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 #include <net/sockaddr_utils.h>
57
58 static int mptcp_usr_attach(struct socket *, int, struct proc *);
59 static int mptcp_usr_detach(struct socket *);
60 static int mptcp_attach(struct socket *, struct proc *);
61 static int mptcp_usr_connectx(struct socket *, struct sockaddr *,
62 struct sockaddr *, struct proc *, uint32_t, sae_associd_t,
63 sae_connid_t *, uint32_t, void *, uint32_t, struct uio *, user_ssize_t *);
64 static int mptcp_getassocids(struct mptses *, uint32_t *, user_addr_t);
65 static int mptcp_getconnids(struct mptses *, sae_associd_t, uint32_t *,
66 user_addr_t);
67 static int mptcp_getconninfo(struct mptses *, sae_connid_t *, uint32_t *,
68 uint32_t *, int32_t *, user_addr_t, socklen_t *, user_addr_t, socklen_t *,
69 uint32_t *, user_addr_t, uint32_t *);
70 static int mptcp_usr_control(struct socket *, u_long cmd,
71 caddr_t __sized_by(IOCPARM_LEN(cmd)), struct ifnet *,
72 struct proc *);
73 static int mptcp_disconnect(struct mptses *);
74 static int mptcp_usr_disconnect(struct socket *);
75 static int mptcp_usr_disconnectx(struct socket *, sae_associd_t, sae_connid_t);
76 static struct mptses *mptcp_usrclosed(struct mptses *);
77 static int mptcp_usr_rcvd(struct socket *, int);
78 static int mptcp_usr_send(struct socket *, int, struct mbuf *,
79 struct sockaddr *, struct mbuf *, struct proc *);
80 static int mptcp_usr_shutdown(struct socket *);
81 static int mptcp_usr_sosend(struct socket *, struct sockaddr *, struct uio *,
82 struct mbuf *, struct mbuf *, int);
83 static int mptcp_usr_socheckopt(struct socket *, struct sockopt *);
84 static int mptcp_usr_preconnect(struct socket *so);
85
86 struct pr_usrreqs mptcp_usrreqs = {
87 .pru_attach = mptcp_usr_attach,
88 .pru_connectx = mptcp_usr_connectx,
89 .pru_control = mptcp_usr_control,
90 .pru_detach = mptcp_usr_detach,
91 .pru_disconnect = mptcp_usr_disconnect,
92 .pru_disconnectx = mptcp_usr_disconnectx,
93 .pru_peeraddr = mp_getpeeraddr,
94 .pru_rcvd = mptcp_usr_rcvd,
95 .pru_send = mptcp_usr_send,
96 .pru_shutdown = mptcp_usr_shutdown,
97 .pru_sockaddr = mp_getsockaddr,
98 .pru_sosend = mptcp_usr_sosend,
99 .pru_soreceive = soreceive,
100 .pru_socheckopt = mptcp_usr_socheckopt,
101 .pru_preconnect = mptcp_usr_preconnect,
102 };
103
104
105 int mptcp_developer_mode = 0;
106 SYSCTL_INT(_net_inet_mptcp, OID_AUTO, allow_aggregate, CTLFLAG_RW | CTLFLAG_LOCKED,
107 &mptcp_developer_mode, 0, "Allow the Multipath aggregation mode");
108
109 int mptcp_no_first_party = 0;
110 SYSCTL_INT(_net_inet_mptcp, OID_AUTO, no_first_party, CTLFLAG_RW | CTLFLAG_LOCKED,
111 &mptcp_no_first_party, 0, "Do not do first-party app exemptions");
112
113 static unsigned long mptcp_expected_progress_headstart = 5000;
114 SYSCTL_ULONG(_net_inet_mptcp, OID_AUTO, expected_progress_headstart, CTLFLAG_RW | CTLFLAG_LOCKED,
115 &mptcp_expected_progress_headstart, "Headstart to give MPTCP before meeting the progress deadline");
116
117
118 /*
119 * Attaches an MPTCP control block to a socket.
120 */
121 static int
mptcp_usr_attach(struct socket * mp_so,int proto,struct proc * p)122 mptcp_usr_attach(struct socket *mp_so, int proto, struct proc *p)
123 {
124 #pragma unused(proto)
125 int error;
126
127 VERIFY(mpsotomppcb(mp_so) == NULL);
128
129 error = mptcp_attach(mp_so, p);
130 if (error) {
131 goto out;
132 }
133
134 if ((mp_so->so_options & SO_LINGER) && mp_so->so_linger == 0) {
135 mp_so->so_linger = (short)(TCP_LINGERTIME * hz);
136 }
137 out:
138 return error;
139 }
140
141 /*
142 * Detaches an MPTCP control block from a socket.
143 */
144 static int
mptcp_usr_detach(struct socket * mp_so)145 mptcp_usr_detach(struct socket *mp_so)
146 {
147 struct mptses *__single mpte = mpsotompte(mp_so);
148 struct mppcb *__single mpp = mpsotomppcb(mp_so);
149
150 if (mpp == NULL || mpp->mpp_state == MPPCB_STATE_DEAD) {
151 os_log_error(mptcp_log_handle, "%s - %lx: state: %d\n",
152 __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte),
153 mpp ? mpp->mpp_state : -1);
154 return EINVAL;
155 }
156
157 /*
158 * We are done with this MPTCP socket (it has been closed);
159 * trigger all subflows to be disconnected, if not already,
160 * by initiating the PCB detach sequence (SOF_PCBCLEARING
161 * will be set.)
