt->data_len, !!(mp_opt->suboptions & OPTION_MPTCP_CSUMREQD),
231 mp_opt->csum);
648ef4b8
MM
232 }
233
eda7acdd
PK
234 break;
235
3df523ab
PK
236 case MPTCPOPT_ADD_ADDR:
237 mp_opt->echo = (*ptr++) & MPTCP_ADDR_ECHO;
238 if (!mp_opt->echo) {
239 if (opsize == TCPOLEN_MPTCP_ADD_ADDR ||
240 opsize == TCPOLEN_MPTCP_ADD_ADDR_PORT)
1b1a6ef5 241 mp_opt->addr.family = AF_INET;
3df523ab
PK
242#if IS_ENABLED(CONFIG_MPTCP_IPV6)
243 else if (opsize == TCPOLEN_MPTCP_ADD_ADDR6 ||
244 opsize == TCPOLEN_MPTCP_ADD_ADDR6_PORT)
1b1a6ef5 245 mp_opt->addr.family = AF_INET6;
3df523ab
PK
246#endif
247 else
248 break;
249 } else {
250 if (opsize == TCPOLEN_MPTCP_ADD_ADDR_BASE ||
251 opsize == TCPOLEN_MPTCP_ADD_ADDR_BASE_PORT)
1b1a6ef5 252 mp_opt->addr.family = AF_INET;
3df523ab
PK
253#if IS_ENABLED(CONFIG_MPTCP_IPV6)
254 else if (opsize == TCPOLEN_MPTCP_ADD_ADDR6_BASE ||
255 opsize == TCPOLEN_MPTCP_ADD_ADDR6_BASE_PORT)
1b1a6ef5 256 mp_opt->addr.family = AF_INET6;
3df523ab
PK
257#endif
258 else
259 break;
260 }
261
74c7dfbe 262 mp_opt->suboptions |= OPTION_MPTCP_ADD_ADDR;
f7dafee1 263 mp_opt->addr.id = *ptr++;
a086aeba
PA
264 mp_opt->addr.port = 0;
265 mp_opt->ahmac = 0;
1b1a6ef5 266 if (mp_opt->addr.family == AF_INET) {
f7dafee1 267 memcpy((u8 *)&mp_opt->addr.addr.s_addr, (u8 *)ptr, 4);
3df523ab
PK
268 ptr += 4;
269 if (opsize == TCPOLEN_MPTCP_ADD_ADDR_PORT ||
270 opsize == TCPOLEN_MPTCP_ADD_ADDR_BASE_PORT) {
f7dafee1 271 mp_opt->addr.port = htons(get_unaligned_be16(ptr));
3df523ab
PK
272 ptr += 2;
273 }
274 }
275#if IS_ENABLED(CONFIG_MPTCP_IPV6)
276 else {
f7dafee1 277 memcpy(mp_opt->addr.addr6.s6_addr, (u8 *)ptr, 16);
3df523ab
PK
278 ptr += 16;
279 if (opsize == TCPOLEN_MPTCP_ADD_ADDR6_PORT ||
280 opsize == TCPOLEN_MPTCP_ADD_ADDR6_BASE_PORT) {
f7dafee1 281 mp_opt->addr.port = htons(get_unaligned_be16(ptr));
3df523ab
PK
282 ptr += 2;
283 }
284 }
285#endif
286 if (!mp_opt->echo) {
287 mp_opt->ahmac = get_unaligned_be64(ptr);
288 ptr += 8;
289 }
90a4aea8 290 pr_debug("ADD_ADDR%s: id=%d, ahmac=%llu, echo=%d, port=%d",
1b1a6ef5 291 (mp_opt->addr.family == AF_INET6) ? "6" : "",
f7dafee1 292 mp_opt->addr.id, mp_opt->ahmac, mp_opt->echo, ntohs(mp_opt->addr.port));
3df523ab
PK
293 break;
294
295 case MPTCPOPT_RM_ADDR:
5c4a824d
GT
296 if (opsize < TCPOLEN_MPTCP_RM_ADDR_BASE + 1 ||
297 opsize > TCPOLEN_MPTCP_RM_ADDR_BASE + MPTCP_RM_IDS_MAX)
3df523ab
PK
298 break;
299
8e60eed6
GT
300 ptr++;
301
74c7dfbe 302 mp_opt->suboptions |= OPTION_MPTCP_RM_ADDR;
5c4a824d
GT
303 mp_opt->rm_list.nr = opsize - TCPOLEN_MPTCP_RM_ADDR_BASE;
304 for (i = 0; i < mp_opt->rm_list.nr; i++)
305 mp_opt->rm_list.ids[i] = *ptr++;
306 pr_debug("RM_ADDR: rm_list_nr=%d", mp_opt->rm_list.nr);
3df523ab
PK
307 break;
308
40453a5c
GT
309 case MPTCPOPT_MP_PRIO:
310 if (opsize != TCPOLEN_MPTCP_PRIO)
311 break;
312
74c7dfbe 313 mp_opt->suboptions |= OPTION_MPTCP_PRIO;
40453a5c
GT
314 mp_opt->backup = *ptr++ & MPTCP_PRIO_BKUP;
315 pr_debug("MP_PRIO: prio=%d", mp_opt->backup);
316 break;
317
50c504a2
FW
318 case MPTCPOPT_MP_FASTCLOSE:
319 if (opsize != TCPOLEN_MPTCP_FASTCLOSE)
320 break;
321
322 ptr += 2;
323 mp_opt->rcvr_key = get_unaligned_be64(ptr);
324 ptr += 8;
74c7dfbe 325 mp_opt->suboptions |= OPTION_MPTCP_FASTCLOSE;
50c504a2
FW
326 break;
327
dc87efdb
FW
328 case MPTCPOPT_RST:
329 if (opsize != TCPOLEN_MPTCP_RST)
330 break;
331
332 if (!(TCP_SKB_CB(skb)->tcp_flags & TCPHDR_RST))
333 break;
74c7dfbe
PA
334
335 mp_opt->suboptions |= OPTION_MPTCP_RST;
dc87efdb
FW
336 flags = *ptr++;
337 mp_opt->reset_transient = flags & MPTCP_RST_TRANSIENT;
338 mp_opt->reset_reason = *ptr;
339 break;
340
5580d41b
GT
341 case MPTCPOPT_MP_FAIL:
342 if (opsize != TCPOLEN_MPTCP_FAIL)
343 break;
344
345 ptr += 2;
74c7dfbe 346 mp_opt->suboptions |= OPTION_MPTCP_FAIL;
5580d41b
GT
347 mp_opt->fail_seq = get_unaligned_be64(ptr);
348 pr_debug("MP_FAIL: data_seq=%llu", mp_opt->fail_seq);
349 break;
350
eda7acdd
PK
351 default:
352 break;
353 }
354}
355
c863225b
GT
356void mptcp_get_options(const struct sock *sk,
357 const struct sk_buff *skb,
cfde141e 358 struct mptcp_options_received *mp_opt)
cec37a6e 359{
cec37a6e 360 const struct tcphdr *th = tcp_hdr(skb);
cfde141e
PA
361 const unsigned char *ptr;
362 int length;
cec37a6e 363
cfde141e 364 /* initialize option status */
74c7dfbe 365 mp_opt->suboptions = 0;
cfde141e
PA
366
367 length = (th->doff * 4) - sizeof(struct tcphdr);
cec37a6e
PK
368 ptr = (const unsigned char *)(th + 1);
369
370 while (length > 0) {
371 int opcode = *ptr++;
372 int opsize;
373
374 switch (opcode) {
375 case TCPOPT_EOL:
376 return;
377 case TCPOPT_NOP: /* Ref: RFC 793 section 3.1 */
378 length--;
379 continue;
380 default:
07718be2
MM
381 if (length < 2)
382 return;
cec37a6e
PK
383 opsize = *ptr++;
384 if (opsize < 2) /* "silly options" */
385 return;
386 if (opsize > length)
387 return; /* don't parse partial options */
388 if (opcode == TCPOPT_MPTCP)
cfde141e 389 mptcp_parse_option(skb, ptr, opsize, mp_opt);
cec37a6e
PK
390 ptr += opsize - 2;
391 length -= opsize;
392 }
393 }
394}
395
cc7972ea
CP
396bool mptcp_syn_options(struct sock *sk, const struct sk_buff *skb,
397 unsigned int *size, struct mptcp_out_options *opts)
cec37a6e
PK
398{
399 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
400
cc7972ea
CP
401 /* we will use snd_isn to detect first pkt [re]transmission
402 * in mptcp_established_options_mp()
403 */
404 subflow->snd_isn = TCP_SKB_CB(skb)->end_seq;
