xref: /xnu-10002.41.9/bsd/netinet/mptcp_usrreq.c (revision 699cd48037512bf4380799317ca44ca453c82f57)
1 /*
2  * Copyright (c) 2012-2020 Apple Inc. All rights reserved.
3  *
4  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
5  *
6  * This file contains Original Code and/or Modifications of Original Code
7  * as defined in and that are subject to the Apple Public Source License
8  * Version 2.0 (the 'License'). You may not use this file except in
9  * compliance with the License. The rights granted to you under the License
10  * may not be used to create, or enable the creation or redistribution of,
11  * unlawful or unlicensed copies of an Apple operating system, or to
12  * circumvent, violate, or enable the circumvention or violation of, any
13  * terms of an Apple operating system software license agreement.
14  *
15  * Please obtain a copy of the License at
16  * http://www.opensource.apple.com/apsl/ and read it before using this file.
17  *
18  * The Original Code and all software distributed under the License are
19  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
20  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
21  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
22  * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
23  * Please see the License for the specific language governing rights and
24  * limitations under the License.
25  *
26  * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
27  */
28 
29 #include <sys/param.h>
30 #include <sys/systm.h>
31 #include <sys/kernel.h>
32 #include <sys/socket.h>
33 #include <sys/socketvar.h>
34 #include <sys/protosw.h>
35 #include <sys/mcache.h>
36 #include <sys/syslog.h>
37 #include <sys/proc.h>
38 #include <sys/proc_internal.h>
39 #include <sys/resourcevar.h>
40 #include <sys/kauth.h>
41 #include <sys/priv.h>
42 
43 #include <net/if.h>
44 #include <netinet/in.h>
45 #include <netinet/in_var.h>
46 #include <netinet/tcp.h>
47 #include <netinet/tcp_fsm.h>
48 #include <netinet/tcp_seq.h>
49 #include <netinet/tcp_var.h>
50 #include <netinet/tcp_timer.h>
51 #include <netinet/mptcp.h>
52 #include <netinet/mptcp_var.h>
53 #include <netinet/mptcp_timer.h>
54 
55 #include <mach/sdt.h>
56 
57 static int mptcp_usr_attach(struct socket *, int, struct proc *);
58 static int mptcp_usr_detach(struct socket *);
59 static int mptcp_attach(struct socket *, struct proc *);
60 static int mptcp_usr_connectx(struct socket *, struct sockaddr *,
61     struct sockaddr *, struct proc *, uint32_t, sae_associd_t,
62     sae_connid_t *, uint32_t, void *, uint32_t, struct uio *, user_ssize_t *);
63 static int mptcp_getassocids(struct mptses *, uint32_t *, user_addr_t);
64 static int mptcp_getconnids(struct mptses *, sae_associd_t, uint32_t *,
65     user_addr_t);
66 static int mptcp_getconninfo(struct mptses *, sae_connid_t *, uint32_t *,
67     uint32_t *, int32_t *, user_addr_t, socklen_t *, user_addr_t, socklen_t *,
68     uint32_t *, user_addr_t, uint32_t *);
69 static int mptcp_usr_control(struct socket *, u_long, caddr_t, struct ifnet *,
70     struct proc *);
71 static int mptcp_disconnect(struct mptses *);
72 static int mptcp_usr_disconnect(struct socket *);
73 static int mptcp_usr_disconnectx(struct socket *, sae_associd_t, sae_connid_t);
74 static struct mptses *mptcp_usrclosed(struct mptses *);
75 static int mptcp_usr_rcvd(struct socket *, int);
76 static int mptcp_usr_send(struct socket *, int, struct mbuf *,
77     struct sockaddr *, struct mbuf *, struct proc *);
78 static int mptcp_usr_shutdown(struct socket *);
79 static int mptcp_usr_sosend(struct socket *, struct sockaddr *, struct uio *,
80     struct mbuf *, struct mbuf *, int);
81 static int mptcp_usr_socheckopt(struct socket *, struct sockopt *);
82 static int mptcp_usr_preconnect(struct socket *so);
83 
84 struct pr_usrreqs mptcp_usrreqs = {
85 	.pru_attach =           mptcp_usr_attach,
86 	.pru_connectx =         mptcp_usr_connectx,
87 	.pru_control =          mptcp_usr_control,
88 	.pru_detach =           mptcp_usr_detach,
89 	.pru_disconnect =       mptcp_usr_disconnect,
90 	.pru_disconnectx =      mptcp_usr_disconnectx,
91 	.pru_peeraddr =         mp_getpeeraddr,
92 	.pru_rcvd =             mptcp_usr_rcvd,
93 	.pru_send =             mptcp_usr_send,
94 	.pru_shutdown =         mptcp_usr_shutdown,
95 	.pru_sockaddr =         mp_getsockaddr,
96 	.pru_sosend =           mptcp_usr_sosend,
97 	.pru_soreceive =        soreceive,
98 	.pru_socheckopt =       mptcp_usr_socheckopt,
99 	.pru_preconnect =       mptcp_usr_preconnect,
100 };
101 
102 
103 int mptcp_developer_mode = 0;
104 SYSCTL_INT(_net_inet_mptcp, OID_AUTO, allow_aggregate, CTLFLAG_RW | CTLFLAG_LOCKED,
105     &mptcp_developer_mode, 0, "Allow the Multipath aggregation mode");
106 
107 int mptcp_no_first_party = 0;
108 SYSCTL_INT(_net_inet_mptcp, OID_AUTO, no_first_party, CTLFLAG_RW | CTLFLAG_LOCKED,
109     &mptcp_no_first_party, 0, "Do not do first-party app exemptions");
110 
111 static unsigned long mptcp_expected_progress_headstart = 5000;
112 SYSCTL_ULONG(_net_inet_mptcp, OID_AUTO, expected_progress_headstart, CTLFLAG_RW | CTLFLAG_LOCKED,
113     &mptcp_expected_progress_headstart, "Headstart to give MPTCP before meeting the progress deadline");
114 
115 
116 /*
117  * Attaches an MPTCP control block to a socket.
118  */
119 static int
mptcp_usr_attach(struct socket * mp_so,int proto,struct proc * p)120 mptcp_usr_attach(struct socket *mp_so, int proto, struct proc *p)
121 {
122 #pragma unused(proto)
123 	int error;
124 
125 	VERIFY(mpsotomppcb(mp_so) == NULL);
126 
127 	error = mptcp_attach(mp_so, p);
128 	if (error) {
129 		goto out;
130 	}
131 
132 	if ((mp_so->so_options & SO_LINGER) && mp_so->so_linger == 0) {
133 		mp_so->so_linger = (short)(TCP_LINGERTIME * hz);
134 	}
135 out:
136 	return error;
137 }
138 
139 /*
140  * Detaches an MPTCP control block from a socket.
141  */
142 static int
mptcp_usr_detach(struct socket * mp_so)143 mptcp_usr_detach(struct socket *mp_so)
144 {
145 	struct mptses *mpte = mpsotompte(mp_so);
146 	struct mppcb *mpp = mpsotomppcb(mp_so);
147 
148 	if (mpp == NULL || mpp->mpp_state == MPPCB_STATE_DEAD) {
149 		os_log_error(mptcp_log_handle, "%s - %lx: state: %d\n",
150 		    __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte),
151 		    mpp ? mpp->mpp_state : -1);
152 		return EINVAL;
153 	}
154 
155 	/*
156 	 * We are done with this MPTCP socket (it has been closed);
157 	 * trigger all subflows to be disconnected, if not already,
158 	 * by initiating the PCB detach sequence (SOF_PCBCLEARING
159 	 * will be set.)
160 	 */
161 	mp_pcbdetach(mp_so);
162 
163 	mptcp_disconnect(mpte);
164 
165 	return 0;
166 }
167 
168 /*
169  * Attach MPTCP protocol to socket, allocating MP control block,
170  * MPTCP session, control block, buffer space, etc.
171  */
172 static int
mptcp_attach(struct socket * mp_so,struct proc * p)173 mptcp_attach(struct socket *mp_so, struct proc *p)
174 {
175 #pragma unused(p)
176 	struct mptses *mpte = NULL;
177 	struct mptcb *mp_tp = NULL;
178 	struct mppcb *mpp = NULL;
179 	int error = 0;
180 
181 	if (mp_so->so_snd.sb_hiwat == 0 || mp_so->so_rcv.sb_hiwat == 0) {
182 		error = soreserve(mp_so, tcp_sendspace, tcp_recvspace);
183 		if (error != 0) {
184 			goto out;
185 		}
186 	}
187 
188 	if (mp_so->so_snd.sb_preconn_hiwat == 0) {
189 		soreserve_preconnect(mp_so, 2048);
190 	}
191 
192 	if ((mp_so->so_rcv.sb_flags & SB_USRSIZE) == 0) {
193 		mp_so->so_rcv.sb_flags |= SB_AUTOSIZE;
194 	}
195 	if ((mp_so->so_snd.sb_flags & SB_USRSIZE) == 0) {
196 		mp_so->so_snd.sb_flags |= SB_AUTOSIZE;
197 	}
198 
199 	/*
200 	 * MPTCP send-socket buffers cannot be compressed, due to the
201 	 * fact that each mbuf chained via m_next is a M_PKTHDR
202 	 * which carries some MPTCP metadata.
203 	 */
204 	mp_so->so_snd.sb_flags |= SB_NOCOMPRESS;
205 
206 	if ((error = mp_pcballoc(mp_so, &mtcbinfo)) != 0) {
207 		goto out;
208 	}
209 
210 	mpp = mpsotomppcb(mp_so);
211 	mpte = (struct mptses *)mpp->mpp_pcbe;
212 	mp_tp = mpte->mpte_mptcb;
213 
214 	VERIFY(mp_tp != NULL);
215 out:
216 	return error;
217 }
218 
219 static int
mptcp_entitlement_check(struct socket * mp_so,uint8_t svctype)220 mptcp_entitlement_check(struct socket *mp_so, uint8_t svctype)
221 {
222 	struct mptses *mpte = mpsotompte(mp_so);
223 
224 	if (mptcp_no_first_party) {
225 		return 0;
226 	}
227 
228 	/* First, check for mptcp_extended without delegation */
229 	if (soopt_cred_check(mp_so, PRIV_NET_RESTRICTED_MULTIPATH_EXTENDED, TRUE, FALSE) == 0) {
230 		/*
231 		 * This means the app has the extended entitlement. Thus,
232 		 * it's a first party app and can run without restrictions.
233 		 */
234 		mpte->mpte_flags |= MPTE_FIRSTPARTY;
235 		return 0;
236 	}
237 
238 	/* Now with delegation */
239 	if (mp_so->so_flags & SOF_DELEGATED &&
240 	    soopt_cred_check(mp_so, PRIV_NET_RESTRICTED_MULTIPATH_EXTENDED, TRUE, TRUE) == 0) {
241 		/*
242 		 * This means the app has the extended entitlement. Thus,
243 		 * it's a first party app and can run without restrictions.
244 		 */
245 		mpte->mpte_flags |= MPTE_FIRSTPARTY;
246 		return 0;
247 	}
248 
249 	if (svctype == MPTCP_SVCTYPE_AGGREGATE) {
250 		if (mptcp_developer_mode) {
251 			return 0;
252 		}
253 
254 		os_log_error(mptcp_log_handle, "%s - %lx: MPTCP prohibited on svc %u\n",
255 		    __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte), mpte->mpte_svctype);
256 		return -1;
257 	}
258 
259 	return 0;
260 }
261 
262 /*
263  * Common subroutine to open a MPTCP connection to one of the remote hosts
264  * specified by dst_sl.  This includes allocating and establishing a
265  * subflow TCP connection, either initially to establish MPTCP connection,
266  * or to join an existing one.  Returns a connection handle upon success.
267  */
268 static int
mptcp_connectx(struct mptses * mpte,struct sockaddr * src,struct sockaddr * dst,uint32_t ifscope,sae_connid_t * pcid)269 mptcp_connectx(struct mptses *mpte, struct sockaddr *src,
270     struct sockaddr *dst, uint32_t ifscope, sae_connid_t *pcid)
271 {
272 	int error = 0;
273 
274 	VERIFY(dst != NULL);
275 	VERIFY(pcid != NULL);
276 
277 	error = mptcp_subflow_add(mpte, src, dst, ifscope, pcid);
278 
279 	return error;
280 }
281 
282 /*
283  * User-protocol pru_connectx callback.