162 */
163 mp_pcbdetach(mp_so);
164
165 mptcp_disconnect(mpte);
166
167 return 0;
168 }
169
170 /*
171 * Attach MPTCP protocol to socket, allocating MP control block,
172 * MPTCP session, control block, buffer space, etc.
173 */
174 static int
mptcp_attach(struct socket * mp_so,struct proc * p)175 mptcp_attach(struct socket *mp_so, struct proc *p)
176 {
177 #pragma unused(p)
178 struct mptses *__single mpte = NULL;
179 struct mptcb *__single mp_tp = NULL;
180 struct mppcb *__single mpp = NULL;
181 int error = 0;
182
183 if (mp_so->so_snd.sb_hiwat == 0 || mp_so->so_rcv.sb_hiwat == 0) {
184 error = soreserve(mp_so, tcp_sendspace, tcp_recvspace);
185 if (error != 0) {
186 goto out;
187 }
188 }
189
190 if (mp_so->so_snd.sb_preconn_hiwat == 0) {
191 soreserve_preconnect(mp_so, 2048);
192 }
193
194 if ((mp_so->so_rcv.sb_flags & SB_USRSIZE) == 0) {
195 mp_so->so_rcv.sb_flags |= SB_AUTOSIZE;
196 }
197 if ((mp_so->so_snd.sb_flags & SB_USRSIZE) == 0) {
198 mp_so->so_snd.sb_flags |= SB_AUTOSIZE;
199 }
200
201 /*
202 * MPTCP send-socket buffers cannot be compressed, due to the
203 * fact that each mbuf chained via m_next is a M_PKTHDR
204 * which carries some MPTCP metadata.
205 */
206 mp_so->so_snd.sb_flags |= SB_NOCOMPRESS;
207
208 if ((error = mp_pcballoc(mp_so, &mtcbinfo)) != 0) {
209 goto out;
210 }
211
212 mpp = mpsotomppcb(mp_so);
213 mpte = (struct mptses *)mpp->mpp_pcbe;
214 mp_tp = mpte->mpte_mptcb;
215
216 VERIFY(mp_tp != NULL);
217 out:
218 return error;
219 }
220
221 static int
mptcp_entitlement_check(struct socket * mp_so,uint8_t svctype)222 mptcp_entitlement_check(struct socket *mp_so, uint8_t svctype)
223 {
224 struct mptses *mpte = mpsotompte(mp_so);
225
226 if (mptcp_no_first_party) {
227 return 0;
228 }
229
230 /* First, check for mptcp_extended without delegation */
231 if (soopt_cred_check(mp_so, PRIV_NET_RESTRICTED_MULTIPATH_EXTENDED, TRUE, FALSE) == 0) {
232 /*
233 * This means the app has the extended entitlement. Thus,
234 * it's a first party app and can run without restrictions.
235 */
236 mpte->mpte_flags |= MPTE_FIRSTPARTY;
237 return 0;
238 }
239
240 /* Now with delegation */
241 if (mp_so->so_flags & SOF_DELEGATED &&
242 soopt_cred_check(mp_so, PRIV_NET_RESTRICTED_MULTIPATH_EXTENDED, TRUE, TRUE) == 0) {
243 /*
244 * This means the app has the extended entitlement. Thus,
245 * it's a first party app and can run without restrictions.
246 */
247 mpte->mpte_flags |= MPTE_FIRSTPARTY;
248 return 0;
249 }
250
251 if (svctype == MPTCP_SVCTYPE_AGGREGATE) {
252 if (mptcp_developer_mode) {
253 return 0;
254 }
255
256 os_log_error(mptcp_log_handle, "%s - %lx: MPTCP prohibited on svc %u\n",
257 __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte), mpte->mpte_svctype);
258 return -1;
259 }
260
261 return 0;
262 }
263
264 /*
265 * Common subroutine to open a MPTCP connection to one of the remote hosts
266 * specified by dst_sl. This includes allocating and establishing a
267 * subflow TCP connection, either initially to establish MPTCP connection,
268 * or to join an existing one. Returns a connection handle upon success.
269 */
270 static int
mptcp_connectx(struct mptses * mpte,struct sockaddr * src,struct sockaddr * dst,uint32_t ifscope,sae_connid_t * pcid)271 mptcp_connectx(struct mptses *mpte, struct sockaddr *src,
272 struct sockaddr *dst, uint32_t ifscope, sae_connid_t *pcid)
273 {
274 int error = 0;
275
276 VERIFY(dst != NULL);
277 VERIFY(pcid != NULL);
278
279 error = mptcp_subflow_add(mpte, src, dst, ifscope, pcid);
280
281 return error;
282 }
283
284 /*
285 * User-protocol pru_connectx callback.