cec37a6e 405 if (subflow->request_mptcp) {
cec37a6e 406 opts->suboptions = OPTION_MPTCP_MPC_SYN;
06fe1719 407 opts->csum_reqd = mptcp_is_checksum_enabled(sock_net(sk));
bab6b88e 408 opts->allow_join_id0 = mptcp_allow_join_id0(sock_net(sk));
cec37a6e
PK
409 *size = TCPOLEN_MPTCP_MPC_SYN;
410 return true;
ec3edaa7
PK
411 } else if (subflow->request_join) {
412 pr_debug("remote_token=%u, nonce=%u", subflow->remote_token,
413 subflow->local_nonce);
414 opts->suboptions = OPTION_MPTCP_MPJ_SYN;
415 opts->join_id = subflow->local_id;
416 opts->token = subflow->remote_token;
417 opts->nonce = subflow->local_nonce;
418 opts->backup = subflow->request_bkup;
419 *size = TCPOLEN_MPTCP_MPJ_SYN;
420 return true;
cec37a6e
PK
421 }
422 return false;
423}
424
ec3edaa7
PK
425static void clear_3rdack_retransmission(struct sock *sk)
426{
427 struct inet_connection_sock *icsk = inet_csk(sk);
428
429 sk_stop_timer(sk, &icsk->icsk_delack_timer);
430 icsk->icsk_ack.timeout = 0;
431 icsk->icsk_ack.ato = 0;
432 icsk->icsk_ack.pending &= ~(ICSK_ACK_SCHED | ICSK_ACK_TIMER);
433}
434
cc7972ea 435static bool mptcp_established_options_mp(struct sock *sk, struct sk_buff *skb,
d87903b6 436 bool snd_data_fin_enable,
cc7972ea 437 unsigned int *size,
6d0060f6
MM
438 unsigned int remaining,
439 struct mptcp_out_options *opts)
cec37a6e
PK
440{
441 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
06fe1719 442 struct mptcp_sock *msk = mptcp_sk(subflow->conn);
cc7972ea
CP
443 struct mptcp_ext *mpext;
444 unsigned int data_len;
c94b1f96 445 u8 len;
cc7972ea 446
ec3edaa7
PK
447 /* When skb is not available, we better over-estimate the emitted
448 * options len. A full DSS option (28 bytes) is longer than
449 * TCPOLEN_MPTCP_MPC_ACK_DATA(22) or TCPOLEN_MPTCP_MPJ_ACK(24), so
450 * tell the caller to defer the estimate to
451 * mptcp_established_options_dss(), which will reserve enough space.
452 */
453 if (!skb)
454 return false;
cc7972ea 455
d87903b6
PA
456 /* MPC/MPJ needed only on 3rd ack packet, DATA_FIN and TCP shutdown take precedence */
457 if (subflow->fully_established || snd_data_fin_enable ||
458 subflow->snd_isn != TCP_SKB_CB(skb)->seq ||
459 sk->sk_state != TCP_ESTABLISHED)
ec3edaa7
PK
460 return false;
461
462 if (subflow->mp_capable) {
cc7972ea
CP
463 mpext = mptcp_get_ext(skb);
464 data_len = mpext ? mpext->data_len : 0;
cec37a6e 465
f7cc8890 466 /* we will check ops->data_len in mptcp_write_options() to
cc7972ea
CP
467 * discriminate between TCPOLEN_MPTCP_MPC_ACK_DATA and
468 * TCPOLEN_MPTCP_MPC_ACK
469 */
f7cc8890 470 opts->data_len = data_len;
cec37a6e
PK
471 opts->suboptions = OPTION_MPTCP_MPC_ACK;
472 opts->sndr_key = subflow->local_key;
473 opts->rcvr_key = subflow->remote_key;
06fe1719 474 opts->csum_reqd = READ_ONCE(msk->csum_enabled);
bab6b88e 475 opts->allow_join_id0 = mptcp_allow_join_id0(sock_net(sk));
cc7972ea
CP
476
477 /* Section 3.1.
478 * The MP_CAPABLE option is carried on the SYN, SYN/ACK, and ACK
479 * packets that start the first subflow of an MPTCP connection,
480 * as well as the first packet that carries data
481 */
c94b1f96
GT
482 if (data_len > 0) {
483 len = TCPOLEN_MPTCP_MPC_ACK_DATA;
484 if (opts->csum_reqd) {
c5b39e26 485 /* we need to propagate more info to csum the pseudo hdr */
f7cc8890
DC
486 opts->data_seq = mpext->data_seq;
487 opts->subflow_seq = mpext->subflow_seq;
488 opts->csum = mpext->csum;
c94b1f96
GT
489 len += TCPOLEN_MPTCP_DSS_CHECKSUM;
490 }
491 *size = ALIGN(len, 4);
492 } else {
cc7972ea 493 *size = TCPOLEN_MPTCP_MPC_ACK;
c94b1f96 494 }
cc7972ea
CP
495
496 pr_debug("subflow=%p, local_key=%llu, remote_key=%llu map_len=%d",
497 subflow, subflow->local_key, subflow->remote_key,
498 data_len);
499
cec37a6e 500 return true;
ec3edaa7
PK
501 } else if (subflow->mp_join) {
502 opts->suboptions = OPTION_MPTCP_MPJ_ACK;
503 memcpy(opts->hmac, subflow->hmac, MPTCPOPT_HMAC_LEN);
504 *size = TCPOLEN_MPTCP_MPJ_ACK;
505 pr_debug("subflow=%p", subflow);
506
bcd97734
PA
507 /* we can use the full delegate action helper only from BH context
508 * If we are in process context - sk is flushing the backlog at
509 * socket lock release time - just set the appropriate flag, will
510 * be handled by the release callback
511 */
512 if (sock_owned_by_user(sk))
513 set_bit(MPTCP_DELEGATE_ACK, &subflow->delegated_status);
514 else
515 mptcp_subflow_delegate(subflow, MPTCP_DELEGATE_ACK);
ec3edaa7 516 return true;
cec37a6e
PK
517 }
518 return false;
519}
520
6d0060f6 521static void mptcp_write_data_fin(struct mptcp_subflow_context *subflow,
9c29e361 522 struct sk_buff *skb, struct mptcp_ext *ext)
6d0060f6 523{
017512a0
PA
524 /* The write_seq value has already been incremented, so the actual
525 * sequence number for the DATA_FIN is one less.
526 */
527 u64 data_fin_tx_seq = READ_ONCE(mptcp_sk(subflow->conn)->write_seq) - 1;
7279da61 528
9c29e361 529 if (!ext->use_map || !skb->len) {
6d0060f6
MM
530 /* RFC6824 requires a DSS mapping with specific values
531 * if DATA_FIN is set but no data payload is mapped
532 */
6d37a0b8 533 ext->data_fin = 1;
6d0060f6
MM
534 ext->use_map = 1;
535 ext->dsn64 = 1;
017512a0 536 ext->data_seq = data_fin_tx_seq;
6d0060f6
MM
537 ext->subflow_seq = 0;
538 ext->data_len = 1;
7279da61 539 } else if (ext->data_seq + ext->data_len == data_fin_tx_seq) {
6d37a0b8
MM
540 /* If there's an existing DSS mapping and it is the
541 * final mapping, DATA_FIN consumes 1 additional byte of
542 * mapping space.