284  */
285 static int
mptcp_usr_connectx(struct socket * mp_so,struct sockaddr * src,struct sockaddr * dst,struct proc * p,uint32_t ifscope,sae_associd_t aid,sae_connid_t * pcid,uint32_t flags,void * arg,uint32_t arglen,struct uio * auio,user_ssize_t * bytes_written)286 mptcp_usr_connectx(struct socket *mp_so, struct sockaddr *src,
287     struct sockaddr *dst, struct proc *p, uint32_t ifscope,
288     sae_associd_t aid, sae_connid_t *pcid, uint32_t flags, void *arg,
289     uint32_t arglen, struct uio *auio, user_ssize_t *bytes_written)
290 {
291 #pragma unused(p, aid, flags, arg, arglen)
292 	struct mppcb *mpp = mpsotomppcb(mp_so);
293 	struct mptses *mpte = NULL;
294 	struct mptcb *mp_tp = NULL;
295 	user_ssize_t    datalen;
296 	int error = 0;
297 
298 	if (mpp == NULL || mpp->mpp_state == MPPCB_STATE_DEAD) {
299 		os_log_error(mptcp_log_handle, "%s - %lx: state %d\n",
300 		    __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte),
301 		    mpp ? mpp->mpp_state : -1);
302 		error = EINVAL;
303 		goto out;
304 	}
305 	mpte = mptompte(mpp);
306 	mp_tp = mpte->mpte_mptcb;
307 
308 	if (mp_tp->mpt_flags &  MPTCPF_FALLBACK_TO_TCP) {
309 		os_log_error(mptcp_log_handle, "%s - %lx: fell back to TCP\n",
310 		    __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte));
311 		error = EINVAL;
312 		goto out;
313 	}
314 
315 	if (dst->sa_family != AF_INET && dst->sa_family != AF_INET6) {
316 		error = EAFNOSUPPORT;
317 		goto out;
318 	}
319 
320 	if (dst->sa_family == AF_INET &&
321 	    dst->sa_len != sizeof(mpte->__mpte_dst_v4)) {
322 		os_log_error(mptcp_log_handle, "%s - %lx: IPv4 dst len %u\n",
323 		    __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte), dst->sa_len);
324 		error = EINVAL;
325 		goto out;
326 	}
327 
328 	if (dst->sa_family == AF_INET6 &&
329 	    dst->sa_len != sizeof(mpte->__mpte_dst_v6)) {
330 		os_log_error(mptcp_log_handle, "%s - %lx: IPv6 dst len %u\n",
331 		    __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte), dst->sa_len);
332 		error = EINVAL;
333 		goto out;
334 	}
335 
336 	if (!(mpte->mpte_flags & MPTE_SVCTYPE_CHECKED)) {
337 		if (mptcp_entitlement_check(mp_so, mpte->mpte_svctype) < 0) {
338 			error = EPERM;
339 			goto out;
340 		}
341 
342 		mpte->mpte_flags |= MPTE_SVCTYPE_CHECKED;
343 	}
344 
345 	if ((mp_so->so_state & (SS_ISCONNECTED | SS_ISCONNECTING)) == 0) {
346 		memcpy(&mpte->mpte_u_dst, dst, dst->sa_len);
347 
348 		if (dst->sa_family == AF_INET) {
349 			memcpy(&mpte->mpte_sub_dst_v4, dst, dst->sa_len);
350 		} else {
351 			memcpy(&mpte->mpte_sub_dst_v6, dst, dst->sa_len);
352 		}
353 	}
354 
355 	if (src) {
356 		if (src->sa_family != AF_INET && src->sa_family != AF_INET6) {
357 			error = EAFNOSUPPORT;
358 			goto out;
359 		}
360 
361 		if (src->sa_family == AF_INET &&
362 		    src->sa_len != sizeof(mpte->__mpte_src_v4)) {
363 			os_log_error(mptcp_log_handle, "%s - %lx: IPv4 src len %u\n",
364 			    __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte), src->sa_len);
365 			error = EINVAL;
366 			goto out;
367 		}
368 
369 		if (src->sa_family == AF_INET6 &&
370 		    src->sa_len != sizeof(mpte->__mpte_src_v6)) {
371 			os_log_error(mptcp_log_handle, "%s - %lx: IPv6 src len %u\n",
372 			    __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte), src->sa_len);
373 			error = EINVAL;
374 			goto out;
375 		}
376 
377 		if ((mp_so->so_state & (SS_ISCONNECTED | SS_ISCONNECTING)) == 0) {
378 			memcpy(&mpte->mpte_u_src, src, src->sa_len);
379 		}
380 	}
381 
382 	error = mptcp_connectx(mpte, src, dst, ifscope, pcid);
383 
384 	/* If there is data, copy it */
385 	if (auio != NULL) {
386 		datalen = uio_resid(auio);
387 		socket_unlock(mp_so, 0);
388 		error = mp_so->so_proto->pr_usrreqs->pru_sosend(mp_so, NULL,
389 		    (uio_t) auio, NULL, NULL, 0);
390 
391 		if (error == 0 || error == EWOULDBLOCK) {
392 			*bytes_written = datalen - uio_resid(auio);
393 		}
394 
395 		if (error == EWOULDBLOCK) {
396 			error = EINPROGRESS;
397 		}
398 
399 		socket_lock(mp_so, 0);
400 	}
401 
402 out:
403 	return error;
404 }
405 
406 /*
407  * Handle SIOCGASSOCIDS ioctl for PF_MULTIPATH domain.
408  */
409 static int
mptcp_getassocids(struct mptses * mpte,uint32_t * cnt,user_addr_t aidp)410 mptcp_getassocids(struct mptses *mpte, uint32_t *cnt, user_addr_t aidp)
411 {
412 	/* MPTCP has at most 1 association */
413 	*cnt = (mpte->mpte_associd != SAE_ASSOCID_ANY) ? 1 : 0;
414 
415 	/* just asking how many there are? */
416 	if (aidp == USER_ADDR_NULL) {
417 		return 0;
418 	}
419 
420 	return copyout(&mpte->mpte_associd, aidp,
421 	           sizeof(mpte->mpte_associd));
422 }
423 
424 /*
425  * Handle SIOCGCONNIDS ioctl for PF_MULTIPATH domain.
426  */
427 static int
mptcp_getconnids(struct mptses * mpte,sae_associd_t aid,uint32_t * cnt,user_addr_t cidp)428 mptcp_getconnids(struct mptses *mpte, sae_associd_t aid, uint32_t *cnt,
429     user_addr_t cidp)
430 {
431 	struct mptsub *mpts;
432 	int error = 0;
433 
434 	if (aid != SAE_ASSOCID_ANY && aid != SAE_ASSOCID_ALL &&
435 	    aid != mpte->mpte_associd) {
436 		return EINVAL;
437 	}
438 
439 	*cnt = mpte->mpte_numflows;
440 
441 	/* just asking how many there are? */
442 	if (cidp == USER_ADDR_NULL) {
443 		return 0;
444 	}
445 
446 	TAILQ_FOREACH(mpts, &mpte->mpte_subflows, mpts_entry) {
447 		if ((error = copyout(&mpts->mpts_connid, cidp,
448 		    sizeof(mpts->mpts_connid))) != 0) {
449 			break;
450 		}
451 
452 		cidp += sizeof(mpts->mpts_connid);
453 	}
454 
455 	return error;
456 }
457 
458 /*
459  * Handle SIOCGCONNINFO ioctl for PF_MULTIPATH domain.
460  */
461 static int
mptcp_getconninfo(struct mptses * mpte,sae_connid_t * cid,uint32_t * flags,uint32_t * ifindex,int32_t * soerror,user_addr_t src,socklen_t * src_len,user_addr_t dst,socklen_t * dst_len,uint32_t * aux_type,user_addr_t aux_data,uint32_t * aux_len)462 mptcp_getconninfo(struct mptses *mpte, sae_connid_t *cid, uint32_t *flags,
463     uint32_t *ifindex, int32_t *soerror, user_addr_t src, socklen_t *src_len,
464     user_addr_t dst, socklen_t *dst_len, uint32_t *aux_type,
465     user_addr_t aux_data, uint32_t *aux_len)
466 {
467 	*flags = 0;
468 	*aux_type = 0;
469 	*ifindex = 0;
470 	*soerror = 0;
471 	struct mptcb *mp_tp = mpte->mpte_mptcb;
472 
473 	/* MPTCP-level global stats */
474 	if (*cid == SAE_CONNID_ALL) {
475 		struct socket *mp_so = mptetoso(mpte);
476 		struct conninfo_multipathtcp mptcp_ci;
477 		int error = 0;
478 
479 		if (*aux_len != 0 && *aux_len != sizeof(mptcp_ci)) {
480 			return EINVAL;
481 		}
482 
483 		if (mp_so->so_state & SS_ISCONNECTING) {
484 			*flags |= CIF_CONNECTING;
485 		}
486 		if (mp_so->so_state & SS_ISCONNECTED) {
487 			*flags |= CIF_CONNECTED;
488 		}
489 		if (mp_so->so_state & SS_ISDISCONNECTING) {
490 			*flags |= CIF_DISCONNECTING;
491 		}
492 		if (mp_so->so_state & SS_ISDISCONNECTED) {
493 			*flags |= CIF_DISCONNECTED;
494 		}
495 		if (!(mp_tp->mpt_flags & MPTCPF_FALLBACK_TO_TCP)) {
496 			*flags |= CIF_MP_CAPABLE;
497 		}
498 		if (mp_tp->mpt_flags & MPTCPF_FALLBACK_TO_TCP) {
499 			*flags |= CIF_MP_DEGRADED;
500 		}
501 		if (mp_tp->mpt_version == MPTCP_VERSION_1) {
502 			*flags |= CIF_MP_V1;
503 		}
504 
505 		*src_len = 0;
506 		*dst_len = 0;
507 
508 		*aux_type = CIAUX_MPTCP;
509 		*aux_len = sizeof(mptcp_ci);
510 
511 		if (aux_data != USER_ADDR_NULL) {
512 			const struct mptsub *mpts;
513 			int initial_info_set = 0;
514 			unsigned long i = 0;
515 
516 			bzero(&mptcp_ci, sizeof(mptcp_ci));
517 			mptcp_ci.mptcpci_subflow_count = mpte->mpte_numflows;
518 			mptcp_ci.mptcpci_switch_count = mpte->mpte_subflow_switches;
519 
520 			VERIFY(sizeof(mptcp_ci.mptcpci_itfstats) == sizeof(mpte->mpte_itfstats));
521 			memcpy(mptcp_ci.mptcpci_itfstats, mpte->mpte_itfstats, sizeof(mptcp_ci.mptcpci_itfstats));
522 
523 			TAILQ_FOREACH(mpts, &mpte->mpte_subflows, mpts_entry) {
524 				if (i >= sizeof(mptcp_ci.mptcpci_subflow_connids) / sizeof(sae_connid_t)) {
525 					break;
526 				}
527 				mptcp_ci.mptcpci_subflow_connids[i] = mpts->mpts_connid;
528 
529 				if (mpts->mpts_flags & MPTSF_INITIAL_SUB) {
530 					const struct inpcb *inp;
531 
532 					inp = sotoinpcb(mpts->mpts_socket);
533 
534 					mptcp_ci.mptcpci_init_rxbytes = inp->inp_stat->rxbytes;