286 */
287 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)288 mptcp_usr_connectx(struct socket *mp_so, struct sockaddr *src,
289 struct sockaddr *dst, struct proc *p, uint32_t ifscope,
290 sae_associd_t aid, sae_connid_t *pcid, uint32_t flags, void *arg,
291 uint32_t arglen, struct uio *auio, user_ssize_t *bytes_written)
292 {
293 #pragma unused(p, aid, flags, arg, arglen)
294 struct mppcb *mpp = mpsotomppcb(mp_so);
295 struct mptses *mpte = NULL;
296 struct mptcb *mp_tp = NULL;
297 user_ssize_t datalen;
298 int error = 0;
299
300 if (mpp == NULL || mpp->mpp_state == MPPCB_STATE_DEAD) {
301 os_log_error(mptcp_log_handle, "%s - %lx: state %d\n",
302 __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte),
303 mpp ? mpp->mpp_state : -1);
304 error = EINVAL;
305 goto out;
306 }
307 mpte = mptompte(mpp);
308 mp_tp = mpte->mpte_mptcb;
309
310 if (mp_tp->mpt_flags & MPTCPF_FALLBACK_TO_TCP) {
311 os_log_error(mptcp_log_handle, "%s - %lx: fell back to TCP\n",
312 __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte));
313 error = EINVAL;
314 goto out;
315 }
316
317 if (dst->sa_family != AF_INET && dst->sa_family != AF_INET6) {
318 error = EAFNOSUPPORT;
319 goto out;
320 }
321
322 if (dst->sa_family == AF_INET &&
323 dst->sa_len != sizeof(mpte->__mpte_dst_v4)) {
324 os_log_error(mptcp_log_handle, "%s - %lx: IPv4 dst len %u\n",
325 __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte), dst->sa_len);
326 error = EINVAL;
327 goto out;
328 }
329
330 if (dst->sa_family == AF_INET6 &&
331 dst->sa_len != sizeof(mpte->__mpte_dst_v6)) {
332 os_log_error(mptcp_log_handle, "%s - %lx: IPv6 dst len %u\n",
333 __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte), dst->sa_len);
334 error = EINVAL;
335 goto out;
336 }
337
338 if (!(mpte->mpte_flags & MPTE_SVCTYPE_CHECKED)) {
339 if (mptcp_entitlement_check(mp_so, mpte->mpte_svctype) < 0) {
340 error = EPERM;
341 goto out;
342 }
343
344 mpte->mpte_flags |= MPTE_SVCTYPE_CHECKED;
345 }
346
347 if ((mp_so->so_state & (SS_ISCONNECTED | SS_ISCONNECTING)) == 0) {
348 SOCKADDR_COPY(dst, &mpte->mpte_dst, dst->sa_len);
349
350 if (dst->sa_family == AF_INET) {
351 SOCKADDR_COPY(dst, &mpte->mpte_sub_dst_v4, dst->sa_len);
352 } else {
353 SOCKADDR_COPY(dst, &mpte->mpte_sub_dst_v6, dst->sa_len);
354 }
355 }
356
357 if (src) {
358 if (src->sa_family != AF_INET && src->sa_family != AF_INET6) {
359 error = EAFNOSUPPORT;
360 goto out;
361 }
362
363 if (src->sa_family == AF_INET &&
364 src->sa_len != sizeof(mpte->__mpte_src_v4)) {
365 os_log_error(mptcp_log_handle, "%s - %lx: IPv4 src len %u\n",
366 __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte), src->sa_len);
367 error = EINVAL;
368 goto out;
369 }
370
371 if (src->sa_family == AF_INET6 &&
372 src->sa_len != sizeof(mpte->__mpte_src_v6)) {
373 os_log_error(mptcp_log_handle, "%s - %lx: IPv6 src len %u\n",
374 __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte), src->sa_len);
375 error = EINVAL;
376 goto out;
377 }
378
379 if ((mp_so->so_state & (SS_ISCONNECTED | SS_ISCONNECTING)) == 0) {
380 SOCKADDR_COPY(src, &mpte->mpte_src, src->sa_len);
381 }
382 }
383
384 error = mptcp_connectx(mpte, src, dst, ifscope, pcid);
385
386 /* If there is data, copy it */
387 if (auio != NULL) {
388 datalen = uio_resid(auio);
389 socket_unlock(mp_so, 0);
390 error = mp_so->so_proto->pr_usrreqs->pru_sosend(mp_so, NULL,
391 (uio_t) auio, NULL, NULL, 0);
392
393 if (error == 0 || error == EWOULDBLOCK) {
394 *bytes_written = datalen - uio_resid(auio);
395 }
396
397 if (error == EWOULDBLOCK) {
398 error = EINPROGRESS;
399 }
400
401 socket_lock(mp_so, 0);
402 }
403
404 out:
405 return error;
406 }
407
408 /*
409 * Handle SIOCGASSOCIDS ioctl for PF_MULTIPATH domain.
410 */
411 static int
mptcp_getassocids(struct mptses * mpte,uint32_t * cnt,user_addr_t aidp)412 mptcp_getassocids(struct mptses *mpte, uint32_t *cnt, user_addr_t aidp)
413 {
414 /* MPTCP has at most 1 association */
415 *cnt = (mpte->mpte_associd != SAE_ASSOCID_ANY) ? 1 : 0;
416
417 /* just asking how many there are? */
418 if (aidp == USER_ADDR_NULL) {
419 return 0;
420 }
421
422 return copyout(&mpte->mpte_associd, aidp,
423 sizeof(mpte->mpte_associd));
424 }
425
426 /*
427 * Handle SIOCGCONNIDS ioctl for PF_MULTIPATH domain.