6d0060f6 543 */
6d37a0b8 544 ext->data_fin = 1;
6d0060f6
MM
545 ext->data_len++;
546 }
547}
548
549static bool mptcp_established_options_dss(struct sock *sk, struct sk_buff *skb,
d87903b6 550 bool snd_data_fin_enable,
6d0060f6
MM
551 unsigned int *size,
552 unsigned int remaining,
553 struct mptcp_out_options *opts)
554{
555 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
7279da61 556 struct mptcp_sock *msk = mptcp_sk(subflow->conn);
6d0060f6
MM
557 unsigned int dss_size = 0;
558 struct mptcp_ext *mpext;
6d0060f6 559 unsigned int ack_size;
d22f4988 560 bool ret = false;
d87903b6 561 u64 ack_seq;
6d0060f6 562
c5b39e26 563 opts->csum_reqd = READ_ONCE(msk->csum_enabled);
0bac966a 564 mpext = skb ? mptcp_get_ext(skb) : NULL;
6d0060f6 565
7279da61 566 if (!skb || (mpext && mpext->use_map) || snd_data_fin_enable) {
c5b39e26 567 unsigned int map_size = TCPOLEN_MPTCP_DSS_BASE + TCPOLEN_MPTCP_DSS_MAP64;
6d0060f6 568
c5b39e26
GT
569 if (mpext) {
570 if (opts->csum_reqd)
571 map_size += TCPOLEN_MPTCP_DSS_CHECKSUM;
6d0060f6 572
6d0060f6 573 opts->ext_copy = *mpext;
c5b39e26 574 }
6d0060f6 575
c5b39e26
GT
576 remaining -= map_size;
577 dss_size = map_size;
7279da61 578 if (skb && snd_data_fin_enable)
9c29e361 579 mptcp_write_data_fin(subflow, skb, &opts->ext_copy);
1bff1e43 580 opts->suboptions = OPTION_MPTCP_DSS;
d22f4988
CP
581 ret = true;
582 }
583
2398e399
PA
584 /* passive sockets msk will set the 'can_ack' after accept(), even
585 * if the first subflow may have the already the remote key handy
586 */
d22f4988 587 opts->ext_copy.use_ack = 0;
dc093db5 588 if (!READ_ONCE(msk->can_ack)) {
d22f4988
CP
589 *size = ALIGN(dss_size, 4);
590 return ret;
6d0060f6
MM
591 }
592
e3859603 593 ack_seq = READ_ONCE(msk->ack_seq);
37198e93 594 if (READ_ONCE(msk->use_64bit_ack)) {
a0c1d0ea 595 ack_size = TCPOLEN_MPTCP_DSS_ACK64;
e3859603 596 opts->ext_copy.data_ack = ack_seq;
a0c1d0ea
CP
597 opts->ext_copy.ack64 = 1;
598 } else {
599 ack_size = TCPOLEN_MPTCP_DSS_ACK32;
e3859603 600 opts->ext_copy.data_ack32 = (uint32_t)ack_seq;
a0c1d0ea
CP
601 opts->ext_copy.ack64 = 0;
602 }
603 opts->ext_copy.use_ack = 1;
1bff1e43 604 opts->suboptions = OPTION_MPTCP_DSS;
ea4ca586 605 WRITE_ONCE(msk->old_wspace, __mptcp_space((struct sock *)msk));
6d0060f6
MM
606
607 /* Add kind/length/subtype/flag overhead if mapping is not populated */
608 if (dss_size == 0)
609 ack_size += TCPOLEN_MPTCP_DSS_BASE;
610
611 dss_size += ack_size;
612
6d0060f6
MM
613 *size = ALIGN(dss_size, 4);
614 return true;
615}
616
761c124e
GT
617static u64 add_addr_generate_hmac(u64 key1, u64 key2,
618 struct mptcp_addr_info *addr)
3df523ab 619{
761c124e 620 u16 port = ntohs(addr->port);
bd697222 621 u8 hmac[SHA256_DIGEST_SIZE];
3df523ab 622 u8 msg[19];
761c124e 623 int i = 0;
3df523ab 624
761c124e
GT
625 msg[i++] = addr->id;
626 if (addr->family == AF_INET) {
627 memcpy(&msg[i], &addr->addr.s_addr, 4);
628 i += 4;
629 }
630#if IS_ENABLED(CONFIG_MPTCP_IPV6)
631 else if (addr->family == AF_INET6) {
632 memcpy(&msg[i], &addr->addr6.s6_addr, 16);
633 i += 16;
634 }
635#endif
636 msg[i++] = port >> 8;
637 msg[i++] = port & 0xFF;
3df523ab 638
761c124e 639 mptcp_crypto_hmac_sha(key1, key2, msg, i, hmac);
3df523ab 640
bd697222 641 return get_unaligned_be64(&hmac[SHA256_DIGEST_SIZE - sizeof(u64)]);
3df523ab 642}
3df523ab 643
84dfe367 644static bool mptcp_established_options_add_addr(struct sock *sk, struct sk_buff *skb,
f643b803
GT
645 unsigned int *size,
646 unsigned int remaining,
647 struct mptcp_out_options *opts)
3df523ab
PK
648{
649 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
650 struct mptcp_sock *msk = mptcp_sk(subflow->conn);
84dfe367
GT
651 bool drop_other_suboptions = false;
652 unsigned int opt_size = *size;
6a6c05a8 653 bool echo;
4a2777a8 654 bool port;
1b1c7a0e 655 int len;
3df523ab 656
1f5e9e2f
YL
657 /* add addr will strip the existing options, be sure to avoid breaking
658 * MPC/MPJ handshakes
659 */
f643b803 660 if (!mptcp_pm_should_add_signal(msk) ||
1f5e9e2f
YL
661 (opts->suboptions & (OPTION_MPTCP_MPJ_ACK | OPTION_MPTCP_MPC_ACK)) ||
662 !mptcp_pm_add_addr_signal(msk, skb, opt_size, remaining, &opts->addr,
663 &echo, &port, &drop_other_suboptions))
1b1c7a0e
PK
664 return false;
665
1f5e9e2f
YL
666 if (drop_other_suboptions)
667 remaining += opt_size;
30f60bae 668 len = mptcp_add_addr_len(opts->addr.family, echo, port);
1b1c7a0e
PK
669 if (remaining < len)
670 return false;
3df523ab 671
1b1c7a0e 672 *size = len;
1f5e9e2f
YL
673 if (drop_other_suboptions) {
674 pr_debug("drop other suboptions");
675 opts->suboptions = 0;
1bff1e43
PA
676
677 /* note that e.g. DSS could have written into the memory
678 * aliased by ahmac, we must reset the field here
679 * to avoid appending the hmac even for ADD_ADDR echo
680 * options
681 */
682 opts->ahmac = 0;
84dfe367 683 *size -= opt_size;
1f5e9e2f 684 }
fef6b7ec 685 opts->suboptions |= OPTION_MPTCP_ADD_ADDR;
761c124e
GT
686 if (!echo) {
687 opts->ahmac = add_addr_generate_hmac(msk->local_key,
688 msk->remote_key,
689 &opts->addr);
3df523ab 690 }
4a2777a8 691 pr_debug("addr_id=%d, ahmac=%llu, echo=%d, port=%d",
30f60bae 692 opts->addr.id, opts->ahmac, echo, ntohs(opts->addr.port));
3df523ab
PK
693
694 return true;
695}
696
5cb104ae
GT
697static bool mptcp_established_options_rm_addr(struct sock *sk,
698 unsigned int *size,
699 unsigned int remaining,
700 struct mptcp_out_options *opts)
701{
702 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
703 struct mptcp_sock *msk = mptcp_sk(subflow->conn);
6445e17a
GT
704 struct mptcp_rm_list rm_list;
705 int i, len;
5cb104ae
GT
706
707 if (!mptcp_pm_should_rm_signal(msk) ||
6445e17a 708 !(mptcp_pm_rm_addr_signal(msk, remaining, &rm_list)))
5cb104ae
GT
709 return false;
710
6445e17a
GT
711 len = mptcp_rm_addr_len(&rm_list);
712 if (len < 0)
713 return false;
714 if (remaining < len)
5cb104ae
GT
715 return false;
716
6445e17a 717 *size = len;
5cb104ae 718 opts->suboptions |= OPTION_MPTCP_RM_ADDR;
6445e17a 719 opts->rm_list = rm_list;
5cb104ae 720
6445e17a
GT
721 for (i = 0; i < opts->rm_list.nr; i++)
722 pr_debug("rm_list_ids[%d]=%d", i, opts->rm_list.ids[i]);
5cb104ae
GT
723
724 return true;
725}
726
06706542
GT
727static bool mptcp_established_options_mp_prio(struct sock *sk,
728 unsigned int *size,
729 unsigned int remaining,
730 struct mptcp_out_options *opts)
731{
732 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
733
1bff1e43
PA
734 /* can't send MP_PRIO with MPC, as they share the same option space:
735 * 'backup'. Also it makes no sense at all
736 */
13ac17a3 737 if (!subflow->send_mp_prio || (opts->suboptions & OPTIONS_MPTCP_MPC))
06706542
GT
738 return false;
739
ec99a470
DC
740 /* account for the trailing 'nop' option */
741 if (remaining < TCPOLEN_MPTCP_PRIO_ALIGN)
06706542
GT
742 return false;
743