535 					mptcp_ci.mptcpci_init_txbytes = inp->inp_stat->txbytes;
536 					initial_info_set = 1;
537 				}
538 
539 				mptcpstats_update(mptcp_ci.mptcpci_itfstats, mpts);
540 
541 				i++;
542 			}
543 
544 			if (initial_info_set == 0) {
545 				mptcp_ci.mptcpci_init_rxbytes = mpte->mpte_init_rxbytes;
546 				mptcp_ci.mptcpci_init_txbytes = mpte->mpte_init_txbytes;
547 			}
548 
549 			if (mpte->mpte_flags & MPTE_FIRSTPARTY) {
550 				mptcp_ci.mptcpci_flags |= MPTCPCI_FIRSTPARTY;
551 			}
552 
553 			error = copyout(&mptcp_ci, aux_data, sizeof(mptcp_ci));
554 			if (error != 0) {
555 				os_log_error(mptcp_log_handle, "%s - %lx: copyout failed: %d\n",
556 				    __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte), error);
557 				return error;
558 			}
559 		}
560 
561 		return 0;
562 	}
563 
564 	/* Any stats of any subflow */
565 	if (*cid == SAE_CONNID_ANY) {
566 		const struct mptsub *mpts;
567 		struct socket *so;
568 		const struct inpcb *inp;
569 		int error = 0;
570 
571 		mpts = TAILQ_FIRST(&mpte->mpte_subflows);
572 		if (mpts == NULL) {
573 			return ENXIO;
574 		}
575 
576 		so = mpts->mpts_socket;
577 		inp = sotoinpcb(so);
578 
579 		if (inp->inp_vflag & INP_IPV4) {
580 			error = in_getconninfo(so, SAE_CONNID_ANY, flags, ifindex,
581 			    soerror, src, src_len, dst, dst_len,
582 			    aux_type, aux_data, aux_len);
583 		} else {
584 			error = in6_getconninfo(so, SAE_CONNID_ANY, flags, ifindex,
585 			    soerror, src, src_len, dst, dst_len,
586 			    aux_type, aux_data, aux_len);
587 		}
588 
589 		if (error != 0) {
590 			os_log_error(mptcp_log_handle, "%s - %lx:error from in_getconninfo %d\n",
591 			    __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte), error);
592 			return error;
593 		}
594 
595 		if (mpts->mpts_flags & MPTSF_MP_CAPABLE) {
596 			*flags |= CIF_MP_CAPABLE;
597 		}
598 		if (mpts->mpts_flags & MPTSF_MP_DEGRADED) {
599 			*flags |= CIF_MP_DEGRADED;
600 		}
601 		if (mpts->mpts_flags & MPTSF_MP_READY) {
602 			*flags |= CIF_MP_READY;
603 		}
604 		if (mpts->mpts_flags & MPTSF_ACTIVE) {
605 			*flags |= CIF_MP_ACTIVE;
606 		}
607 		if (mp_tp->mpt_version == MPTCP_VERSION_1) {
608 			*flags |= CIF_MP_V1;
609 		}
610 
611 		return 0;
612 	} else {
613 		/* Per-interface stats */
614 		const struct mptsub *mpts, *orig_mpts = NULL;
615 		struct conninfo_tcp tcp_ci;
616 		const struct inpcb *inp;
617 		struct socket *so;
618 		int error = 0;
619 		int index;
620 
621 		/* cid is thus an ifindex - range-check first! */
622 		if (*cid > USHRT_MAX) {
623 			return EINVAL;
624 		}
625 
626 		bzero(&tcp_ci, sizeof(tcp_ci));
627 
628 		/* First, get a subflow to fill in the "regular" info. */
629 		TAILQ_FOREACH(mpts, &mpte->mpte_subflows, mpts_entry) {
630 			const struct ifnet *ifp = sotoinpcb(mpts->mpts_socket)->inp_last_outifp;
631 
632 			if (ifp && ifp->if_index == *cid) {
633 				break;
634 			}
635 		}
636 
637 		if (mpts == NULL) {
638 			/* No subflow there - well, let's just get the basic itf-info */
639 			goto interface_info;
640 		}
641 
642 		so = mpts->mpts_socket;
643 		inp = sotoinpcb(so);
644 
645 		/* Give it USER_ADDR_NULL, because we are doing this on our own */
646 		if (inp->inp_vflag & INP_IPV4) {
647 			error = in_getconninfo(so, SAE_CONNID_ANY, flags, ifindex,
648 			    soerror, src, src_len, dst, dst_len,
649 			    aux_type, USER_ADDR_NULL, aux_len);
650 		} else {
651 			error = in6_getconninfo(so, SAE_CONNID_ANY, flags, ifindex,
652 			    soerror, src, src_len, dst, dst_len,
653 			    aux_type, USER_ADDR_NULL, aux_len);
654 		}
655 
656 		if (error != 0) {
657 			os_log_error(mptcp_log_handle, "%s - %lx:error from in_getconninfo %d\n",
658 			    __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte), error);
659 			return error;
660 		}
661 
662 		/* ToDo: Nobody is reading these flags on subflows. Why bother ? */
663 		if (mpts->mpts_flags & MPTSF_MP_CAPABLE) {
664 			*flags |= CIF_MP_CAPABLE;
665 		}
666 		if (mpts->mpts_flags & MPTSF_MP_DEGRADED) {
667 			*flags |= CIF_MP_DEGRADED;
668 		}
669 		if (mpts->mpts_flags & MPTSF_MP_READY) {
670 			*flags |= CIF_MP_READY;
671 		}
672 		if (mpts->mpts_flags & MPTSF_ACTIVE) {
673 			*flags |= CIF_MP_ACTIVE;
674 		}
675 		if (mp_tp->mpt_version == MPTCP_VERSION_1) {
676 			*flags |= CIF_MP_V1;
677 		}
678 
679 		/*
680 		 * Now, we gather the metrics (aka., tcp_info) and roll them in
681 		 * across all subflows of this interface to build an aggregated
682 		 * view.
683 		 *
684 		 * We take the TCP_INFO from the first subflow as the "master",
685 		 * feeding into those fields that we do not roll.
686 		 */
687 		if (aux_data != USER_ADDR_NULL) {
688 			tcp_getconninfo(so, &tcp_ci);
689 
690 			orig_mpts = mpts;
691 			TAILQ_FOREACH(mpts, &mpte->mpte_subflows, mpts_entry) {
692 				const struct inpcb *mptsinp = sotoinpcb(mpts->mpts_socket);
693 				const struct ifnet *ifp;
694 
695 				ifp = mptsinp->inp_last_outifp;
696 
697 				if (ifp == NULL || ifp->if_index != *cid || mpts == orig_mpts) {
698 					continue;
699 				}
700 
701 				/* Roll the itf-stats into the tcp_info */
702 				tcp_ci.tcpci_tcp_info.tcpi_txbytes +=
703 				    mptsinp->inp_stat->txbytes;
704 				tcp_ci.tcpci_tcp_info.tcpi_rxbytes +=
705 				    mptsinp->inp_stat->rxbytes;
706 
707 				tcp_ci.tcpci_tcp_info.tcpi_wifi_txbytes +=
708 				    mptsinp->inp_wstat->txbytes;
709 				tcp_ci.tcpci_tcp_info.tcpi_wifi_rxbytes +=
710 				    mptsinp->inp_wstat->rxbytes;
711 
712 				tcp_ci.tcpci_tcp_info.tcpi_wired_txbytes +=
713 				    mptsinp->inp_Wstat->txbytes;
714 				tcp_ci.tcpci_tcp_info.tcpi_wired_rxbytes +=
715 				    mptsinp->inp_Wstat->rxbytes;
716 
717 				tcp_ci.tcpci_tcp_info.tcpi_cell_txbytes +=
718 				    mptsinp->inp_cstat->txbytes;
719 				tcp_ci.tcpci_tcp_info.tcpi_cell_rxbytes +=
720 				    mptsinp->inp_cstat->rxbytes;
721 			}
722 		}
723 
724 interface_info:
725 		*aux_type = CIAUX_TCP;
726 		if (*aux_len == 0) {
727 			*aux_len = sizeof(tcp_ci);
728 		} else if (aux_data != USER_ADDR_NULL) {
729 			boolean_t create;
730 
731 			/*
732 			 * Finally, old subflows might have been closed - we
733 			 * want this data as well, so grab it from the interface
734 			 * stats.
735 			 */
736 			create = orig_mpts != NULL;
737 
738 			/*
739 			 * When we found a subflow, we are willing to create a stats-index
740 			 * because we have some data to return. If there isn't a subflow,
741 			 * nor anything in the stats, return EINVAL. Because the
742 			 * ifindex belongs to something that doesn't exist.
743 			 */
744 			index = mptcpstats_get_index_by_ifindex(mpte->mpte_itfstats, (u_short)(*cid), false);
745 			if (index == -1) {
746 				os_log_error(mptcp_log_handle,
747 				    "%s - %lx: Asking for too many ifindex: %u subcount %u, mpts? %s\n",
748 				    __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte),
749 				    *cid, mpte->mpte_numflows,
750 				    orig_mpts ? "yes" : "no");
751 
752 				if (orig_mpts == NULL) {
753 					return EINVAL;
754 				}
755 			} else {
756 				struct mptcp_itf_stats *stats;
757 
758 				stats = &mpte->mpte_itfstats[index];
759 
760 				/* Roll the itf-stats into the tcp_info */
761 				tcp_ci.tcpci_tcp_info.tcpi_last_outif = *cid;
762 				tcp_ci.tcpci_tcp_info.tcpi_txbytes +=
763 				    stats->mpis_txbytes;
764 				tcp_ci.tcpci_tcp_info.tcpi_rxbytes +=
765 				    stats->mpis_rxbytes;
766 
767 				tcp_ci.tcpci_tcp_info.tcpi_wifi_txbytes +=
768 				    stats->mpis_wifi_txbytes;
769 				tcp_ci.tcpci_tcp_info.tcpi_wifi_rxbytes +=
770 				    stats->mpis_wifi_rxbytes;
771 
772 				tcp_ci.tcpci_tcp_info.tcpi_wired_txbytes +=
773 				    stats->mpis_wired_txbytes;
774 				tcp_ci.tcpci_tcp_info.tcpi_wired_rxbytes +=
775 				    stats->mpis_wired_rxbytes;
776 
777 				tcp_ci.tcpci_tcp_info.tcpi_cell_txbytes +=
778 				    stats->mpis_cell_txbytes;
779 				tcp_ci.tcpci_tcp_info.tcpi_cell_rxbytes +=
780 				    stats->mpis_cell_rxbytes;
781 			}
782 
783 			*aux_len = min(*aux_len, sizeof(tcp_ci));
784 			error = copyout(&tcp_ci, aux_data, *aux_len);
785 			if (error != 0) {
786 				return error;
787 			}
788 		}
789 	}
790 
791 	return 0;
792 }
793 
794 /*
795  * User-protocol pru_control callback.