428 */
429 static int
mptcp_getconnids(struct mptses * mpte,sae_associd_t aid,uint32_t * cnt,user_addr_t cidp)430 mptcp_getconnids(struct mptses *mpte, sae_associd_t aid, uint32_t *cnt,
431 user_addr_t cidp)
432 {
433 struct mptsub *mpts;
434 int error = 0;
435
436 if (aid != SAE_ASSOCID_ANY && aid != SAE_ASSOCID_ALL &&
437 aid != mpte->mpte_associd) {
438 return EINVAL;
439 }
440
441 *cnt = mpte->mpte_numflows;
442
443 /* just asking how many there are? */
444 if (cidp == USER_ADDR_NULL) {
445 return 0;
446 }
447
448 TAILQ_FOREACH(mpts, &mpte->mpte_subflows, mpts_entry) {
449 if ((error = copyout(&mpts->mpts_connid, cidp,
450 sizeof(mpts->mpts_connid))) != 0) {
451 break;
452 }
453
454 cidp += sizeof(mpts->mpts_connid);
455 }
456
457 return error;
458 }
459
460 /*
461 * Handle SIOCGCONNINFO ioctl for PF_MULTIPATH domain.
462 */
463 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)464 mptcp_getconninfo(struct mptses *mpte, sae_connid_t *cid, uint32_t *flags,
465 uint32_t *ifindex, int32_t *soerror, user_addr_t src, socklen_t *src_len,
466 user_addr_t dst, socklen_t *dst_len, uint32_t *aux_type,
467 user_addr_t aux_data, uint32_t *aux_len)
468 {
469 *flags = 0;
470 *aux_type = 0;
471 *ifindex = 0;
472 *soerror = 0;
473 struct mptcb *mp_tp = mpte->mpte_mptcb;
474
475 /* MPTCP-level global stats */
476 if (*cid == SAE_CONNID_ALL) {
477 struct socket *mp_so = mptetoso(mpte);
478 struct conninfo_multipathtcp mptcp_ci;
479 int error = 0;
480
481 if (*aux_len != 0 && *aux_len != sizeof(mptcp_ci)) {
482 return EINVAL;
483 }
484
485 if (mp_so->so_state & SS_ISCONNECTING) {
486 *flags |= CIF_CONNECTING;
487 }
488 if (mp_so->so_state & SS_ISCONNECTED) {
489 *flags |= CIF_CONNECTED;
490 }
491 if (mp_so->so_state & SS_ISDISCONNECTING) {
492 *flags |= CIF_DISCONNECTING;
493 }
494 if (mp_so->so_state & SS_ISDISCONNECTED) {
495 *flags |= CIF_DISCONNECTED;
496 }
497 if (!(mp_tp->mpt_flags & MPTCPF_FALLBACK_TO_TCP)) {
498 *flags |= CIF_MP_CAPABLE;
499 }
500 if (mp_tp->mpt_flags & MPTCPF_FALLBACK_TO_TCP) {
501 *flags |= CIF_MP_DEGRADED;
502 }
503 if (mp_tp->mpt_version == MPTCP_VERSION_1) {
504 *flags |= CIF_MP_V1;
505 }
506
507 *src_len = 0;
508 *dst_len = 0;
509
510 *aux_type = CIAUX_MPTCP;
511 *aux_len = sizeof(mptcp_ci);
512
513 if (aux_data != USER_ADDR_NULL) {
514 const struct mptsub *mpts;
515 int initial_info_set = 0;
516 unsigned long i = 0;
517
518 bzero(&mptcp_ci, sizeof(mptcp_ci));
519 mptcp_ci.mptcpci_subflow_count = mpte->mpte_numflows;
520 mptcp_ci.mptcpci_switch_count = mpte->mpte_subflow_switches;
521
522 VERIFY(sizeof(mptcp_ci.mptcpci_itfstats) == sizeof(mpte->mpte_itfstats));
523 memcpy(mptcp_ci.mptcpci_itfstats, mpte->mpte_itfstats, sizeof(mptcp_ci.mptcpci_itfstats));
524
525 TAILQ_FOREACH(mpts, &mpte->mpte_subflows, mpts_entry) {
526 if (i >= sizeof(mptcp_ci.mptcpci_subflow_connids) / sizeof(sae_connid_t)) {
527 break;
528 }
529 mptcp_ci.mptcpci_subflow_connids[i] = mpts->mpts_connid;
530
531 if (mpts->mpts_flags & MPTSF_INITIAL_SUB) {
532 const struct inpcb *inp;
533
534 inp = sotoinpcb(mpts->mpts_socket);
535
536 mptcp_ci.mptcpci_init_rxbytes = inp->inp_mstat.ms_total.ts_rxbytes;
537 mptcp_ci.mptcpci_init_txbytes = inp->inp_mstat.ms_total.ts_txbytes;
538 initial_info_set = 1;
539 }
540
541 mptcpstats_update(mptcp_ci.mptcpci_itfstats, MPTCP_ITFSTATS_SIZE, mpts);
542
543 i++;
544 }
545
546 if (initial_info_set == 0) {
547 mptcp_ci.mptcpci_init_rxbytes = mpte->mpte_init_rxbytes;
548 mptcp_ci.mptcpci_init_txbytes = mpte->mpte_init_txbytes;
549 }
550
551 if (mpte->mpte_flags & MPTE_FIRSTPARTY) {
552 mptcp_ci.mptcpci_flags |= MPTCPCI_FIRSTPARTY;
553 }
554
555 error = copyout(&mptcp_ci, aux_data, sizeof(mptcp_ci));
556 if (error != 0) {