ec99a470 744 *size = TCPOLEN_MPTCP_PRIO_ALIGN;
06706542
GT
745 opts->suboptions |= OPTION_MPTCP_PRIO;
746 opts->backup = subflow->request_bkup;
747
748 pr_debug("prio=%d", opts->backup);
749
750 return true;
751}
752
c25aeb4e 753static noinline bool mptcp_established_options_rst(struct sock *sk, struct sk_buff *skb,
dc87efdb
FW
754 unsigned int *size,
755 unsigned int remaining,
756 struct mptcp_out_options *opts)
757{
758 const struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
759
760 if (remaining < TCPOLEN_MPTCP_RST)
c25aeb4e 761 return false;
dc87efdb
FW
762
763 *size = TCPOLEN_MPTCP_RST;
764 opts->suboptions |= OPTION_MPTCP_RST;
765 opts->reset_transient = subflow->reset_transient;
766 opts->reset_reason = subflow->reset_reason;
c25aeb4e
GT
767
768 return true;
769}
770
f284c0c7
PA
771static bool mptcp_established_options_fastclose(struct sock *sk,
772 unsigned int *size,
773 unsigned int remaining,
774 struct mptcp_out_options *opts)
775{
776 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
777 struct mptcp_sock *msk = mptcp_sk(subflow->conn);
778
779 if (likely(!subflow->send_fastclose))
780 return false;
781
782 if (remaining < TCPOLEN_MPTCP_FASTCLOSE)
783 return false;
784
785 *size = TCPOLEN_MPTCP_FASTCLOSE;
786 opts->suboptions |= OPTION_MPTCP_FASTCLOSE;
787 opts->rcvr_key = msk->remote_key;
788
789 pr_debug("FASTCLOSE key=%llu", opts->rcvr_key);
790 return true;
791}
792
c25aeb4e
GT
793static bool mptcp_established_options_mp_fail(struct sock *sk,
794 unsigned int *size,
795 unsigned int remaining,
796 struct mptcp_out_options *opts)
797{
798 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
799
800 if (likely(!subflow->send_mp_fail))
801 return false;
802
803 if (remaining < TCPOLEN_MPTCP_FAIL)
804 return false;
805
806 *size = TCPOLEN_MPTCP_FAIL;
807 opts->suboptions |= OPTION_MPTCP_FAIL;
808 opts->fail_seq = subflow->map_seq;
809
810 pr_debug("MP_FAIL fail_seq=%llu", opts->fail_seq);
811
812 return true;
dc87efdb
FW
813}
814
6d0060f6
MM
815bool mptcp_established_options(struct sock *sk, struct sk_buff *skb,
816 unsigned int *size, unsigned int remaining,
817 struct mptcp_out_options *opts)
818{
d87903b6
PA
819 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
820 struct mptcp_sock *msk = mptcp_sk(subflow->conn);
6d0060f6 821 unsigned int opt_size = 0;
d87903b6 822 bool snd_data_fin;
6d0060f6
MM
823 bool ret = false;
824
3df523ab
PK
825 opts->suboptions = 0;
826
d87903b6 827 if (unlikely(__mptcp_check_fallback(msk)))
e1ff9e82
DC
828 return false;
829
dc87efdb 830 if (unlikely(skb && TCP_SKB_CB(skb)->tcp_flags & TCPHDR_RST)) {
f284c0c7
PA
831 if (mptcp_established_options_fastclose(sk, &opt_size, remaining, opts) ||
832 mptcp_established_options_mp_fail(sk, &opt_size, remaining, opts)) {
c25aeb4e
GT
833 *size += opt_size;
834 remaining -= opt_size;
835 }
f284c0c7 836 /* MP_RST can be used with MP_FASTCLOSE and MP_FAIL if there is room */
c25aeb4e
GT
837 if (mptcp_established_options_rst(sk, skb, &opt_size, remaining, opts)) {
838 *size += opt_size;
839 remaining -= opt_size;
840 }
dc87efdb
FW
841 return true;
842 }
d5824847 843
d87903b6
PA
844 snd_data_fin = mptcp_data_fin_enabled(msk);
845 if (mptcp_established_options_mp(sk, skb, snd_data_fin, &opt_size, remaining, opts))
6d0060f6 846 ret = true;
c25aeb4e 847 else if (mptcp_established_options_dss(sk, skb, snd_data_fin, &opt_size, remaining, opts)) {
04fac2ca
MB
848 unsigned int mp_fail_size;
849
6d0060f6 850 ret = true;
04fac2ca
MB
851 if (mptcp_established_options_mp_fail(sk, &mp_fail_size,
852 remaining - opt_size, opts)) {
853 *size += opt_size + mp_fail_size;
854 remaining -= opt_size - mp_fail_size;
c25aeb4e
GT
855 return true;
856 }
857 }
6d0060f6
MM
858
859 /* we reserved enough space for the above options, and exceeding the
860 * TCP option space would be fatal
861 */
862 if (WARN_ON_ONCE(opt_size > remaining))
863 return false;
864
865 *size += opt_size;
866 remaining -= opt_size;
84dfe367 867 if (mptcp_established_options_add_addr(sk, skb, &opt_size, remaining, opts)) {
3df523ab
PK
868 *size += opt_size;
869 remaining -= opt_size;
870 ret = true;
5cb104ae
GT
871 } else if (mptcp_established_options_rm_addr(sk, &opt_size, remaining, opts)) {
872 *size += opt_size;
873 remaining -= opt_size;
874 ret = true;
3df523ab 875 }
6d0060f6 876
06706542
GT
877 if (mptcp_established_options_mp_prio(sk, &opt_size, remaining, opts)) {
878 *size += opt_size;
879 remaining -= opt_size;
880 ret = true;
881 }
882
6d0060f6
MM
883 return ret;
884}
885
cec37a6e
PK
886bool mptcp_synack_options(const struct request_sock *req, unsigned int *size,
887 struct mptcp_out_options *opts)
888{
889 struct mptcp_subflow_request_sock *subflow_req = mptcp_subflow_rsk(req);
890
891 if (subflow_req->mp_capable) {
892 opts->suboptions = OPTION_MPTCP_MPC_SYNACK;
893 opts->sndr_key = subflow_req->local_key;
06fe1719 894 opts->csum_reqd = subflow_req->csum_reqd;
bab6b88e 895 opts->allow_join_id0 = subflow_req->allow_join_id0;
cec37a6e
PK
896 *size = TCPOLEN_MPTCP_MPC_SYNACK;
897 pr_debug("subflow_req=%p, local_key=%llu",
898 subflow_req, subflow_req->local_key);
899 return true;
f296234c
PK
900 } else if (subflow_req->mp_join) {
901 opts->suboptions = OPTION_MPTCP_MPJ_SYNACK;
902 opts->backup = subflow_req->backup;
903 opts->join_id = subflow_req->local_id;
904 opts->thmac = subflow_req->thmac;
905 opts->nonce = subflow_req->local_nonce;
906 pr_debug("req=%p, bkup=%u, id=%u, thmac=%llu, nonce=%u",
907 subflow_req, opts->backup, opts->join_id,
908 opts->thmac, opts->nonce);
909 *size = TCPOLEN_MPTCP_MPJ_SYNACK;
910 return true;
cec37a6e
PK
911 }
912 return false;
913}
914
d5824847 915static bool check_fully_established(struct mptcp_sock *msk, struct sock *ssk,
f296234c 916 struct mptcp_subflow_context *subflow,
0be534f5
PA
917 struct sk_buff *skb,
918 struct mptcp_options_received *mp_opt)
d22f4988
CP
919{
920 /* here we can process OoO, in-window pkts, only in-sequence 4th ack
f296234c 921 * will make the subflow fully established
d22f4988 922 */
f296234c
PK
923 if (likely(subflow->fully_established)) {
924 /* on passive sockets, check for 3rd ack retransmission
925 * note that msk is always set by subflow_syn_recv_sock()
926 * for mp_join subflows
927 */
928 if (TCP_SKB_CB(skb)->seq == subflow->ssn_offset + 1 &&
929 TCP_SKB_CB(skb)->end_seq == TCP_SKB_CB(skb)->seq &&
74c7dfbe 930 subflow->mp_join && (mp_opt->suboptions & OPTIONS_MPTCP_MPJ) &&
f296234c 931 READ_ONCE(msk->pm.server_side))
d5824847 932 tcp_send_ack(ssk);
f296234c
PK
933 goto fully_established;
934 }
935
d5824847
PA
936 /* we must process OoO packets before the first subflow is fully
937 * established. OoO packets are instead a protocol violation
938 * for MP_JOIN subflows as the peer must not send any data
939 * before receiving the forth ack - cfr. RFC 8684 section 3.2.