796  */
797 static int
mptcp_usr_control(struct socket * mp_so,u_long cmd,caddr_t data,struct ifnet * ifp,struct proc * p)798 mptcp_usr_control(struct socket *mp_so, u_long cmd, caddr_t data,
799     struct ifnet *ifp, struct proc *p)
800 {
801 #pragma unused(ifp, p)
802 	struct mppcb *mpp = mpsotomppcb(mp_so);
803 	struct mptses *mpte;
804 	int error = 0;
805 
806 	if (mpp == NULL || mpp->mpp_state == MPPCB_STATE_DEAD) {
807 		error = EINVAL;
808 		goto out;
809 	}
810 	mpte = mptompte(mpp);
811 
812 	switch (cmd) {
813 	case SIOCGASSOCIDS32: {         /* struct so_aidreq32 */
814 		struct so_aidreq32 aidr;
815 		bcopy(data, &aidr, sizeof(aidr));
816 		error = mptcp_getassocids(mpte, &aidr.sar_cnt,
817 		    aidr.sar_aidp);
818 		if (error == 0) {
819 			bcopy(&aidr, data, sizeof(aidr));
820 		}
821 		break;
822 	}
823 
824 	case SIOCGASSOCIDS64: {         /* struct so_aidreq64 */
825 		struct so_aidreq64 aidr;
826 		bcopy(data, &aidr, sizeof(aidr));
827 		error = mptcp_getassocids(mpte, &aidr.sar_cnt,
828 		    (user_addr_t)aidr.sar_aidp);
829 		if (error == 0) {
830 			bcopy(&aidr, data, sizeof(aidr));
831 		}
832 		break;
833 	}
834 
835 	case SIOCGCONNIDS32: {          /* struct so_cidreq32 */
836 		struct so_cidreq32 cidr;
837 		bcopy(data, &cidr, sizeof(cidr));
838 		error = mptcp_getconnids(mpte, cidr.scr_aid, &cidr.scr_cnt,
839 		    cidr.scr_cidp);
840 		if (error == 0) {
841 			bcopy(&cidr, data, sizeof(cidr));
842 		}
843 		break;
844 	}
845 
846 	case SIOCGCONNIDS64: {          /* struct so_cidreq64 */
847 		struct so_cidreq64 cidr;
848 		bcopy(data, &cidr, sizeof(cidr));
849 		error = mptcp_getconnids(mpte, cidr.scr_aid, &cidr.scr_cnt,
850 		    (user_addr_t)cidr.scr_cidp);
851 		if (error == 0) {
852 			bcopy(&cidr, data, sizeof(cidr));
853 		}
854 		break;
855 	}
856 
857 	case SIOCGCONNINFO32: {         /* struct so_cinforeq32 */
858 		struct so_cinforeq32 cifr;
859 		bcopy(data, &cifr, sizeof(cifr));
860 		error = mptcp_getconninfo(mpte, &cifr.scir_cid,
861 		    &cifr.scir_flags, &cifr.scir_ifindex, &cifr.scir_error,
862 		    cifr.scir_src, &cifr.scir_src_len, cifr.scir_dst,
863 		    &cifr.scir_dst_len, &cifr.scir_aux_type, cifr.scir_aux_data,
864 		    &cifr.scir_aux_len);
865 		if (error == 0) {
866 			bcopy(&cifr, data, sizeof(cifr));
867 		}
868 		break;
869 	}
870 
871 	case SIOCGCONNINFO64: {         /* struct so_cinforeq64 */
872 		struct so_cinforeq64 cifr;
873 		bcopy(data, &cifr, sizeof(cifr));
874 		error = mptcp_getconninfo(mpte, &cifr.scir_cid,
875 		    &cifr.scir_flags, &cifr.scir_ifindex, &cifr.scir_error,
876 		    (user_addr_t)cifr.scir_src, &cifr.scir_src_len,
877 		    (user_addr_t)cifr.scir_dst, &cifr.scir_dst_len,
878 		    &cifr.scir_aux_type, (user_addr_t)cifr.scir_aux_data,
879 		    &cifr.scir_aux_len);
880 		if (error == 0) {
881 			bcopy(&cifr, data, sizeof(cifr));
882 		}
883 		break;
884 	}
885 
886 	default:
887 		error = EOPNOTSUPP;
888 		break;
889 	}
890 out:
891 	return error;
892 }
893 
894 static int
mptcp_disconnect(struct mptses * mpte)895 mptcp_disconnect(struct mptses *mpte)
896 {
897 	struct socket *mp_so;
898 	struct mptcb *mp_tp;
899 	int error = 0;
900 
901 	mp_so = mptetoso(mpte);
902 	mp_tp = mpte->mpte_mptcb;
903 
904 	/* if we're not detached, go thru socket state checks */
905 	if (!(mp_so->so_flags & SOF_PCBCLEARING) && !(mp_so->so_flags & SOF_DEFUNCT)) {
906 		if (!(mp_so->so_state & (SS_ISCONNECTED |
907 		    SS_ISCONNECTING))) {
908 			error = ENOTCONN;
909 			goto out;
910 		}
911 		if (mp_so->so_state & SS_ISDISCONNECTING) {
912 			error = EALREADY;
913 			goto out;
914 		}
915 	}
916 
917 	mptcp_cancel_all_timers(mp_tp);
918 	if (mp_tp->mpt_state < MPTCPS_ESTABLISHED) {
919 		mptcp_close(mpte, mp_tp);
920 	} else if (((mp_so->so_options & SO_LINGER) &&
921 	    mp_so->so_linger == 0) ||
922 	    (mp_so->so_flags1 & SOF1_DEFUNCTINPROG)) {
923 		mptcp_drop(mpte, mp_tp, 0);
924 	} else {
925 		soisdisconnecting(mp_so);
926 		sbflush(&mp_so->so_rcv);
927 		if (mptcp_usrclosed(mpte) != NULL) {
928 			mptcp_output(mpte);
929 		}
930 	}
931 
932 	if (error == 0) {
933 		mptcp_subflow_workloop(mpte);
934 	}
935 
936 out:
937 	return error;
938 }
939 
940 /*
941  * Wrapper function to support disconnect on socket
942  */
943 static int
mptcp_usr_disconnect(struct socket * mp_so)944 mptcp_usr_disconnect(struct socket *mp_so)
945 {
946 	return mptcp_disconnect(mpsotompte(mp_so));
947 }
948 
949 /*
950  * User-protocol pru_disconnectx callback.
951  */
952 static int
mptcp_usr_disconnectx(struct socket * mp_so,sae_associd_t aid,sae_connid_t cid)953 mptcp_usr_disconnectx(struct socket *mp_so, sae_associd_t aid, sae_connid_t cid)
954 {
955 	if (aid != SAE_ASSOCID_ANY && aid != SAE_ASSOCID_ALL) {
956 		return EINVAL;
957 	}
958 
959 	if (cid != SAE_CONNID_ANY && cid != SAE_CONNID_ALL) {
960 		return EINVAL;
961 	}
962 
963 	return mptcp_usr_disconnect(mp_so);
964 }
965 
966 void
mptcp_finish_usrclosed(struct mptses * mpte)967 mptcp_finish_usrclosed(struct mptses *mpte)
968 {
969 	struct mptcb *mp_tp = mpte->mpte_mptcb;
970 	struct socket *mp_so = mptetoso(mpte);
971 
972 	if (mp_tp->mpt_state == MPTCPS_CLOSED || mp_tp->mpt_state == MPTCPS_TERMINATE) {
973 		mpte = mptcp_close(mpte, mp_tp);
974 	} else if (mp_tp->mpt_state >= MPTCPS_FIN_WAIT_2) {
975 		soisdisconnected(mp_so);
976 	} else {
977 		struct mptsub *mpts;
978 
979 		TAILQ_FOREACH(mpts, &mpte->mpte_subflows, mpts_entry) {
980 			if ((mp_so->so_state & (SS_CANTRCVMORE | SS_CANTSENDMORE)) ==
981 			    (SS_CANTRCVMORE | SS_CANTSENDMORE)) {
982 				mptcp_subflow_disconnect(mpte, mpts);
983 			} else {
984 				mptcp_subflow_shutdown(mpte, mpts);
985 			}
986 		}
987 	}
988 }
989 
990 /*
991  * User issued close, and wish to trail thru shutdown states.
992  */
993 static struct mptses *
mptcp_usrclosed(struct mptses * mpte)994 mptcp_usrclosed(struct mptses *mpte)
995 {
996 	struct mptcb *mp_tp = mpte->mpte_mptcb;
997 
998 	mptcp_close_fsm(mp_tp, MPCE_CLOSE);
999 
1000 	/* Not everything has been acknowledged - don't close the subflows! */
1001 	if (mp_tp->mpt_state != MPTCPS_TERMINATE &&
1002 	    mp_tp->mpt_sndnxt + 1 != mp_tp->mpt_sndmax) {
1003 		return mpte;
1004 	}
1005 
1006 	mptcp_finish_usrclosed(mpte);
1007 
1008 	return mpte;
1009 }
1010 
1011 /*
1012  * After a receive, possible send some update to peer.
1013  */
1014 static int
mptcp_usr_rcvd(struct socket * mp_so,int flags)1015 mptcp_usr_rcvd(struct socket *mp_so, int flags)
1016 {
1017 #pragma unused(flags)
1018 	struct mppcb *mpp = mpsotomppcb(mp_so);
1019 	struct mptses *mpte;
1020 	struct mptsub *mpts;
1021 	int error = 0;
1022 
1023 	if (mpp == NULL || mpp->mpp_state == MPPCB_STATE_DEAD) {
1024 		error = EINVAL;
1025 		goto out;
1026 	}
1027 
1028 	mpte = mptompte(mpp);
1029 
1030 	TAILQ_FOREACH(mpts, &mpte->mpte_subflows, mpts_entry) {
1031 		struct socket *so = mpts->mpts_socket;
1032 
1033 		if (so->so_proto->pr_flags & PR_WANTRCVD && so->so_pcb != NULL) {
1034 			(*so->so_proto->pr_usrreqs->pru_rcvd)(so, 0);
1035 		}
1036 	}
1037 
1038 	error = mptcp_output(mpte);
1039 out:
1040 	return error;
1041 }
1042 
1043 /*
1044  * Do a send by putting data in the output queue.
1045  */
1046 static int
mptcp_usr_send(struct socket * mp_so,int prus_flags,struct mbuf * m,struct sockaddr * nam,struct mbuf * control,struct proc * p)1047 mptcp_usr_send(struct socket *mp_so, int prus_flags, struct mbuf *m,
1048     struct sockaddr *nam, struct mbuf *control, struct proc *p)
1049 {
1050 #pragma unused(nam, p)
1051 	struct mppcb *mpp = mpsotomppcb(mp_so);
1052 	struct mptses *mpte;
1053 	int error = 0;
1054 
1055 	if (prus_flags & (PRUS_OOB | PRUS_EOF)) {
1056 		error = EOPNOTSUPP;
1057 		goto out;
1058 	}
1059 
1060 	if (nam != NULL) {
1061 		error = EOPNOTSUPP;
1062 		goto out;
1063 	}
1064 
1065 	if (control != NULL && control->m_len != 0) {
1066 		error = EOPNOTSUPP;
1067 		goto out;
1068 	}
1069 
1070 	if (mpp == NULL || mpp->mpp_state == MPPCB_STATE_DEAD) {
1071 		error = ECONNRESET;
1072 		goto out;
1073 	}
1074 	mpte = mptompte(mpp);
1075 	VERIFY(mpte != NULL);
1076 
1077 	if (!(mp_so->so_state & SS_ISCONNECTED) &&
1078 	    !(mp_so->so_flags1 & SOF1_PRECONNECT_DATA)) {
1079 		error = ENOTCONN;
1080 		goto out;
1081 	}
1082 
1083 	mptcp_insert_dsn(mpp, m);
1084 	VERIFY(mp_so->so_snd.sb_flags & SB_NOCOMPRESS);
1085 	sbappendstream(&mp_so->so_snd, m);
1086 	m = NULL;
1087 
1088 	error = mptcp_output(mpte);
1089 	if (error != 0) {
1090 		goto out;
1091 	}
1092 
1093 	if (mp_so->so_state & SS_ISCONNECTING) {
1094 		if (mp_so->so_state & SS_NBIO) {
1095 			error = EWOULDBLOCK;
1096 		} else {
1097 			error = sbwait(&mp_so->so_snd);
1098 		}
1099 	}
1100 
1101 out:
1102 	if (error) {
1103 		if (m != NULL) {
1104 			m_freem(m);
1105 		}
1106 		if (control != NULL) {
1107 			m_freem(control);
1108 		}
1109 	}
1110 	return error;
1111 }
1112 
1113 /*
1114  * Mark the MPTCP connection as being incapable of further output.
1115  */
1116 static int
mptcp_usr_shutdown(struct socket * mp_so)1117 mptcp_usr_shutdown(struct socket *mp_so)
1118 {
1119 	struct mppcb *mpp = mpsotomppcb(mp_so);
1120 	struct mptses *mpte;
1121 	int error = 0;
1122 
1123 	if (mpp == NULL || mpp->mpp_state == MPPCB_STATE_DEAD) {
1124 		error = EINVAL;
1125 		goto out;
1126 	}
1127 	mpte = mptompte(mpp);
1128 	VERIFY(mpte != NULL);
1129 
1130 	socantsendmore(mp_so);
1131 
1132 	mpte = mptcp_usrclosed(mpte);
1133 	if (mpte != NULL) {
1134 		error = mptcp_output(mpte);
1135 	}
1136 out:
1137 	return error;
1138 }
1139 
1140 /*
1141  * Copy the contents of uio into a properly sized mbuf chain.
1142  */
1143 static int
mptcp_uiotombuf(struct uio * uio,int how,user_ssize_t space,struct mbuf ** top)1144 mptcp_uiotombuf(struct uio *uio, int how, user_ssize_t space, struct mbuf **top)
1145 {
1146 	struct mbuf *m, *mb, *nm = NULL, *mtail = NULL;
1147 	int progress, len, error;
1148 	user_ssize_t resid, tot;
1149 
1150 	VERIFY(top != NULL && *top == NULL);
1151 
1152 	/*
1153 	 * space can be zero or an arbitrary large value bound by
1154 	 * the total data supplied by the uio.