557 os_log_error(mptcp_log_handle, "%s - %lx: copyout failed: %d\n",
558 __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte), error);
559 return error;
560 }
561 }
562
563 return 0;
564 }
565
566 /* Any stats of any subflow */
567 if (*cid == SAE_CONNID_ANY) {
568 const struct mptsub *mpts;
569 struct socket *so;
570 const struct inpcb *inp;
571 int error = 0;
572
573 mpts = TAILQ_FIRST(&mpte->mpte_subflows);
574 if (mpts == NULL) {
575 return ENXIO;
576 }
577
578 so = mpts->mpts_socket;
579 inp = sotoinpcb(so);
580
581 if (inp->inp_vflag & INP_IPV4) {
582 error = in_getconninfo(so, SAE_CONNID_ANY, flags, ifindex,
583 soerror, src, src_len, dst, dst_len,
584 aux_type, aux_data, aux_len);
585 } else {
586 error = in6_getconninfo(so, SAE_CONNID_ANY, flags, ifindex,
587 soerror, src, src_len, dst, dst_len,
588 aux_type, aux_data, aux_len);
589 }
590
591 if (error != 0) {
592 os_log_error(mptcp_log_handle, "%s - %lx:error from in_getconninfo %d\n",
593 __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte), error);
594 return error;
595 }
596
597 if (mpts->mpts_flags & MPTSF_MP_CAPABLE) {
598 *flags |= CIF_MP_CAPABLE;
599 }
600 if (mpts->mpts_flags & MPTSF_MP_DEGRADED) {
601 *flags |= CIF_MP_DEGRADED;
602 }
603 if (mpts->mpts_flags & MPTSF_MP_READY) {
604 *flags |= CIF_MP_READY;
605 }
606 if (mpts->mpts_flags & MPTSF_ACTIVE) {
607 *flags |= CIF_MP_ACTIVE;
608 }
609 if (mp_tp->mpt_version == MPTCP_VERSION_1) {
610 *flags |= CIF_MP_V1;
611 }
612
613 return 0;
614 } else {
615 /* Per-interface stats */
616 const struct mptsub *mpts, *orig_mpts = NULL;
617 struct conninfo_tcp tcp_ci;
618 const struct inpcb *inp;
619 struct socket *so;
620 int error = 0;
621 int index;
622
623 /* cid is thus an ifindex - range-check first! */
624 if (*cid > USHRT_MAX) {
625 return EINVAL;
626 }
627
628 bzero(&tcp_ci, sizeof(tcp_ci));
629
630 /* First, get a subflow to fill in the "regular" info. */
631 TAILQ_FOREACH(mpts, &mpte->mpte_subflows, mpts_entry) {
632 const struct ifnet *ifp = sotoinpcb(mpts->mpts_socket)->inp_last_outifp;
633
634 if (ifp && ifp->if_index == *cid) {
635 break;
636 }
637 }
638
639 if (mpts == NULL) {
640 /* No subflow there - well, let's just get the basic itf-info */
641 goto interface_info;
642 }
643
644 so = mpts->mpts_socket;
645 inp = sotoinpcb(so);
646
647 /* Give it USER_ADDR_NULL, because we are doing this on our own */
648 if (inp->inp_vflag & INP_IPV4) {
649 error = in_getconninfo(so, SAE_CONNID_ANY, flags, ifindex,
650 soerror, src, src_len, dst, dst_len,
651 aux_type, USER_ADDR_NULL, aux_len);
652 } else {
653 error = in6_getconninfo(so, SAE_CONNID_ANY, flags, ifindex,
654 soerror, src, src_len, dst, dst_len,
655 aux_type, USER_ADDR_NULL, aux_len);
656 }
657
658 if (error != 0) {
659 os_log_error(mptcp_log_handle, "%s - %lx:error from in_getconninfo %d\n",
660 __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte), error);
661 return error;
662 }
663
664 /* ToDo: Nobody is reading these flags on subflows. Why bother ? */
665 if (mpts->mpts_flags & MPTSF_MP_CAPABLE) {
666 *flags |= CIF_MP_CAPABLE;
667 }
668 if (mpts->mpts_flags & MPTSF_MP_DEGRADED) {
669 *flags |= CIF_MP_DEGRADED;
670 }
671 if (mpts->mpts_flags & MPTSF_MP_READY) {
672 *flags |= CIF_MP_READY;
673 }
674 if (mpts->mpts_flags & MPTSF_ACTIVE) {
675 *flags |= CIF_MP_ACTIVE;
676 }
677 if (mp_tp->mpt_version == MPTCP_VERSION_1) {
678 *flags |= CIF_MP_V1;
679 }
680
681 /*
682 * Now, we gather the metrics (aka., tcp_info) and roll them in
683 * across all subflows of this interface to build an aggregated
684 * view.
685 *
686 * We take the TCP_INFO from the first subflow as the "master",
687 * feeding into those fields that we do not roll.