f296234c 940 */
d5824847
PA
941 if (TCP_SKB_CB(skb)->seq != subflow->ssn_offset + 1) {
942 if (subflow->mp_join)
943 goto reset;
f296234c 944 return subflow->mp_capable;
d5824847 945 }
d22f4988 946
74c7dfbe
PA
947 if (((mp_opt->suboptions & OPTION_MPTCP_DSS) && mp_opt->use_ack) ||
948 ((mp_opt->suboptions & OPTION_MPTCP_ADD_ADDR) && !mp_opt->echo)) {
f296234c 949 /* subflows are fully established as soon as we get any
67b12f79 950 * additional ack, including ADD_ADDR.
f296234c 951 */
0be534f5 952 subflow->fully_established = 1;
b93df08c 953 WRITE_ONCE(msk->fully_established, true);
f296234c
PK
954 goto fully_established;
955 }
d22f4988 956
d22f4988 957 /* If the first established packet does not contain MP_CAPABLE + data
d5824847
PA
958 * then fallback to TCP. Fallback scenarios requires a reset for
959 * MP_JOIN subflows.
d22f4988 960 */
74c7dfbe 961 if (!(mp_opt->suboptions & OPTIONS_MPTCP_MPC)) {
d5824847
PA
962 if (subflow->mp_join)
963 goto reset;
d22f4988 964 subflow->mp_capable = 0;
e1ff9e82
DC
965 pr_fallback(msk);
966 __mptcp_do_fallback(msk);
d22f4988
CP
967 return false;
968 }
f296234c 969
df377be3
GT
970 if (mp_opt->deny_join_id0)
971 WRITE_ONCE(msk->pm.remote_deny_join_id0, true);
972
d6085fe1
PA
973 if (unlikely(!READ_ONCE(msk->pm.server_side)))
974 pr_warn_once("bogus mpc option on established client sk");
b93df08c 975 mptcp_subflow_fully_established(subflow, mp_opt);
f296234c
PK
976
977fully_established:
5b950ff4
PA
978 /* if the subflow is not already linked into the conn_list, we can't
979 * notify the PM: this subflow is still on the listener queue
980 * and the PM possibly acquiring the subflow lock could race with
981 * the listener close
982 */
983 if (likely(subflow->pm_notified) || list_empty(&subflow->node))
f296234c
PK
984 return true;
985
986 subflow->pm_notified = 1;
ec3edaa7 987 if (subflow->mp_join) {
d5824847 988 clear_3rdack_retransmission(ssk);
62535200 989 mptcp_pm_subflow_established(msk);
ec3edaa7 990 } else {
6c714f1b 991 mptcp_pm_fully_established(msk, ssk, GFP_ATOMIC);
ec3edaa7 992 }
d22f4988 993 return true;
d5824847
PA
994
995reset:
996 mptcp_subflow_reset(ssk);
997 return false;
d22f4988
CP
998}
999
1502328f 1000u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq)
cc9d2566 1001{
1502328f
PA
1002 u32 old_seq32, cur_seq32;
1003
1004 old_seq32 = (u32)old_seq;
1005 cur_seq32 = (u32)cur_seq;
1006 cur_seq = (old_seq & GENMASK_ULL(63, 32)) + cur_seq32;
1007 if (unlikely(cur_seq32 < old_seq32 && before(old_seq32, cur_seq32)))
1008 return cur_seq + (1LL << 32);
1009
1010 /* reverse wrap could happen, too */
1011 if (unlikely(cur_seq32 > old_seq32 && after(old_seq32, cur_seq32)))
1012 return cur_seq - (1LL << 32);
1013 return cur_seq;
cc9d2566
PA
1014}
1015
6f8a612a 1016static void ack_update_msk(struct mptcp_sock *msk,
6e628cd3 1017 struct sock *ssk,
6f8a612a 1018 struct mptcp_options_received *mp_opt)
cc9d2566 1019{
7439d687 1020 u64 new_wnd_end, new_snd_una, snd_nxt = READ_ONCE(msk->snd_nxt);
6f8a612a 1021 struct sock *sk = (struct sock *)msk;
7439d687
PA
1022 u64 old_snd_una;
1023
1024 mptcp_data_lock(sk);
cc9d2566
PA
1025
1026 /* avoid ack expansion on update conflict, to reduce the risk of
1027 * wrongly expanding to a future ack sequence number, which is way
1028 * more dangerous than missing an ack
1029 */
7439d687 1030 old_snd_una = msk->snd_una;
1502328f 1031 new_snd_una = mptcp_expand_seq(old_snd_una, mp_opt->data_ack, mp_opt->ack64);
cc9d2566 1032
0d199e43
FW
1033 /* ACK for data not even sent yet? Ignore.*/
1034 if (unlikely(after64(new_snd_una, snd_nxt)))
1035 new_snd_una = old_snd_una;
cc9d2566 1036
6f8a612a
FW
1037 new_wnd_end = new_snd_una + tcp_sk(ssk)->snd_wnd;
1038
219d0499 1039 if (after64(new_wnd_end, msk->wnd_end))
7439d687 1040 msk->wnd_end = new_wnd_end;
219d0499
PA
1041
1042 /* this assumes mptcp_incoming_options() is invoked after tcp_ack() */
d09d818e 1043 if (after64(msk->wnd_end, READ_ONCE(msk->snd_nxt)))
219d0499 1044 __mptcp_check_push(sk, ssk);
6f8a612a 1045
7439d687
PA
1046 if (after64(new_snd_una, old_snd_una)) {
1047 msk->snd_una = new_snd_una;
1048 __mptcp_data_acked(sk);
cc9d2566 1049 }
7439d687 1050 mptcp_data_unlock(sk);
ed66bfb4
GT
1051
1052 trace_ack_update_msk(mp_opt->data_ack,
1053 old_snd_una, new_snd_una,
1054 new_wnd_end, msk->wnd_end);
cc9d2566
PA
1055}
1056
1a49b2c2 1057bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit)
3721b9b6
MM
1058{
1059 /* Skip if DATA_FIN was already received.
1060 * If updating simultaneously with the recvmsg loop, values
1061 * should match. If they mismatch, the peer is misbehaving and
1062 * we will prefer the most recent information.