1155 	 */
1156 	resid = uio_resid(uio);
1157 	if (space > 0) {
1158 		tot = MIN(resid, space);
1159 	} else {
1160 		tot = resid;
1161 	}
1162 
1163 	if (tot < 0 || tot > INT_MAX) {
1164 		return EINVAL;
1165 	}
1166 
1167 	len = (int)tot;
1168 	if (len == 0) {
1169 		len = 1;
1170 	}
1171 
1172 	/* Loop and append maximum sized mbufs to the chain tail. */
1173 	while (len > 0) {
1174 		uint32_t m_needed = 1;
1175 
1176 		if (njcl > 0 && len > MBIGCLBYTES) {
1177 			mb = m_getpackets_internal(&m_needed, 1,
1178 			    how, 1, M16KCLBYTES);
1179 		} else if (len > MCLBYTES) {
1180 			mb = m_getpackets_internal(&m_needed, 1,
1181 			    how, 1, MBIGCLBYTES);
1182 		} else if (len >= (signed)MINCLSIZE) {
1183 			mb = m_getpackets_internal(&m_needed, 1,
1184 			    how, 1, MCLBYTES);
1185 		} else {
1186 			mb = m_gethdr(how, MT_DATA);
1187 		}
1188 
1189 		/* Fail the whole operation if one mbuf can't be allocated. */
1190 		if (mb == NULL) {
1191 			if (nm != NULL) {
1192 				m_freem(nm);
1193 			}
1194 			return ENOBUFS;
1195 		}
1196 
1197 		/* Book keeping. */
1198 		VERIFY(mb->m_flags & M_PKTHDR);
1199 		len -= ((mb->m_flags & M_EXT) ? mb->m_ext.ext_size : MHLEN);
1200 		if (mtail != NULL) {
1201 			mtail->m_next = mb;
1202 		} else {
1203 			nm = mb;
1204 		}
1205 		mtail = mb;
1206 	}
1207 
1208 	m = nm;
1209 
1210 	progress = 0;
1211 	/* Fill all mbufs with uio data and update header information. */
1212 	for (mb = m; mb != NULL; mb = mb->m_next) {
1213 		/* tot >= 0 && tot <= INT_MAX (see above) */
1214 		len = MIN((int)M_TRAILINGSPACE(mb), (int)(tot - progress));
1215 
1216 		error = uiomove(mtod(mb, char *), len, uio);
1217 		if (error != 0) {
1218 			m_freem(m);
1219 			return error;
1220 		}
1221 
1222 		/* each mbuf is M_PKTHDR chained via m_next */
1223 		mb->m_len = len;
1224 		mb->m_pkthdr.len = len;
1225 
1226 		progress += len;
1227 	}
1228 	VERIFY(progress == tot);
1229 	*top = m;
1230 	return 0;
1231 }
1232 
1233 /*
1234  * MPTCP socket protocol-user socket send routine, derived from sosend().
1235  */
1236 static int
mptcp_usr_sosend(struct socket * mp_so,struct sockaddr * addr,struct uio * uio,struct mbuf * top,struct mbuf * control,int flags)1237 mptcp_usr_sosend(struct socket *mp_so, struct sockaddr *addr, struct uio *uio,
1238     struct mbuf *top, struct mbuf *control, int flags)
1239 {
1240 #pragma unused(addr)
1241 	user_ssize_t resid, space;
1242 	int error, sendflags;
1243 	struct proc *p = current_proc();
1244 	int sblocked = 0;
1245 
1246 	/* UIO is required for now, due to per-mbuf M_PKTHDR constrains */
1247 	if (uio == NULL || top != NULL) {
1248 		error = EINVAL;
1249 		goto out;
1250 	}
1251 	resid = uio_resid(uio);
1252 
1253 	socket_lock(mp_so, 1);
1254 	so_update_last_owner_locked(mp_so, p);
1255 	so_update_policy(mp_so);
1256 
1257 	VERIFY(mp_so->so_type == SOCK_STREAM);
1258 	VERIFY(!(mp_so->so_flags & SOF_MP_SUBFLOW));
1259 
1260 	if (flags & (MSG_OOB | MSG_DONTROUTE)) {
1261 		error = EOPNOTSUPP;
1262 		socket_unlock(mp_so, 1);
1263 		goto out;
1264 	}
1265 
1266 	/*
1267 	 * In theory resid should be unsigned.  However, space must be
1268 	 * signed, as it might be less than 0 if we over-committed, and we
1269 	 * must use a signed comparison of space and resid.  On the other
1270 	 * hand, a negative resid causes us to loop sending 0-length
1271 	 * segments to the protocol.
1272 	 */
1273 	if (resid < 0 || resid > INT_MAX ||
1274 	    (flags & MSG_EOR) || control != NULL) {
1275 		error = EINVAL;
1276 		socket_unlock(mp_so, 1);
1277 		goto out;
1278 	}
1279 
1280 	OSIncrementAtomicLong(&p->p_stats->p_ru.ru_msgsnd);
1281 
1282 	do {
1283 		error = sosendcheck(mp_so, NULL, resid, 0, 0, flags,
1284 		    &sblocked);
1285 		if (error != 0) {
1286 			goto release;
1287 		}
1288 
1289 		space = sbspace(&mp_so->so_snd);
1290 		do {
1291 			socket_unlock(mp_so, 0);
1292 			/*
1293 			 * Copy the data from userland into an mbuf chain.
1294 			 */
1295 			error = mptcp_uiotombuf(uio, M_WAITOK, space, &top);
1296 			if (error != 0) {
1297 				socket_lock(mp_so, 0);
1298 				goto release;
1299 			}
1300 			VERIFY(top != NULL);
1301 			space -= resid - uio_resid(uio);
1302 			resid = uio_resid(uio);
1303 			socket_lock(mp_so, 0);
1304 
1305 			/*
1306 			 * Compute flags here, for pru_send and NKEs.
1307 			 */
1308 			sendflags = (resid > 0 && space > 0) ?
1309 			    PRUS_MORETOCOME : 0;
1310 
1311 			/*
1312 			 * Socket filter processing
1313 			 */
1314 			VERIFY(control == NULL);
1315 			error = sflt_data_out(mp_so, NULL, &top, &control, 0);
1316 			if (error != 0) {
1317 				if (error == EJUSTRETURN) {
1318 					error = 0;
1319 					top = NULL;
1320 					/* always free control if any */
1321 				}
1322 				goto release;
1323 			}
1324 			if (control != NULL) {
1325 				m_freem(control);
1326 				control = NULL;
1327 			}
1328 
1329 			/*
1330 			 * Pass data to protocol.
1331 			 */
1332 			error = (*mp_so->so_proto->pr_usrreqs->pru_send)
1333 			    (mp_so, sendflags, top, NULL, NULL, p);
1334 
1335 			top = NULL;
1336 			if (error != 0) {
1337 				goto release;
1338 			}
1339 		} while (resid != 0 && space > 0);
1340 	} while (resid != 0);
1341 
1342 release:
1343 	if (sblocked) {
1344 		sbunlock(&mp_so->so_snd, FALSE); /* will unlock socket */
1345 	} else {
1346 		socket_unlock(mp_so, 1);
1347 	}
1348 out:
1349 	if (top != NULL) {
1350 		m_freem(top);
1351 	}
1352 	if (control != NULL) {
1353 		m_freem(control);
1354 	}
1355 
1356 	soclearfastopen(mp_so);
1357 
1358 	return error;
1359 }
1360 
1361 /*
1362  * Called to filter SOPT_{SET,GET} for SOL_SOCKET level socket options.
1363  * This routine simply indicates to the caller whether or not to proceed
1364  * further with the given socket option.  This is invoked by sosetoptlock()
1365  * and sogetoptlock().
1366  */
1367 static int
mptcp_usr_socheckopt(struct socket * mp_so,struct sockopt * sopt)1368 mptcp_usr_socheckopt(struct socket *mp_so, struct sockopt *sopt)
1369 {
1370 #pragma unused(mp_so)
1371 	int error = 0;
1372 
1373 	VERIFY(sopt->sopt_level == SOL_SOCKET);
1374 
1375 	/*
1376 	 * We could check for sopt_dir (set/get) here, but we'll just
1377 	 * let the caller deal with it as appropriate; therefore the
1378 	 * following is a superset of the socket options which we
1379 	 * allow for set/get.
1380 	 *
1381 	 * XXX: [email protected]
1382 	 *
1383 	 * Need to consider the following cases:
1384 	 *
1385 	 *   a.	Certain socket options don't have a clear definition
1386 	 *	on the expected behavior post connect(2).  At the time
1387 	 *	those options are issued on the MP socket, there may
1388 	 *	be existing subflow sockets that are already connected.
1389 	 */
1390 	switch (sopt->sopt_name) {
1391 	case SO_LINGER:                         /* MP */
1392 	case SO_LINGER_SEC:                     /* MP */
1393 	case SO_TYPE:                           /* MP */
1394 	case SO_NREAD:                          /* MP */
1395 	case SO_NWRITE:                         /* MP */
1396 	case SO_ERROR:                          /* MP */
1397 	case SO_SNDBUF:                         /* MP */
1398 	case SO_RCVBUF:                         /* MP */
1399 	case SO_SNDLOWAT:                       /* MP */
1400 	case SO_RCVLOWAT:                       /* MP */
1401 	case SO_SNDTIMEO:                       /* MP */
1402 	case SO_RCVTIMEO:                       /* MP */
1403 	case SO_NKE:                            /* MP */
1404 	case SO_NOSIGPIPE:                      /* MP */
1405 	case SO_NOADDRERR:                      /* MP */
1406 	case SO_LABEL:                          /* MP */
1407 	case SO_PEERLABEL:                      /* MP */
1408 	case SO_DEFUNCTIT:                      /* MP */
1409 	case SO_DEFUNCTOK:                      /* MP */
1410 	case SO_ISDEFUNCT:                      /* MP */
1411 	case SO_TRAFFIC_CLASS_DBG:              /* MP */
1412 	case SO_DELEGATED:                      /* MP */
1413 	case SO_DELEGATED_UUID:                 /* MP */
1414 #if NECP
1415 	case SO_NECP_ATTRIBUTES:
1416 	case SO_NECP_CLIENTUUID:
1417 #endif /* NECP */
1418 	case SO_MPKL_SEND_INFO:
1419 		/*
1420 		 * Tell the caller that these options are to be processed.
1421 		 */
1422 		break;
1423 
1424 	case SO_DEBUG:                          /* MP + subflow */
1425 	case SO_KEEPALIVE:                      /* MP + subflow */
1426 	case SO_USELOOPBACK:                    /* MP + subflow */
1427 	case SO_RANDOMPORT:                     /* MP + subflow */
1428 	case SO_TRAFFIC_CLASS:                  /* MP + subflow */
1429 	case SO_RECV_TRAFFIC_CLASS:             /* MP + subflow */
1430 	case SO_PRIVILEGED_TRAFFIC_CLASS:       /* MP + subflow */
1431 	case SO_RECV_ANYIF:                     /* MP + subflow */
1432 	case SO_RESTRICTIONS:                   /* MP + subflow */
1433 	case SO_FLUSH:                          /* MP + subflow */
1434 	case SO_NOWAKEFROMSLEEP:
1435 	case SO_NOAPNFALLBK:
1436 	case SO_MARK_CELLFALLBACK:
1437 	case SO_MARK_CELLFALLBACK_UUID:
1438 	case SO_MARK_KNOWN_TRACKER:
1439 	case SO_MARK_KNOWN_TRACKER_NON_APP_INITIATED:
1440 	case SO_MARK_APPROVED_APP_DOMAIN:
1441 	case SO_FALLBACK_MODE:
1442 		/*
1443 		 * Tell the caller that these options are to be processed;
1444 		 * these will also be recorded later by mptcp_setopt().
1445 		 *
1446 		 * NOTE: Only support integer option value for now.
1447 		 */
1448 		if (sopt->sopt_valsize != sizeof(int)) {
1449 			error = EINVAL;
1450 		}
1451 		break;
1452 
1453 	default:
1454 		/*
1455 		 * Tell the caller to stop immediately and return an error.
1456 		 */
1457 		error = ENOPROTOOPT;
1458 		break;
1459 	}
1460 
1461 	return error;
1462 }
1463 
1464 /*
1465  * Issue SOPT_SET for all MPTCP subflows (for integer option values.)
1466  */
1467 static int
mptcp_setopt_apply(struct mptses * mpte,struct mptopt * mpo)1468 mptcp_setopt_apply(struct mptses *mpte, struct mptopt *mpo)
1469 {
1470 	struct socket *mp_so;
1471 	struct mptsub *mpts;
1472 	struct mptopt smpo;
1473 	int error = 0;
1474 
1475 	/* just bail now if this isn't applicable to subflow sockets */
1476 	if (!(mpo->mpo_flags & MPOF_SUBFLOW_OK)) {
1477 		error = ENOPROTOOPT;
1478 		goto out;
1479 	}
1480 
1481 	/*
1482 	 * Skip those that are handled internally; these options
1483 	 * should not have been recorded and marked with the
1484 	 * MPOF_SUBFLOW_OK by mptcp_setopt(), but just in case.
1485 	 */
1486 	if (mpo->mpo_level == SOL_SOCKET &&
1487 	    (mpo->mpo_name == SO_NOSIGPIPE || mpo->mpo_name == SO_NOADDRERR)) {
1488 		error = ENOPROTOOPT;
1489 		goto out;
1490 	}
1491 
1492 	mp_so = mptetoso(mpte);
1493 
1494 	/*
1495 	 * Don't bother going further if there's no subflow; mark the option
1496 	 * with MPOF_INTERIM so that we know whether or not to remove this
1497 	 * option upon encountering an error while issuing it during subflow
1498 	 * socket creation.