688 */
689 if (aux_data != USER_ADDR_NULL) {
690 tcp_getconninfo(so, &tcp_ci);
691
692 orig_mpts = mpts;
693 TAILQ_FOREACH(mpts, &mpte->mpte_subflows, mpts_entry) {
694 const struct inpcb *mptsinp = sotoinpcb(mpts->mpts_socket);
695 const struct ifnet *ifp;
696
697 ifp = mptsinp->inp_last_outifp;
698
699 if (ifp == NULL || ifp->if_index != *cid || mpts == orig_mpts) {
700 continue;
701 }
702
703 /* Roll the itf-stats into the tcp_info */
704 tcp_ci.tcpci_tcp_info.tcpi_txbytes +=
705 mptsinp->inp_mstat.ms_total.ts_txbytes;
706 tcp_ci.tcpci_tcp_info.tcpi_rxbytes +=
707 mptsinp->inp_mstat.ms_total.ts_rxbytes;
708
709 tcp_ci.tcpci_tcp_info.tcpi_wifi_txbytes +=
710 mptsinp->inp_mstat.ms_wifi_infra.ts_txbytes +
711 mptsinp->inp_mstat.ms_wifi_non_infra.ts_txbytes;
712 tcp_ci.tcpci_tcp_info.tcpi_wifi_rxbytes +=
713 mptsinp->inp_mstat.ms_wifi_infra.ts_rxbytes +
714 mptsinp->inp_mstat.ms_wifi_non_infra.ts_rxbytes;
715
716 tcp_ci.tcpci_tcp_info.tcpi_wired_txbytes +=
717 mptsinp->inp_mstat.ms_wired.ts_txbytes;
718 tcp_ci.tcpci_tcp_info.tcpi_wired_rxbytes +=
719 mptsinp->inp_mstat.ms_wired.ts_rxbytes;
720
721 tcp_ci.tcpci_tcp_info.tcpi_cell_txbytes +=
722 mptsinp->inp_mstat.ms_cellular.ts_txbytes;
723 tcp_ci.tcpci_tcp_info.tcpi_cell_rxbytes +=
724 mptsinp->inp_mstat.ms_cellular.ts_rxbytes;
725 }
726 }
727
728 interface_info:
729 *aux_type = CIAUX_TCP;
730 if (*aux_len == 0) {
731 *aux_len = sizeof(tcp_ci);
732 } else if (aux_data != USER_ADDR_NULL) {
733 boolean_t create;
734
735 /*
736 * Finally, old subflows might have been closed - we
737 * want this data as well, so grab it from the interface
738 * stats.
739 */
740 create = orig_mpts != NULL;
741
742 /*
743 * When we found a subflow, we are willing to create a stats-index
744 * because we have some data to return. If there isn't a subflow,
745 * nor anything in the stats, return EINVAL. Because the
746 * ifindex belongs to something that doesn't exist.
747 */
748 index = mptcpstats_get_index_by_ifindex(mpte->mpte_itfstats, MPTCP_ITFSTATS_SIZE, (u_short)(*cid), false);
749 if (index == -1) {
750 os_log_error(mptcp_log_handle,
751 "%s - %lx: Asking for too many ifindex: %u subcount %u, mpts? %s\n",
752 __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte),
753 *cid, mpte->mpte_numflows,
754 orig_mpts ? "yes" : "no");
755
756 if (orig_mpts == NULL) {
757 return EINVAL;
758 }
759 } else {
760 struct mptcp_itf_stats *stats;
761
762 stats = &mpte->mpte_itfstats[index];
763
764 /* Roll the itf-stats into the tcp_info */
765 tcp_ci.tcpci_tcp_info.tcpi_last_outif = *cid;
766 tcp_ci.tcpci_tcp_info.tcpi_txbytes +=
767 stats->mpis_txbytes;
768 tcp_ci.tcpci_tcp_info.tcpi_rxbytes +=
769 stats->mpis_rxbytes;
770
771 tcp_ci.tcpci_tcp_info.tcpi_wifi_txbytes +=
772 stats->mpis_wifi_txbytes;
773 tcp_ci.tcpci_tcp_info.tcpi_wifi_rxbytes +=
774 stats->mpis_wifi_rxbytes;
775
776 tcp_ci.tcpci_tcp_info.tcpi_wired_txbytes +=
777 stats->mpis_wired_txbytes;
778 tcp_ci.tcpci_tcp_info.tcpi_wired_rxbytes +=
779 stats->mpis_wired_rxbytes;
780
781 tcp_ci.tcpci_tcp_info.tcpi_cell_txbytes +=
782 stats->mpis_cell_txbytes;
783 tcp_ci.tcpci_tcp_info.tcpi_cell_rxbytes +=
784 stats->mpis_cell_rxbytes;
785 }
786
787 *aux_len = min(*aux_len, sizeof(tcp_ci));
788 error = copyout(&tcp_ci, aux_data, *aux_len);
789 if (error != 0) {
790 return error;
791 }
792 }
793 }
794
795 return 0;
796 }
797
798 /*
799 * User-protocol pru_control callback.