1063 */
781bf13d 1064 if (READ_ONCE(msk->rcv_data_fin))
3721b9b6
MM
1065 return false;
1066
1a49b2c2 1067 WRITE_ONCE(msk->rcv_data_fin_seq,
1502328f 1068 mptcp_expand_seq(READ_ONCE(msk->ack_seq), data_fin_seq, use_64bit));
3721b9b6
MM
1069 WRITE_ONCE(msk->rcv_data_fin, 1);
1070
1071 return true;
1072}
1073
1b1c7a0e
PK
1074static bool add_addr_hmac_valid(struct mptcp_sock *msk,
1075 struct mptcp_options_received *mp_opt)
1076{
1077 u64 hmac = 0;
1078
1079 if (mp_opt->echo)
1080 return true;
1081
761c124e
GT
1082 hmac = add_addr_generate_hmac(msk->remote_key,
1083 msk->local_key,
1084 &mp_opt->addr);
1b1c7a0e
PK
1085
1086 pr_debug("msk=%p, ahmac=%llu, mp_opt->ahmac=%llu\n",
1087 msk, (unsigned long long)hmac,
1088 (unsigned long long)mp_opt->ahmac);
1089
1090 return hmac == mp_opt->ahmac;
1091}
1092
6787b7e3
JW
1093/* Return false if a subflow has been reset, else return true */
1094bool mptcp_incoming_options(struct sock *sk, struct sk_buff *skb)
648ef4b8 1095{
d22f4988 1096 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1b1c7a0e 1097 struct mptcp_sock *msk = mptcp_sk(subflow->conn);
cfde141e 1098 struct mptcp_options_received mp_opt;
648ef4b8
MM
1099 struct mptcp_ext *mpext;
1100
6e628cd3
PA
1101 if (__mptcp_check_fallback(msk)) {
1102 /* Keep it simple and unconditionally trigger send data cleanup and
1103 * pending queue spooling. We will need to acquire the data lock
1104 * for more accurate checks, and once the lock is acquired, such
1105 * helpers are cheap.
1106 */
1107 mptcp_data_lock(subflow->conn);
219d0499
PA
1108 if (sk_stream_memory_free(sk))
1109 __mptcp_check_push(subflow->conn, sk);
6e628cd3
PA
1110 __mptcp_data_acked(subflow->conn);
1111 mptcp_data_unlock(subflow->conn);
6787b7e3 1112 return true;
6e628cd3 1113 }
e1ff9e82 1114
c863225b 1115 mptcp_get_options(sk, skb, &mp_opt);
6787b7e3
JW
1116
1117 /* The subflow can be in close state only if check_fully_established()
1118 * just sent a reset. If so, tell the caller to ignore the current packet.
1119 */
cfde141e 1120 if (!check_fully_established(msk, sk, subflow, skb, &mp_opt))
6787b7e3 1121 return sk->sk_state != TCP_CLOSE;
648ef4b8 1122
f6c2ef59
PA
1123 if (unlikely(mp_opt.suboptions != OPTION_MPTCP_DSS)) {
1124 if ((mp_opt.suboptions & OPTION_MPTCP_FASTCLOSE) &&
1125 msk->local_key == mp_opt.rcvr_key) {
1126 WRITE_ONCE(msk->rcv_fastclose, true);
1127 mptcp_schedule_work((struct sock *)msk);
1128 }
50c504a2 1129
f6c2ef59
PA
1130 if ((mp_opt.suboptions & OPTION_MPTCP_ADD_ADDR) &&
1131 add_addr_hmac_valid(msk, &mp_opt)) {
1132 if (!mp_opt.echo) {
1133 mptcp_pm_add_addr_received(msk, &mp_opt.addr);
1134 MPTCP_INC_STATS(sock_net(sk), MPTCP_MIB_ADDADDR);
1135 } else {
1136 mptcp_pm_add_addr_echoed(msk, &mp_opt.addr);
1137 mptcp_pm_del_add_timer(msk, &mp_opt.addr, true);
1138 MPTCP_INC_STATS(sock_net(sk), MPTCP_MIB_ECHOADD);
1139 }
1140
1141 if (mp_opt.addr.port)
1142 MPTCP_INC_STATS(sock_net(sk), MPTCP_MIB_PORTADD);
a877de06 1143 }
2fbdd9ea 1144
f6c2ef59
PA
1145 if (mp_opt.suboptions & OPTION_MPTCP_RM_ADDR)
1146 mptcp_pm_rm_addr_received(msk, &mp_opt.rm_list);
1b1c7a0e 1147
f6c2ef59
PA
1148 if (mp_opt.suboptions & OPTION_MPTCP_PRIO) {
1149 mptcp_pm_mp_prio_received(sk, mp_opt.backup);
1150 MPTCP_INC_STATS(sock_net(sk), MPTCP_MIB_MPPRIORX);
1151 }
d0876b22 1152
f6c2ef59
PA
1153 if (mp_opt.suboptions & OPTION_MPTCP_FAIL) {
1154 mptcp_pm_mp_fail_received(sk, mp_opt.fail_seq);
1155 MPTCP_INC_STATS(sock_net(sk), MPTCP_MIB_MPFAILRX);
1156 }
40453a5c 1157
f6c2ef59
PA
1158 if (mp_opt.suboptions & OPTION_MPTCP_RST) {
1159 subflow->reset_seen = 1;
1160 subflow->reset_reason = mp_opt.reset_reason;
1161 subflow->reset_transient = mp_opt.reset_transient;
1162 }
5580d41b 1163
f6c2ef59
PA
1164 if (!(mp_opt.suboptions & OPTION_MPTCP_DSS))
1165 return true;
dc87efdb
FW
1166 }
1167
cc9d2566
PA
1168 /* we can't wait for recvmsg() to update the ack_seq, otherwise
1169 * monodirectional flows will stuck
1170 */
cfde141e 1171 if (mp_opt.use_ack)
6f8a612a 1172 ack_update_msk(msk, sk, &mp_opt);
cc9d2566 1173
06827b34
MM
1174 /* Zero-data-length packets are dropped by the caller and not
1175 * propagated to the MPTCP layer, so the skb extension does not
1176 * need to be allocated or populated. DATA_FIN information, if
1177 * present, needs to be updated here before the skb is freed.
43b54c6e
MM
1178 */
1179 if (TCP_SKB_CB(skb)->seq == TCP_SKB_CB(skb)->end_seq) {
1180 if (mp_opt.data_fin && mp_opt.data_len == 1 &&
1a49b2c2 1181 mptcp_update_rcv_data_fin(msk, mp_opt.data_seq, mp_opt.dsn64) &&
43b54c6e
MM
1182 schedule_work(&msk->work))
1183 sock_hold(subflow->conn);
06827b34 1184
6787b7e3 1185 return true;
43b54c6e
MM
1186 }
1187
648ef4b8
MM
1188 mpext = skb_ext_add(skb, SKB_EXT_MPTCP);
1189 if (!mpext)
6787b7e3 1190 return true;
648ef4b8
MM
1191
1192 memset(mpext, 0, sizeof(*mpext));
1193
f6c2ef59 1194 if (likely(mp_opt.use_map)) {
cfde141e 1195 if (mp_opt.mpc_map) {
cc7972ea
CP
1196 /* this is an MP_CAPABLE carrying MPTCP data
1197 * we know this map the first chunk of data
1198 */
1199 mptcp_crypto_key_sha(subflow->remote_key, NULL,
1200 &mpext->data_seq);
1201 mpext->data_seq++;
1202 mpext->subflow_seq = 1;
1203 mpext->dsn64 = 1;
1204 mpext->mpc_map = 1;
a77895db 1205 mpext->data_fin = 0;
cc7972ea 1206 } else {
cfde141e
PA
1207 mpext->data_seq = mp_opt.data_seq;
1208 mpext->subflow_seq = mp_opt.subflow_seq;
1209 mpext->dsn64 = mp_opt.dsn64;
1210 mpext->data_fin = mp_opt.data_fin;
cc7972ea 1211 }
cfde141e 1212 mpext->data_len = mp_opt.data_len;
648ef4b8 1213 mpext->use_map = 1;
74c7dfbe 1214 mpext->csum_reqd = !!(mp_opt.suboptions & OPTION_MPTCP_CSUMREQD);
208e8f66
GT
1215
1216 if (mpext->csum_reqd)
1217 mpext->csum = mp_opt.csum;
648ef4b8 1218 }
6787b7e3
JW
1219
1220 return true;
648ef4b8
MM
1221}
1222
fa3fe2b1
FW
1223static void mptcp_set_rwin(const struct tcp_sock *tp)
1224{
1225 const struct sock *ssk = (const struct sock *)tp;
1226 const struct mptcp_subflow_context *subflow;
1227 struct mptcp_sock *msk;
1228 u64 ack_seq;
1229
1230 subflow = mptcp_subflow_ctx(ssk);
1231 msk = mptcp_sk(subflow->conn);
1232
1233 ack_seq = READ_ONCE(msk->ack_seq) + tp->rcv_wnd;
1234
1235 if (after64(ack_seq, READ_ONCE(msk->rcv_wnd_sent)))
1236 WRITE_ONCE(msk->rcv_wnd_sent, ack_seq);
1237}
1238
c312ee21 1239u16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum)
c94b1f96
GT
1240{
1241 struct csum_pseudo_header header;
1242 __wsum csum;
1243
1244 /* cfr RFC 8684 3.3.1.:
1245 * the data sequence number used in the pseudo-header is
1246 * always the 64-bit value, irrespective of what length is used in the
1247 * DSS option itself.