1499 	 */
1500 	if (mpte->mpte_numflows == 0) {
1501 		VERIFY(TAILQ_EMPTY(&mpte->mpte_subflows));
1502 		mpo->mpo_flags |= MPOF_INTERIM;
1503 		/* return success */
1504 		goto out;
1505 	}
1506 
1507 	bzero(&smpo, sizeof(smpo));
1508 	smpo.mpo_flags |= MPOF_SUBFLOW_OK;
1509 	smpo.mpo_level = mpo->mpo_level;
1510 	smpo.mpo_name = mpo->mpo_name;
1511 
1512 	/* grab exisiting values in case we need to rollback */
1513 	TAILQ_FOREACH(mpts, &mpte->mpte_subflows, mpts_entry) {
1514 		struct socket *so;
1515 
1516 		mpts->mpts_flags &= ~(MPTSF_SOPT_OLDVAL | MPTSF_SOPT_INPROG);
1517 		mpts->mpts_oldintval = 0;
1518 		smpo.mpo_intval = 0;
1519 		VERIFY(mpts->mpts_socket != NULL);
1520 		so = mpts->mpts_socket;
1521 		if (mptcp_subflow_sogetopt(mpte, so, &smpo) == 0) {
1522 			mpts->mpts_flags |= MPTSF_SOPT_OLDVAL;
1523 			mpts->mpts_oldintval = smpo.mpo_intval;
1524 		}
1525 	}
1526 
1527 	/* apply socket option */
1528 	TAILQ_FOREACH(mpts, &mpte->mpte_subflows, mpts_entry) {
1529 		struct socket *so;
1530 
1531 		mpts->mpts_flags |= MPTSF_SOPT_INPROG;
1532 		VERIFY(mpts->mpts_socket != NULL);
1533 		so = mpts->mpts_socket;
1534 		error = mptcp_subflow_sosetopt(mpte, mpts, mpo);
1535 		if (error != 0) {
1536 			break;
1537 		}
1538 	}
1539 
1540 	/* cleanup, and rollback if needed */
1541 	TAILQ_FOREACH(mpts, &mpte->mpte_subflows, mpts_entry) {
1542 		struct socket *so;
1543 
1544 		if (!(mpts->mpts_flags & MPTSF_SOPT_INPROG)) {
1545 			/* clear in case it's set */
1546 			mpts->mpts_flags &= ~MPTSF_SOPT_OLDVAL;
1547 			mpts->mpts_oldintval = 0;
1548 			continue;
1549 		}
1550 		if (!(mpts->mpts_flags & MPTSF_SOPT_OLDVAL)) {
1551 			mpts->mpts_flags &= ~MPTSF_SOPT_INPROG;
1552 			VERIFY(mpts->mpts_oldintval == 0);
1553 			continue;
1554 		}
1555 		/* error during sosetopt, so roll it back */
1556 		if (error != 0) {
1557 			VERIFY(mpts->mpts_socket != NULL);
1558 			so = mpts->mpts_socket;
1559 			smpo.mpo_intval = mpts->mpts_oldintval;
1560 			mptcp_subflow_sosetopt(mpte, mpts, &smpo);
1561 		}
1562 		mpts->mpts_oldintval = 0;
1563 		mpts->mpts_flags &= ~(MPTSF_SOPT_OLDVAL | MPTSF_SOPT_INPROG);
1564 	}
1565 
1566 out:
1567 	return error;
1568 }
1569 
1570 /*
1571  * Handle SOPT_SET for socket options issued on MP socket.
1572  */
1573 static int
mptcp_setopt(struct mptses * mpte,struct sockopt * sopt)1574 mptcp_setopt(struct mptses *mpte, struct sockopt *sopt)
1575 {
1576 	int error = 0, optval = 0, level, optname, rec = 1;
1577 	struct mptopt smpo, *mpo = NULL;
1578 	struct socket *mp_so;
1579 
1580 	level = sopt->sopt_level;
1581 	optname = sopt->sopt_name;
1582 
1583 	mp_so = mptetoso(mpte);
1584 
1585 	VERIFY(!(mpsotomppcb(mp_so)->mpp_flags & MPP_INSIDE_SETGETOPT));
1586 	mpsotomppcb(mp_so)->mpp_flags |= MPP_INSIDE_SETGETOPT;
1587 
1588 	/*
1589 	 * Record socket options which are applicable to subflow sockets so
1590 	 * that we can replay them for new ones; see mptcp_usr_socheckopt()
1591 	 * for the list of eligible socket-level options.
1592 	 */
1593 	if (level == SOL_SOCKET) {
1594 		switch (optname) {
1595 		case SO_DEBUG:
1596 		case SO_KEEPALIVE:
1597 		case SO_USELOOPBACK:
1598 		case SO_RANDOMPORT:
1599 		case SO_TRAFFIC_CLASS:
1600 		case SO_RECV_TRAFFIC_CLASS:
1601 		case SO_PRIVILEGED_TRAFFIC_CLASS:
1602 		case SO_RECV_ANYIF:
1603 		case SO_RESTRICTIONS:
1604 		case SO_NOWAKEFROMSLEEP:
1605 		case SO_NOAPNFALLBK:
1606 		case SO_MARK_CELLFALLBACK:
1607 		case SO_MARK_KNOWN_TRACKER:
1608 		case SO_MARK_KNOWN_TRACKER_NON_APP_INITIATED:
1609 		case SO_MARK_APPROVED_APP_DOMAIN:
1610 		case SO_FALLBACK_MODE:
1611 			/* record it */
1612 			break;
1613 		case SO_FLUSH:
1614 			/* don't record it */
1615 			rec = 0;
1616 			break;
1617 
1618 		/* Next ones, record at MPTCP-level */
1619 		case SO_DELEGATED:
1620 			error = sooptcopyin(sopt, &mpte->mpte_epid,
1621 			    sizeof(int), sizeof(int));
1622 			if (error != 0) {
1623 				goto err_out;
1624 			}
1625 
1626 			goto out;
1627 		case SO_DELEGATED_UUID:
1628 			error = sooptcopyin(sopt, &mpte->mpte_euuid,
1629 			    sizeof(uuid_t), sizeof(uuid_t));
1630 			if (error != 0) {
1631 				goto err_out;
1632 			}
1633 
1634 			goto out;
1635 #if NECP
1636 		case SO_NECP_CLIENTUUID:
1637 			if (!uuid_is_null(mpsotomppcb(mp_so)->necp_client_uuid)) {
1638 				error = EINVAL;
1639 				goto err_out;
1640 			}
1641 
1642 			error = sooptcopyin(sopt, &mpsotomppcb(mp_so)->necp_client_uuid,
1643 			    sizeof(uuid_t), sizeof(uuid_t));
1644 			if (error != 0) {
1645 				goto err_out;
1646 			}
1647 
1648 			mpsotomppcb(mp_so)->necp_cb = mptcp_session_necp_cb;
1649 			error = necp_client_register_multipath_cb(mp_so->last_pid,
1650 			    mpsotomppcb(mp_so)->necp_client_uuid,
1651 			    mpsotomppcb(mp_so));
1652 			if (error) {
1653 				goto err_out;
1654 			}
1655 
1656 			if (uuid_is_null(mpsotomppcb(mp_so)->necp_client_uuid)) {
1657 				error = EINVAL;
1658 				goto err_out;
1659 			}
1660 
1661 			goto out;
1662 		case SO_NECP_ATTRIBUTES:
1663 			error = necp_set_socket_attributes(&mpsotomppcb(mp_so)->inp_necp_attributes, sopt);
1664 			if (error) {
1665 				goto err_out;
1666 			}
1667 
1668 			goto out;
1669 #endif /* NECP */
1670 		default:
1671 			/* nothing to do; just return */
1672 			goto out;
1673 		}
1674 	} else {
1675 		switch (optname) {
1676 		case TCP_NODELAY:
1677 		case TCP_RXT_FINDROP:
1678 		case TCP_KEEPALIVE:
1679 		case TCP_KEEPINTVL:
1680 		case TCP_KEEPCNT:
1681 		case TCP_CONNECTIONTIMEOUT:
1682 		case TCP_RXT_CONNDROPTIME:
1683 		case PERSIST_TIMEOUT:
1684 		case TCP_ADAPTIVE_READ_TIMEOUT:
1685 		case TCP_ADAPTIVE_WRITE_TIMEOUT:
1686 		case TCP_FASTOPEN_FORCE_ENABLE:
1687 			/* eligible; record it */
1688 			break;
1689 		case TCP_NOTSENT_LOWAT:
1690 			/* record at MPTCP level */
1691 			error = sooptcopyin(sopt, &optval, sizeof(optval),
1692 			    sizeof(optval));
1693 			if (error) {
1694 				goto err_out;
1695 			}
1696 			if (optval < 0) {
1697 				error = EINVAL;
1698 				goto err_out;
1699 			} else {
1700 				if (optval == 0) {
1701 					mp_so->so_flags &= ~SOF_NOTSENT_LOWAT;
1702 					error = mptcp_set_notsent_lowat(mpte, 0);
1703 				} else {
1704 					mp_so->so_flags |= SOF_NOTSENT_LOWAT;
1705 					error = mptcp_set_notsent_lowat(mpte,
1706 					    optval);
1707 				}
1708 
1709 				if (error) {
1710 					goto err_out;
1711 				}
1712 			}
1713 			goto out;
1714 		case MPTCP_SERVICE_TYPE:
1715 			/* record at MPTCP level */
1716 			error = sooptcopyin(sopt, &optval, sizeof(optval),
1717 			    sizeof(optval));
1718 			if (error) {
1719 				goto err_out;
1720 			}
1721 			if (optval < 0 || optval >= MPTCP_SVCTYPE_MAX) {
1722 				error = EINVAL;
1723 				goto err_out;
1724 			}
1725 
1726 			if (mptcp_entitlement_check(mp_so, (uint8_t)optval) < 0) {
1727 				error = EACCES;
1728 				goto err_out;
1729 			}
1730 
1731 			mpte->mpte_svctype = (uint8_t)optval;
1732 			mpte->mpte_flags |= MPTE_SVCTYPE_CHECKED;
1733 
1734 			goto out;
1735 		case MPTCP_ALTERNATE_PORT:
1736 			/* record at MPTCP level */
1737 			error = sooptcopyin(sopt, &optval, sizeof(optval),
1738 			    sizeof(optval));
1739 			if (error) {
1740 				goto err_out;
1741 			}
1742 
1743 			if (optval < 0 || optval > UINT16_MAX) {
1744 				error = EINVAL;
1745 				goto err_out;
1746 			}
1747 
1748 			mpte->mpte_alternate_port = (uint16_t)optval;
1749 
1750 			goto out;
1751 		case MPTCP_FORCE_ENABLE:
1752 			/* record at MPTCP level */
1753 			error = sooptcopyin(sopt, &optval, sizeof(optval),
1754 			    sizeof(optval));
1755 			if (error) {
1756 				goto err_out;
1757 			}
1758 
1759 			if (optval < 0 || optval > 1) {
1760 				error = EINVAL;
1761 				goto err_out;
1762 			}
1763 
1764 			if (optval) {
1765 				mpte->mpte_flags |= MPTE_FORCE_ENABLE;
1766 			} else {
1767 				mpte->mpte_flags &= ~MPTE_FORCE_ENABLE;
1768 			}
1769 
1770 			goto out;
1771 		case MPTCP_FORCE_VERSION:
1772 			error = sooptcopyin(sopt, &optval, sizeof(optval),
1773 			    sizeof(optval));
1774 			if (error) {
1775 				goto err_out;
1776 			}
1777 
1778 			if (optval != 0 && optval != 1) {
1779 				error = EINVAL;
1780 				goto err_out;
1781 			}
1782 
1783 			if (optval == 0) {
1784 				mpte->mpte_flags |= MPTE_FORCE_V0;
1785 				mpte->mpte_flags &= ~MPTE_FORCE_V1;
1786 			} else {
1787 				mpte->mpte_flags |= MPTE_FORCE_V1;
1788 				mpte->mpte_flags &= ~MPTE_FORCE_V0;
1789 			}
1790 
1791 			goto out;
1792 		case MPTCP_EXPECTED_PROGRESS_TARGET:
1793 		{
1794 			struct mptcb *mp_tp = mpte->mpte_mptcb;
1795 			uint64_t mach_time_target;
1796 			uint64_t nanoseconds;
1797 
1798 			if (mpte->mpte_svctype != MPTCP_SVCTYPE_TARGET_BASED) {
1799 				os_log(mptcp_log_handle, "%s - %lx: Can't set urgent activity when svctype is %u\n",
1800 				    __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte), mpte->mpte_svctype);
1801 				error = EINVAL;
1802 				goto err_out;
1803 			}
1804 
1805 			error = sooptcopyin(sopt, &mach_time_target, sizeof(mach_time_target), sizeof(mach_time_target));
1806 			if (error) {
1807 				goto err_out;
1808 			}
1809 
1810 			if (!mptcp_ok_to_create_subflows(mp_tp)) {
1811 				os_log(mptcp_log_handle, "%s - %lx: Not ok to create subflows, state %u flags %#x\n",
1812 				    __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte), mp_tp->mpt_state, mp_tp->mpt_flags);
1813 				error = EINVAL;
1814 				goto err_out;
1815 			}
1816 
1817 			if (mach_time_target) {
1818 				uint64_t time_now = 0;
1819 				uint64_t time_now_nanoseconds;
1820 
1821 				absolutetime_to_nanoseconds(mach_time_target, &nanoseconds);
1822 				nanoseconds = nanoseconds - (mptcp_expected_progress_headstart * NSEC_PER_MSEC);
1823 
1824 				time_now = mach_continuous_time();
1825 				absolutetime_to_nanoseconds(time_now, &time_now_nanoseconds);
1826 
1827 				nanoseconds_to_absolutetime(nanoseconds, &mach_time_target);
1828 				/* If the timer is already running and it would
1829 				 * fire in less than mptcp_expected_progress_headstart
1830 				 * seconds, then it's not worth canceling it.