800 */
801 static int
mptcp_usr_control(struct socket * mp_so,u_long cmd,caddr_t __sized_by (IOCPARM_LEN (cmd))data,struct ifnet * ifp,struct proc * p)802 mptcp_usr_control(struct socket *mp_so, u_long cmd,
803 caddr_t __sized_by(IOCPARM_LEN(cmd)) data,
804 struct ifnet *ifp, struct proc *p)
805 {
806 #pragma unused(ifp, p)
807 struct mppcb *mpp = mpsotomppcb(mp_so);
808 struct mptses *mpte;
809 int error = 0;
810
811 if (mpp == NULL || mpp->mpp_state == MPPCB_STATE_DEAD) {
812 error = EINVAL;
813 goto out;
814 }
815 mpte = mptompte(mpp);
816
817 switch (cmd) {
818 case SIOCGASSOCIDS32: { /* struct so_aidreq32 */
819 struct so_aidreq32 aidr;
820 bcopy(data, &aidr, sizeof(aidr));
821 error = mptcp_getassocids(mpte, &aidr.sar_cnt,
822 aidr.sar_aidp);
823 if (error == 0) {
824 bcopy(&aidr, data, sizeof(aidr));
825 }
826 break;
827 }
828
829 case SIOCGASSOCIDS64: { /* struct so_aidreq64 */
830 struct so_aidreq64 aidr;
831 bcopy(data, &aidr, sizeof(aidr));
832 error = mptcp_getassocids(mpte, &aidr.sar_cnt,
833 (user_addr_t)aidr.sar_aidp);
834 if (error == 0) {
835 bcopy(&aidr, data, sizeof(aidr));
836 }
837 break;
838 }
839
840 case SIOCGCONNIDS32: { /* struct so_cidreq32 */
841 struct so_cidreq32 cidr;
842 bcopy(data, &cidr, sizeof(cidr));
843 error = mptcp_getconnids(mpte, cidr.scr_aid, &cidr.scr_cnt,
844 cidr.scr_cidp);
845 if (error == 0) {
846 bcopy(&cidr, data, sizeof(cidr));
847 }
848 break;
849 }
850
851 case SIOCGCONNIDS64: { /* struct so_cidreq64 */
852 struct so_cidreq64 cidr;
853 bcopy(data, &cidr, sizeof(cidr));
854 error = mptcp_getconnids(mpte, cidr.scr_aid, &cidr.scr_cnt,
855 (user_addr_t)cidr.scr_cidp);
856 if (error == 0) {
857 bcopy(&cidr, data, sizeof(cidr));
858 }
859 break;
860 }
861
862 case SIOCGCONNINFO32: { /* struct so_cinforeq32 */
863 struct so_cinforeq32 cifr;
864 bcopy(data, &cifr, sizeof(cifr));
865 error = mptcp_getconninfo(mpte, &cifr.scir_cid,
866 &cifr.scir_flags, &cifr.scir_ifindex, &cifr.scir_error,
867 cifr.scir_src, &cifr.scir_src_len, cifr.scir_dst,
868 &cifr.scir_dst_len, &cifr.scir_aux_type, cifr.scir_aux_data,
869 &cifr.scir_aux_len);
870 if (error == 0) {
871 bcopy(&cifr, data, sizeof(cifr));
872 }
873 break;
874 }
875
876 case SIOCGCONNINFO64: { /* struct so_cinforeq64 */
877 struct so_cinforeq64 cifr;
878 bcopy(data, &cifr, sizeof(cifr));
879 error = mptcp_getconninfo(mpte, &cifr.scir_cid,
880 &cifr.scir_flags, &cifr.scir_ifindex, &cifr.scir_error,
881 (user_addr_t)cifr.scir_src, &cifr.scir_src_len,
882 (user_addr_t)cifr.scir_dst, &cifr.scir_dst_len,
883 &cifr.scir_aux_type, (user_addr_t)cifr.scir_aux_data,
884 &cifr.scir_aux_len);
885 if (error == 0) {
886 bcopy(&cifr, data, sizeof(cifr));
887 }
888 break;
889 }
890
891 default:
892 error = EOPNOTSUPP;
893 break;
894 }
895 out:
896 return error;
897 }
898
899 static int
mptcp_disconnect(struct mptses * mpte)900 mptcp_disconnect(struct mptses *mpte)
901 {
902 struct socket *mp_so;
903 struct mptcb *mp_tp;
904 int error = 0;
905
906 mp_so = mptetoso(mpte);
907 mp_tp = mpte->mpte_mptcb;
908
909 /* if we're not detached, go thru socket state checks */
910 if (!(mp_so->so_flags & SOF_PCBCLEARING) && !(mp_so->so_flags & SOF_DEFUNCT)) {
911 if (!(mp_so->so_state & (SS_ISCONNECTED |
912 SS_ISCONNECTING))) {
913 error = ENOTCONN;
914 goto out;
915 }
916 if (mp_so->so_state & SS_ISDISCONNECTING) {
917 error = EALREADY;
918 goto out;
919 }
920 }
921
922 mptcp_cancel_all_timers(mp_tp);
923 if (mp_tp->mpt_state < MPTCPS_ESTABLISHED) {
924 mptcp_close(mpte, mp_tp);
925 } else if (((mp_so->so_options & SO_LINGER) &&
926 mp_so->so_linger == 0) ||
927 (mp_so->so_flags1 & SOF1_DEFUNCTINPROG)) {
928 mptcp_drop(mpte, mp_tp, 0);
929 } else {
930 soisdisconnecting(mp_so);
931 sbflush(&mp_so->so_rcv);
932 if (mptcp_usrclosed(mpte) != NULL) {
933 mptcp_output(mpte);
934 }
935 }
936
937 if (error == 0) {
938 mptcp_subflow_workloop(mpte);
939 }
940
941 out:
942 return error;
943 }
944
945 /*
946 * Wrapper function to support disconnect on socket
947 */
948 static int
mptcp_usr_disconnect(struct socket * mp_so)949 mptcp_usr_disconnect(struct socket *mp_so)
950 {
951 return mptcp_disconnect(mpsotompte(mp_so));
952 }
953
954 /*
955 * User-protocol pru_disconnectx callback.