1248 */
f7cc8890
DC
1249 header.data_seq = cpu_to_be64(data_seq);
1250 header.subflow_seq = htonl(subflow_seq);
1251 header.data_len = htons(data_len);
c94b1f96
GT
1252 header.csum = 0;
1253
c312ee21 1254 csum = csum_partial(&header, sizeof(header), sum);
c94b1f96
GT
1255 return (__force u16)csum_fold(csum);
1256}
1257
f7cc8890
DC
1258static u16 mptcp_make_csum(const struct mptcp_ext *mpext)
1259{
1260 return __mptcp_make_csum(mpext->data_seq, mpext->subflow_seq, mpext->data_len,
c312ee21 1261 ~csum_unfold(mpext->csum));
f7cc8890
DC
1262}
1263
fa3fe2b1
FW
1264void mptcp_write_options(__be32 *ptr, const struct tcp_sock *tp,
1265 struct mptcp_out_options *opts)
eda7acdd 1266{
c25aeb4e
GT
1267 if (unlikely(OPTION_MPTCP_FAIL & opts->suboptions)) {
1268 const struct sock *ssk = (const struct sock *)tp;
1269 struct mptcp_subflow_context *subflow;
1270
1271 subflow = mptcp_subflow_ctx(ssk);
1272 subflow->send_mp_fail = 0;
1273
1274 *ptr++ = mptcp_option(MPTCPOPT_MP_FAIL,
1275 TCPOLEN_MPTCP_FAIL,
1276 0, 0);
1277 put_unaligned_be64(opts->fail_seq, ptr);
1278 ptr += 2;
1279 }
1280
f284c0c7
PA
1281 /* DSS, MPC, MPJ, ADD_ADDR, FASTCLOSE and RST are mutually exclusive,
1282 * see mptcp_established_options*()
1bff1e43
PA
1283 */
1284 if (likely(OPTION_MPTCP_DSS & opts->suboptions)) {
1285 struct mptcp_ext *mpext = &opts->ext_copy;
1286 u8 len = TCPOLEN_MPTCP_DSS_BASE;
1287 u8 flags = 0;
1288
1289 if (mpext->use_ack) {
1290 flags = MPTCP_DSS_HAS_ACK;
1291 if (mpext->ack64) {
1292 len += TCPOLEN_MPTCP_DSS_ACK64;
1293 flags |= MPTCP_DSS_ACK64;
1294 } else {
1295 len += TCPOLEN_MPTCP_DSS_ACK32;
1296 }
1297 }
1298
1299 if (mpext->use_map) {
1300 len += TCPOLEN_MPTCP_DSS_MAP64;
1301
1302 /* Use only 64-bit mapping flags for now, add
1303 * support for optional 32-bit mappings later.
1304 */
1305 flags |= MPTCP_DSS_HAS_MAP | MPTCP_DSS_DSN64;
1306 if (mpext->data_fin)
1307 flags |= MPTCP_DSS_DATA_FIN;
1308
1309 if (opts->csum_reqd)
1310 len += TCPOLEN_MPTCP_DSS_CHECKSUM;
1311 }
1312
1313 *ptr++ = mptcp_option(MPTCPOPT_DSS, len, 0, flags);
1314
1315 if (mpext->use_ack) {
1316 if (mpext->ack64) {
1317 put_unaligned_be64(mpext->data_ack, ptr);
1318 ptr += 2;
1319 } else {
1320 put_unaligned_be32(mpext->data_ack32, ptr);
1321 ptr += 1;
1322 }
1323 }
1324
1325 if (mpext->use_map) {
1326 put_unaligned_be64(mpext->data_seq, ptr);
1327 ptr += 2;
1328 put_unaligned_be32(mpext->subflow_seq, ptr);
1329 ptr += 1;
1330 if (opts->csum_reqd) {
1331 put_unaligned_be32(mpext->data_len << 16 |
1332 mptcp_make_csum(mpext), ptr);
1333 } else {
1334 put_unaligned_be32(mpext->data_len << 16 |
1335 TCPOPT_NOP << 8 | TCPOPT_NOP, ptr);
1336 }
110b6d1f 1337 ptr += 1;
1bff1e43 1338 }
13ac17a3 1339 } else if (OPTIONS_MPTCP_MPC & opts->suboptions) {
06fe1719 1340 u8 len, flag = MPTCP_CAP_HMAC_SHA256;
eda7acdd 1341
c94b1f96 1342 if (OPTION_MPTCP_MPC_SYN & opts->suboptions) {
eda7acdd 1343 len = TCPOLEN_MPTCP_MPC_SYN;
c94b1f96 1344 } else if (OPTION_MPTCP_MPC_SYNACK & opts->suboptions) {
cec37a6e 1345 len = TCPOLEN_MPTCP_MPC_SYNACK;
f7cc8890 1346 } else if (opts->data_len) {
cc7972ea 1347 len = TCPOLEN_MPTCP_MPC_ACK_DATA;
c94b1f96
GT
1348 if (opts->csum_reqd)
1349 len += TCPOLEN_MPTCP_DSS_CHECKSUM;
1350 } else {
eda7acdd 1351 len = TCPOLEN_MPTCP_MPC_ACK;
c94b1f96 1352 }
eda7acdd 1353
06fe1719
GT
1354 if (opts->csum_reqd)
1355 flag |= MPTCP_CAP_CHECKSUM_REQD;
1356
bab6b88e
GT
1357 if (!opts->allow_join_id0)
1358 flag |= MPTCP_CAP_DENY_JOIN_ID0;
1359
3df523ab
PK
1360 *ptr++ = mptcp_option(MPTCPOPT_MP_CAPABLE, len,
1361 MPTCP_SUPPORTED_VERSION,
06fe1719 1362 flag);
cc7972ea
CP
1363
1364 if (!((OPTION_MPTCP_MPC_SYNACK | OPTION_MPTCP_MPC_ACK) &
1365 opts->suboptions))
1366 goto mp_capable_done;
1367
eda7acdd
PK
1368 put_unaligned_be64(opts->sndr_key, ptr);
1369 ptr += 2;
cc7972ea
CP
1370 if (!((OPTION_MPTCP_MPC_ACK) & opts->suboptions))
1371 goto mp_capable_done;
1372
1373 put_unaligned_be64(opts->rcvr_key, ptr);
1374 ptr += 2;
f7cc8890 1375 if (!opts->data_len)
cc7972ea
CP
1376 goto mp_capable_done;
1377
c94b1f96 1378 if (opts->csum_reqd) {
f7cc8890
DC
1379 put_unaligned_be32(opts->data_len << 16 |
1380 __mptcp_make_csum(opts->data_seq,
1381 opts->subflow_seq,
1382 opts->data_len,
c312ee21 1383 ~csum_unfold(opts->csum)), ptr);
c94b1f96 1384 } else {
f7cc8890 1385 put_unaligned_be32(opts->data_len << 16 |
c94b1f96
GT
1386 TCPOPT_NOP << 8 | TCPOPT_NOP, ptr);
1387 }
cc7972ea 1388 ptr += 1;
6d0060f6 1389
1bff1e43
PA
1390 /* MPC is additionally mutually exclusive with MP_PRIO */
1391 goto mp_capable_done;
71b077e4
PA
1392 } else if (OPTIONS_MPTCP_MPJ & opts->suboptions) {