1831 				 */
1832 				if (mpte->mpte_time_target &&
1833 				    mpte->mpte_time_target < time_now &&
1834 				    time_now_nanoseconds > nanoseconds - (mptcp_expected_progress_headstart * NSEC_PER_MSEC)) {
1835 					os_log(mptcp_log_handle, "%s - %lx: Not rescheduling timer %llu now %llu target %llu\n",
1836 					    __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte),
1837 					    mpte->mpte_time_target,
1838 					    time_now,
1839 					    mach_time_target);
1840 					goto out;
1841 				}
1842 			}
1843 
1844 			mpte->mpte_time_target = mach_time_target;
1845 			mptcp_set_urgency_timer(mpte);
1846 
1847 			goto out;
1848 		}
1849 		default:
1850 			/* not eligible */
1851 			error = ENOPROTOOPT;
1852 			goto err_out;
1853 		}
1854 	}
1855 
1856 	if ((error = sooptcopyin(sopt, &optval, sizeof(optval),
1857 	    sizeof(optval))) != 0) {
1858 		goto err_out;
1859 	}
1860 
1861 	if (rec) {
1862 		/* search for an existing one; if not found, allocate */
1863 		if ((mpo = mptcp_sopt_find(mpte, sopt)) == NULL) {
1864 			mpo = mptcp_sopt_alloc(Z_WAITOK);
1865 		}
1866 
1867 		if (mpo == NULL) {
1868 			error = ENOBUFS;
1869 			goto err_out;
1870 		} else {
1871 			/* initialize or update, as needed */
1872 			mpo->mpo_intval = optval;
1873 			if (!(mpo->mpo_flags & MPOF_ATTACHED)) {
1874 				mpo->mpo_level = level;
1875 				mpo->mpo_name = optname;
1876 				mptcp_sopt_insert(mpte, mpo);
1877 			}
1878 			/* this can be issued on the subflow socket */
1879 			mpo->mpo_flags |= MPOF_SUBFLOW_OK;
1880 		}
1881 	} else {
1882 		bzero(&smpo, sizeof(smpo));
1883 		mpo = &smpo;
1884 		mpo->mpo_flags |= MPOF_SUBFLOW_OK;
1885 		mpo->mpo_level = level;
1886 		mpo->mpo_name = optname;
1887 		mpo->mpo_intval = optval;
1888 	}
1889 
1890 	/* issue this socket option on existing subflows */
1891 	error = mptcp_setopt_apply(mpte, mpo);
1892 	if (error != 0 && (mpo->mpo_flags & MPOF_ATTACHED)) {
1893 		VERIFY(mpo != &smpo);
1894 		mptcp_sopt_remove(mpte, mpo);
1895 		mptcp_sopt_free(mpo);
1896 	}
1897 	if (mpo == &smpo) {
1898 		mpo->mpo_flags &= ~MPOF_INTERIM;
1899 	}
1900 
1901 	if (error) {
1902 		goto err_out;
1903 	}
1904 
1905 out:
1906 
1907 	mpsotomppcb(mp_so)->mpp_flags &= ~MPP_INSIDE_SETGETOPT;
1908 	return 0;
1909 
1910 err_out:
1911 	os_log_error(mptcp_log_handle, "%s - %lx: sopt %s (%d, %d) val %d can't be issued error %d\n",
1912 	    __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte),
1913 	    mptcp_sopt2str(level, optname), level, optname, optval, error);
1914 	mpsotomppcb(mp_so)->mpp_flags &= ~MPP_INSIDE_SETGETOPT;
1915 	return error;
1916 }
1917 
1918 static void
mptcp_fill_info_bytestats(struct tcp_info * ti,struct mptses * mpte)1919 mptcp_fill_info_bytestats(struct tcp_info *ti, struct mptses *mpte)
1920 {
1921 	struct mptsub *mpts;
1922 	int i;
1923 
1924 	TAILQ_FOREACH(mpts, &mpte->mpte_subflows, mpts_entry) {
1925 		const struct inpcb *inp = sotoinpcb(mpts->mpts_socket);
1926 
1927 		if (inp == NULL) {
1928 			continue;
1929 		}
1930 
1931 		ti->tcpi_txbytes += inp->inp_stat->txbytes;
1932 		ti->tcpi_rxbytes += inp->inp_stat->rxbytes;
1933 		ti->tcpi_cell_txbytes += inp->inp_cstat->txbytes;
1934 		ti->tcpi_cell_rxbytes += inp->inp_cstat->rxbytes;
1935 		ti->tcpi_wifi_txbytes += inp->inp_wstat->txbytes;
1936 		ti->tcpi_wifi_rxbytes += inp->inp_wstat->rxbytes;
1937 		ti->tcpi_wired_txbytes += inp->inp_Wstat->txbytes;
1938 		ti->tcpi_wired_rxbytes += inp->inp_Wstat->rxbytes;
1939 	}
1940 
1941 	for (i = 0; i < MPTCP_ITFSTATS_SIZE; i++) {
1942 		struct mptcp_itf_stats *stats = &mpte->mpte_itfstats[i];
1943 
1944 		ti->tcpi_txbytes += stats->mpis_txbytes;
1945 		ti->tcpi_rxbytes += stats->mpis_rxbytes;
1946 
1947 		ti->tcpi_wifi_txbytes += stats->mpis_wifi_txbytes;
1948 		ti->tcpi_wifi_rxbytes += stats->mpis_wifi_rxbytes;
1949 
1950 		ti->tcpi_wired_txbytes += stats->mpis_wired_txbytes;
1951 		ti->tcpi_wired_rxbytes += stats->mpis_wired_rxbytes;
1952 
1953 		ti->tcpi_cell_txbytes += stats->mpis_cell_txbytes;
1954 		ti->tcpi_cell_rxbytes += stats->mpis_cell_rxbytes;
1955 	}
1956 }
1957 
1958 static void
mptcp_fill_info(struct mptses * mpte,struct tcp_info * ti)1959 mptcp_fill_info(struct mptses *mpte, struct tcp_info *ti)
1960 {
1961 	struct mptsub *actsub = mpte->mpte_active_sub;
1962 	struct mptcb *mp_tp = mpte->mpte_mptcb;
1963 	struct tcpcb *acttp = NULL;
1964 
1965 	if (actsub) {
1966 		acttp = sototcpcb(actsub->mpts_socket);
1967 	}
1968 
1969 	bzero(ti, sizeof(*ti));
1970 
1971 	ti->tcpi_state = (uint8_t)mp_tp->mpt_state;
1972 	/* tcpi_options */
1973 	/* tcpi_snd_wscale */
1974 	/* tcpi_rcv_wscale */
1975 	/* tcpi_flags */
1976 	if (acttp) {
1977 		ti->tcpi_rto = acttp->t_timer[TCPT_REXMT] ? acttp->t_rxtcur : 0;
1978 	}
1979 
1980 	/* tcpi_snd_mss */
1981 	/* tcpi_rcv_mss */
1982 	if (acttp) {
1983 		ti->tcpi_rttcur = acttp->t_rttcur;
1984 		ti->tcpi_srtt = acttp->t_srtt >> TCP_RTT_SHIFT;
1985 		ti->tcpi_rttvar = acttp->t_rttvar >> TCP_RTTVAR_SHIFT;
1986 		ti->tcpi_rttbest = acttp->t_rttbest >> TCP_RTT_SHIFT;
1987 		ti->tcpi_rcv_srtt = acttp->rcv_srtt >> TCP_RTT_SHIFT;
1988 	}
1989 	/* tcpi_snd_ssthresh */
1990 	/* tcpi_snd_cwnd */
1991 	/* tcpi_rcv_space */
1992 	ti->tcpi_snd_wnd = mp_tp->mpt_sndwnd;
1993 	ti->tcpi_snd_nxt = (uint32_t)mp_tp->mpt_sndnxt;
1994 	ti->tcpi_rcv_nxt = (uint32_t)mp_tp->mpt_rcvnxt;
1995 	if (acttp) {
1996 		ti->tcpi_last_outif = (acttp->t_inpcb->inp_last_outifp == NULL) ? 0 :
1997 		    acttp->t_inpcb->inp_last_outifp->if_index;
1998 	}
1999 
2000 	mptcp_fill_info_bytestats(ti, mpte);
2001 	/* tcpi_txpackets */
2002 
2003 	/* tcpi_txretransmitbytes */
2004 	/* tcpi_txunacked */
2005 	/* tcpi_rxpackets */
2006 
2007 	/* tcpi_rxduplicatebytes */
2008 	/* tcpi_rxoutoforderbytes */
2009 	/* tcpi_snd_bw */
2010 	/* tcpi_synrexmits */
2011 	/* tcpi_unused1 */
2012 	/* tcpi_unused2 */
2013 	/* tcpi_cell_rxpackets */
2014 
2015 	/* tcpi_cell_txpackets */
2016 
2017 	/* tcpi_wifi_rxpackets */
2018 
2019 	/* tcpi_wifi_txpackets */
2020 
2021 	/* tcpi_wired_rxpackets */
2022 	/* tcpi_wired_txpackets */
2023 	/* tcpi_connstatus */
2024 	/* TFO-stuff */
2025 	/* ECN stuff */
2026 	/* tcpi_ecn_recv_ce */
2027 	/* tcpi_ecn_recv_cwr */
2028 	if (acttp) {
2029 		ti->tcpi_rcvoopack = acttp->t_rcvoopack;
2030 	}
2031 	/* tcpi_pawsdrop */
2032 	/* tcpi_sack_recovery_episode */
2033 	/* tcpi_reordered_pkts */
2034 	/* tcpi_dsack_sent */
2035 	/* tcpi_dsack_recvd */
2036 	/* tcpi_flowhash */
2037 	if (acttp) {
2038 		ti->tcpi_txretransmitpackets = acttp->t_stat.rxmitpkts;
2039 	}
2040 }
2041 
2042 /*
2043  * Handle SOPT_GET for socket options issued on MP socket.
2044  */
2045 static int
mptcp_getopt(struct mptses * mpte,struct sockopt * sopt)2046 mptcp_getopt(struct mptses *mpte, struct sockopt *sopt)
2047 {
2048 	int error = 0, optval = 0;
2049 	struct socket *mp_so;
2050 
2051 	mp_so = mptetoso(mpte);
2052 
2053 	VERIFY(!(mpsotomppcb(mp_so)->mpp_flags & MPP_INSIDE_SETGETOPT));
2054 	mpsotomppcb(mp_so)->mpp_flags |= MPP_INSIDE_SETGETOPT;
2055 
2056 	/*
2057 	 * We only handle SOPT_GET for TCP level socket options; we should
2058 	 * not get here for socket level options since they are already
2059 	 * handled at the socket layer.