1393 if (OPTION_MPTCP_MPJ_SYN & opts->suboptions) {
1394 *ptr++ = mptcp_option(MPTCPOPT_MP_JOIN,
1395 TCPOLEN_MPTCP_MPJ_SYN,
1396 opts->backup, opts->join_id);
1397 put_unaligned_be32(opts->token, ptr);
1398 ptr += 1;
1399 put_unaligned_be32(opts->nonce, ptr);
1400 ptr += 1;
1401 } else if (OPTION_MPTCP_MPJ_SYNACK & opts->suboptions) {
1402 *ptr++ = mptcp_option(MPTCPOPT_MP_JOIN,
1403 TCPOLEN_MPTCP_MPJ_SYNACK,
1404 opts->backup, opts->join_id);
1405 put_unaligned_be64(opts->thmac, ptr);
1406 ptr += 2;
1407 put_unaligned_be32(opts->nonce, ptr);
1408 ptr += 1;
1409 } else {
1410 *ptr++ = mptcp_option(MPTCPOPT_MP_JOIN,
1411 TCPOLEN_MPTCP_MPJ_ACK, 0, 0);
1412 memcpy(ptr, opts->hmac, MPTCPOPT_HMAC_LEN);
1413 ptr += 5;
1414 }
1bff1e43 1415 } else if (OPTION_MPTCP_ADD_ADDR & opts->suboptions) {
6eb3d1e3
GT
1416 u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE;
1417 u8 echo = MPTCP_ADDR_ECHO;
1418
e1ef6832 1419#if IS_ENABLED(CONFIG_MPTCP_IPV6)
fef6b7ec 1420 if (opts->addr.family == AF_INET6)
e1ef6832
GT
1421 len = TCPOLEN_MPTCP_ADD_ADDR6_BASE;
1422#endif
1423
30f60bae 1424 if (opts->addr.port)
22fb85ff
GT
1425 len += TCPOLEN_MPTCP_PORT_LEN;
1426
3df523ab 1427 if (opts->ahmac) {
6eb3d1e3
GT
1428 len += sizeof(opts->ahmac);
1429 echo = 0;
3df523ab 1430 }
3df523ab 1431
6eb3d1e3 1432 *ptr++ = mptcp_option(MPTCPOPT_ADD_ADDR,
30f60bae 1433 len, echo, opts->addr.id);
fef6b7ec 1434 if (opts->addr.family == AF_INET) {
30f60bae 1435 memcpy((u8 *)ptr, (u8 *)&opts->addr.addr.s_addr, 4);
e1ef6832 1436 ptr += 1;
3df523ab 1437 }
3df523ab 1438#if IS_ENABLED(CONFIG_MPTCP_IPV6)
fef6b7ec 1439 else if (opts->addr.family == AF_INET6) {
30f60bae 1440 memcpy((u8 *)ptr, opts->addr.addr6.s6_addr, 16);
e1ef6832 1441 ptr += 4;
3df523ab 1442 }
3df523ab
PK
1443#endif
1444
30f60bae 1445 if (!opts->addr.port) {
22fb85ff
GT
1446 if (opts->ahmac) {
1447 put_unaligned_be64(opts->ahmac, ptr);
1448 ptr += 2;
1449 }
1450 } else {
30f60bae
GT
1451 u16 port = ntohs(opts->addr.port);
1452
22fb85ff
GT
1453 if (opts->ahmac) {
1454 u8 *bptr = (u8 *)ptr;
1455
30f60bae 1456 put_unaligned_be16(port, bptr);
22fb85ff
GT
1457 bptr += 2;
1458 put_unaligned_be64(opts->ahmac, bptr);
1459 bptr += 8;
1460 put_unaligned_be16(TCPOPT_NOP << 8 |
1461 TCPOPT_NOP, bptr);
1462
1463 ptr += 3;
1464 } else {
30f60bae 1465 put_unaligned_be32(port << 16 |
22fb85ff
GT
1466 TCPOPT_NOP << 8 |
1467 TCPOPT_NOP, ptr);
1468 ptr += 1;
1469 }
3df523ab 1470 }
f284c0c7
PA
1471 } else if (unlikely(OPTION_MPTCP_FASTCLOSE & opts->suboptions)) {
1472 /* FASTCLOSE is mutually exclusive with others except RST */
1473 *ptr++ = mptcp_option(MPTCPOPT_MP_FASTCLOSE,
1474 TCPOLEN_MPTCP_FASTCLOSE,
1475 0, 0);
1476 put_unaligned_be64(opts->rcvr_key, ptr);
1477 ptr += 2;
1478
1479 if (OPTION_MPTCP_RST & opts->suboptions)
1480 goto mp_rst;
1481 return;
1482 } else if (unlikely(OPTION_MPTCP_RST & opts->suboptions)) {
1483mp_rst:
1484 *ptr++ = mptcp_option(MPTCPOPT_RST,
1485 TCPOLEN_MPTCP_RST,
1486 opts->reset_transient,
1487 opts->reset_reason);
1488 return;
3df523ab 1489 }
3df523ab 1490
1bff1e43
PA
1491 if (OPTION_MPTCP_PRIO & opts->suboptions) {
1492 const struct sock *ssk = (const struct sock *)tp;
1493 struct mptcp_subflow_context *subflow;
1494
1495 subflow = mptcp_subflow_ctx(ssk);
1496 subflow->send_mp_prio = 0;
1497
1498 *ptr++ = mptcp_option(MPTCPOPT_MP_PRIO,
1499 TCPOLEN_MPTCP_PRIO,
1500 opts->backup, TCPOPT_NOP);
1501 }
1502
1503mp_capable_done:
3df523ab 1504 if (OPTION_MPTCP_RM_ADDR & opts->suboptions) {
6445e17a
GT
1505 u8 i = 1;
1506
3df523ab 1507 *ptr++ = mptcp_option(MPTCPOPT_RM_ADDR,
6445e17a
GT
1508 TCPOLEN_MPTCP_RM_ADDR_BASE + opts->rm_list.nr,
1509 0, opts->rm_list.ids[0]);
1510
1511 while (i < opts->rm_list.nr) {
1512 u8 id1, id2, id3, id4;
1513
1514 id1 = opts->rm_list.ids[i];
1515 id2 = i + 1 < opts->rm_list.nr ? opts->rm_list.ids[i + 1] : TCPOPT_NOP;
1516 id3 = i + 2 < opts->rm_list.nr ? opts->rm_list.ids[i + 2] : TCPOPT_NOP;
1517 id4 = i + 3 < opts->rm_list.nr ? opts->rm_list.ids[i + 3] : TCPOPT_NOP;
1518 put_unaligned_be32(id1 << 24 | id2 << 16 | id3 << 8 | id4, ptr);
1519 ptr += 1;
1520 i += 4;
1521 }
3df523ab
PK
1522 }
1523
fa3fe2b1
FW
1524 if (tp)
1525 mptcp_set_rwin(tp);
eda7acdd 1526}
dc87efdb
FW
1527
1528__be32 mptcp_get_reset_option(const struct sk_buff *skb)
1529{
1530 const struct mptcp_ext *ext = mptcp_get_ext(skb);
1531 u8 flags, reason;
1532
1533 if (ext) {
1534 flags = ext->reset_transient;
1535 reason = ext->reset_reason;
1536
1537 return mptcp_option(MPTCPOPT_RST, TCPOLEN_MPTCP_RST,
1538 flags, reason);
1539 }
1540
1541 return htonl(0u);
1542}
1543EXPORT_SYMBOL_GPL(mptcp_get_reset_option);