2060 	 */
2061 	if (sopt->sopt_level != IPPROTO_TCP) {
2062 		error = ENOPROTOOPT;
2063 		goto out;
2064 	}
2065 
2066 	switch (sopt->sopt_name) {
2067 	case PERSIST_TIMEOUT:
2068 		/* Only case for which we have a non-zero default */
2069 		optval = tcp_max_persist_timeout;
2070 		OS_FALLTHROUGH;
2071 	case TCP_NODELAY:
2072 	case TCP_RXT_FINDROP:
2073 	case TCP_KEEPALIVE:
2074 	case TCP_KEEPINTVL:
2075 	case TCP_KEEPCNT:
2076 	case TCP_CONNECTIONTIMEOUT:
2077 	case TCP_RXT_CONNDROPTIME:
2078 	case TCP_ADAPTIVE_READ_TIMEOUT:
2079 	case TCP_ADAPTIVE_WRITE_TIMEOUT:
2080 	case TCP_FASTOPEN_FORCE_ENABLE:
2081 	{
2082 		struct mptopt *mpo = mptcp_sopt_find(mpte, sopt);
2083 
2084 		if (mpo != NULL) {
2085 			optval = mpo->mpo_intval;
2086 		}
2087 		break;
2088 	}
2089 
2090 	/* The next ones are stored at the MPTCP-level */
2091 	case TCP_NOTSENT_LOWAT:
2092 		if (mptetoso(mpte)->so_flags & SOF_NOTSENT_LOWAT) {
2093 			optval = mptcp_get_notsent_lowat(mpte);
2094 		} else {
2095 			optval = 0;
2096 		}
2097 		break;
2098 	case TCP_INFO:
2099 	{
2100 		struct tcp_info ti;
2101 
2102 		mptcp_fill_info(mpte, &ti);
2103 		error = sooptcopyout(sopt, &ti, sizeof(struct tcp_info));
2104 
2105 		goto out;
2106 	}
2107 	case MPTCP_SERVICE_TYPE:
2108 		optval = mpte->mpte_svctype;
2109 		break;
2110 	case MPTCP_ALTERNATE_PORT:
2111 		optval = mpte->mpte_alternate_port;
2112 		break;
2113 	case MPTCP_FORCE_ENABLE:
2114 		optval = !!(mpte->mpte_flags & MPTE_FORCE_ENABLE);
2115 		break;
2116 	case MPTCP_FORCE_VERSION:
2117 		if (mpte->mpte_flags & MPTE_FORCE_V0) {
2118 			optval = 0;
2119 		} else if (mpte->mpte_flags & MPTE_FORCE_V1) {
2120 			optval = 1;
2121 		} else {
2122 			optval = -1;
2123 		}
2124 		break;
2125 	case MPTCP_EXPECTED_PROGRESS_TARGET:
2126 		error = sooptcopyout(sopt, &mpte->mpte_time_target, sizeof(mpte->mpte_time_target));
2127 
2128 		goto out;
2129 	default:
2130 		/* not eligible */
2131 		error = ENOPROTOOPT;
2132 		break;
2133 	}
2134 
2135 	if (error == 0) {
2136 		error = sooptcopyout(sopt, &optval, sizeof(int));
2137 	}
2138 
2139 out:
2140 	mpsotomppcb(mp_so)->mpp_flags &= ~MPP_INSIDE_SETGETOPT;
2141 	return error;
2142 }
2143 
2144 /*
2145  * MPTCP SOPT_{SET,GET} socket option handler, for options issued on the MP
2146  * socket, at SOL_SOCKET and IPPROTO_TCP levels.  The former is restricted
2147  * to those that are allowed by mptcp_usr_socheckopt().
2148  */
2149 int
mptcp_ctloutput(struct socket * mp_so,struct sockopt * sopt)2150 mptcp_ctloutput(struct socket *mp_so, struct sockopt *sopt)
2151 {
2152 	struct mppcb *mpp = mpsotomppcb(mp_so);
2153 	struct mptses *mpte;
2154 	int error = 0;
2155 
2156 	if (mpp == NULL || mpp->mpp_state == MPPCB_STATE_DEAD) {
2157 		error = EINVAL;
2158 		goto out;
2159 	}
2160 	mpte = mptompte(mpp);
2161 	socket_lock_assert_owned(mp_so);
2162 
2163 	/* we only handle socket and TCP-level socket options for MPTCP */
2164 	if (sopt->sopt_level != SOL_SOCKET && sopt->sopt_level != IPPROTO_TCP) {
2165 		error = EINVAL;
2166 		goto out;
2167 	}
2168 
2169 	switch (sopt->sopt_dir) {
2170 	case SOPT_SET:
2171 		error = mptcp_setopt(mpte, sopt);
2172 		break;
2173 
2174 	case SOPT_GET:
2175 		error = mptcp_getopt(mpte, sopt);
2176 		break;
2177 	}
2178 out:
2179 	return error;
2180 }
2181 
2182 const char *
mptcp_sopt2str(int level,int optname)2183 mptcp_sopt2str(int level, int optname)
2184 {
2185 	switch (level) {
2186 	case SOL_SOCKET:
2187 		switch (optname) {
2188 		case SO_LINGER:
2189 			return "SO_LINGER";
2190 		case SO_LINGER_SEC:
2191 			return "SO_LINGER_SEC";
2192 		case SO_DEBUG:
2193 			return "SO_DEBUG";
2194 		case SO_KEEPALIVE:
2195 			return "SO_KEEPALIVE";
2196 		case SO_USELOOPBACK:
2197 			return "SO_USELOOPBACK";
2198 		case SO_TYPE:
2199 			return "SO_TYPE";
2200 		case SO_NREAD:
2201 			return "SO_NREAD";
2202 		case SO_NWRITE:
2203 			return "SO_NWRITE";
2204 		case SO_ERROR:
2205 			return "SO_ERROR";
2206 		case SO_SNDBUF:
2207 			return "SO_SNDBUF";
2208 		case SO_RCVBUF:
2209 			return "SO_RCVBUF";
2210 		case SO_SNDLOWAT:
2211 			return "SO_SNDLOWAT";
2212 		case SO_RCVLOWAT:
2213 			return "SO_RCVLOWAT";
2214 		case SO_SNDTIMEO:
2215 			return "SO_SNDTIMEO";
2216 		case SO_RCVTIMEO:
2217 			return "SO_RCVTIMEO";
2218 		case SO_NKE:
2219 			return "SO_NKE";
2220 		case SO_NOSIGPIPE:
2221 			return "SO_NOSIGPIPE";
2222 		case SO_NOADDRERR:
2223 			return "SO_NOADDRERR";
2224 		case SO_RESTRICTIONS:
2225 			return "SO_RESTRICTIONS";
2226 		case SO_LABEL:
2227 			return "SO_LABEL";
2228 		case SO_PEERLABEL:
2229 			return "SO_PEERLABEL";
2230 		case SO_RANDOMPORT:
2231 			return "SO_RANDOMPORT";
2232 		case SO_TRAFFIC_CLASS:
2233 			return "SO_TRAFFIC_CLASS";
2234 		case SO_RECV_TRAFFIC_CLASS:
2235 			return "SO_RECV_TRAFFIC_CLASS";
2236 		case SO_TRAFFIC_CLASS_DBG:
2237 			return "SO_TRAFFIC_CLASS_DBG";
2238 		case SO_PRIVILEGED_TRAFFIC_CLASS:
2239 			return "SO_PRIVILEGED_TRAFFIC_CLASS";
2240 		case SO_DEFUNCTIT:
2241 			return "SO_DEFUNCTIT";
2242 		case SO_DEFUNCTOK:
2243 			return "SO_DEFUNCTOK";
2244 		case SO_ISDEFUNCT:
2245 			return "SO_ISDEFUNCT";
2246 		case SO_OPPORTUNISTIC:
2247 			return "SO_OPPORTUNISTIC";
2248 		case SO_FLUSH:
2249 			return "SO_FLUSH";
2250 		case SO_RECV_ANYIF:
2251 			return "SO_RECV_ANYIF";
2252 		case SO_NOWAKEFROMSLEEP:
2253 			return "SO_NOWAKEFROMSLEEP";
2254 		case SO_NOAPNFALLBK:
2255 			return "SO_NOAPNFALLBK";
2256 		case SO_MARK_CELLFALLBACK:
2257 			return "SO_CELLFALLBACK";
2258 		case SO_FALLBACK_MODE:
2259 			return "SO_FALLBACK_MODE";
2260 		case SO_MARK_KNOWN_TRACKER:
2261 			return "SO_MARK_KNOWN_TRACKER";
2262 		case SO_MARK_KNOWN_TRACKER_NON_APP_INITIATED:
2263 			return "SO_MARK_KNOWN_TRACKER_NON_APP_INITIATED";
2264 		case SO_MARK_APPROVED_APP_DOMAIN:
2265 			return "SO_MARK_APPROVED_APP_DOMAIN";
2266 		case SO_DELEGATED:
2267 			return "SO_DELEGATED";
2268 		case SO_DELEGATED_UUID:
2269 			return "SO_DELEGATED_UUID";
2270 #if NECP
2271 		case SO_NECP_ATTRIBUTES:
2272 			return "SO_NECP_ATTRIBUTES";
2273 		case SO_NECP_CLIENTUUID:
2274 			return "SO_NECP_CLIENTUUID";
2275 #endif /* NECP */
2276 		}
2277 
2278 		break;
2279 	case IPPROTO_TCP:
2280 		switch (optname) {
2281 		case TCP_NODELAY:
2282 			return "TCP_NODELAY";
2283 		case TCP_KEEPALIVE:
2284 			return "TCP_KEEPALIVE";
2285 		case TCP_KEEPINTVL:
2286 			return "TCP_KEEPINTVL";
2287 		case TCP_KEEPCNT:
2288 			return "TCP_KEEPCNT";
2289 		case TCP_CONNECTIONTIMEOUT:
2290 			return "TCP_CONNECTIONTIMEOUT";
2291 		case TCP_RXT_CONNDROPTIME:
2292 			return "TCP_RXT_CONNDROPTIME";
2293 		case PERSIST_TIMEOUT:
2294 			return "PERSIST_TIMEOUT";
2295 		case TCP_NOTSENT_LOWAT:
2296 			return "NOTSENT_LOWAT";
2297 		case TCP_ADAPTIVE_READ_TIMEOUT:
2298 			return "ADAPTIVE_READ_TIMEOUT";
2299 		case TCP_ADAPTIVE_WRITE_TIMEOUT:
2300 			return "ADAPTIVE_WRITE_TIMEOUT";
2301 		case TCP_FASTOPEN_FORCE_ENABLE:
2302 			return "TCP_FASTOPEN_FORCE_ENABLE";
2303 		case MPTCP_SERVICE_TYPE:
2304 			return "MPTCP_SERVICE_TYPE";
2305 		case MPTCP_ALTERNATE_PORT:
2306 			return "MPTCP_ALTERNATE_PORT";
2307 		case MPTCP_FORCE_ENABLE:
2308 			return "MPTCP_FORCE_ENABLE";
2309 		case MPTCP_FORCE_VERSION:
2310 			return "MPTCP_FORCE_VERSION";
2311 		case MPTCP_EXPECTED_PROGRESS_TARGET:
2312 			return "MPTCP_EXPECTED_PROGRESS_TARGET";
2313 		}
2314 
2315 		break;
2316 	}
2317 
2318 	return "unknown";
2319 }
2320 
2321 static int
mptcp_usr_preconnect(struct socket * mp_so)2322 mptcp_usr_preconnect(struct socket *mp_so)
2323 {
2324 	struct mptsub *mpts = NULL;
2325 	struct mppcb *mpp = mpsotomppcb(mp_so);
2326 	struct mptses *mpte;
2327 	struct socket *so;
2328 	struct tcpcb *tp = NULL;
2329 	int error;
2330 
2331 	mpte = mptompte(mpp);
2332 
2333 	mpts = mptcp_get_subflow(mpte, NULL);
2334 	if (mpts == NULL) {
2335 		os_log_error(mptcp_log_handle, "%s - %lx: invalid preconnect ",
2336 		    __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte));
2337 		return EINVAL;
2338 	}
2339 	mpts->mpts_flags &= ~MPTSF_TFO_REQD;
2340 	so = mpts->mpts_socket;
2341 	tp = intotcpcb(sotoinpcb(so));
2342 	tp->t_mpflags &= ~TMPF_TFO_REQUEST;
2343 	error = tcp_output(sototcpcb(so));
2344 
2345 	soclearfastopen(mp_so);
2346 
2347 	return error;
2348 }
2349