956 */
957 static int
mptcp_usr_disconnectx(struct socket * mp_so,sae_associd_t aid,sae_connid_t cid)958 mptcp_usr_disconnectx(struct socket *mp_so, sae_associd_t aid, sae_connid_t cid)
959 {
960 if (aid != SAE_ASSOCID_ANY && aid != SAE_ASSOCID_ALL) {
961 return EINVAL;
962 }
963
964 if (cid != SAE_CONNID_ANY && cid != SAE_CONNID_ALL) {
965 return EINVAL;
966 }
967
968 return mptcp_usr_disconnect(mp_so);
969 }
970
971 void
mptcp_finish_usrclosed(struct mptses * mpte)972 mptcp_finish_usrclosed(struct mptses *mpte)
973 {
974 struct mptcb *mp_tp = mpte->mpte_mptcb;
975 struct socket *mp_so = mptetoso(mpte);
976
977 if (mp_tp->mpt_state == MPTCPS_CLOSED || mp_tp->mpt_state == MPTCPS_TERMINATE) {
978 mpte = mptcp_close(mpte, mp_tp);
979 } else if (mp_tp->mpt_state >= MPTCPS_FIN_WAIT_2) {
980 soisdisconnected(mp_so);
981 } else {
982 struct mptsub *mpts;
983
984 TAILQ_FOREACH(mpts, &mpte->mpte_subflows, mpts_entry) {
985 if ((mp_so->so_state & (SS_CANTRCVMORE | SS_CANTSENDMORE)) ==
986 (SS_CANTRCVMORE | SS_CANTSENDMORE)) {
987 mptcp_subflow_disconnect(mpte, mpts);
988 } else {
989 mptcp_subflow_shutdown(mpte, mpts);
990 }
991 }
992 }
993 }
994
995 /*
996 * User issued close, and wish to trail thru shutdown states.
997 */
998 static struct mptses *
mptcp_usrclosed(struct mptses * mpte)999 mptcp_usrclosed(struct mptses *mpte)
1000 {
1001 struct mptcb *mp_tp = mpte->mpte_mptcb;
1002
1003 mptcp_close_fsm(mp_tp, MPCE_CLOSE);
1004
1005 /* Not everything has been acknowledged - don't close the subflows! */
1006 if (mp_tp->mpt_state != MPTCPS_TERMINATE &&
1007 mp_tp->mpt_sndnxt + 1 != mp_tp->mpt_sndmax) {
1008 return mpte;
1009 }
1010
1011 mptcp_finish_usrclosed(mpte);
1012
1013 return mpte;
1014 }
1015
1016 /*
1017 * After a receive, possible send some update to peer.
1018 */
1019 static int
mptcp_usr_rcvd(struct socket * mp_so,int flags)1020 mptcp_usr_rcvd(struct socket *mp_so, int flags)
1021 {
1022 #pragma unused(flags)
1023 struct mppcb *mpp = mpsotomppcb(mp_so);
1024 struct mptses *mpte;
1025 struct mptsub *mpts;
1026 int error = 0;
1027
1028 if (mpp == NULL || mpp->mpp_state == MPPCB_STATE_DEAD) {
1029 error = EINVAL;
1030 goto out;
1031 }
1032
1033 mpte = mptompte(mpp);
1034
1035 TAILQ_FOREACH(mpts, &mpte->mpte_subflows, mpts_entry) {
1036 struct socket *so = mpts->mpts_socket;
1037
1038 if (so->so_proto->pr_flags & PR_WANTRCVD && so->so_pcb != NULL) {
1039 (*so->so_proto->pr_usrreqs->pru_rcvd)(so, 0);
1040 }
1041 }
1042
1043 error = mptcp_output(mpte);
1044 out:
1045 return error;
1046 }
1047
1048 /*
1049 * Do a send by putting data in the output queue.
1050 */
1051 static int
mptcp_usr_send(struct socket * mp_so,int prus_flags,struct mbuf * m,struct sockaddr * nam,struct mbuf * control,struct proc * p)1052 mptcp_usr_send(struct socket *mp_so, int prus_flags, struct mbuf *m,
1053 struct sockaddr *nam, struct mbuf *control, struct proc *p)
1054 {
1055 #pragma unused(nam, p)
1056 struct mppcb *mpp = mpsotomppcb(mp_so);
1057 struct mptses *mpte;
1058 int error = 0;
1059
1060 if (prus_flags & (PRUS_OOB | PRUS_EOF)) {
1061 error = EOPNOTSUPP;
1062 goto out;
1063 }
1064
1065 if (nam != NULL) {
1066 error = EOPNOTSUPP;
1067 goto out;
1068 }
1069
1070 if (control != NULL && control->m_len != 0) {
1071 error = EOPNOTSUPP;
1072 goto out;
1073 }
1074
1075 if (mpp == NULL || mpp->mpp_state == MPPCB_STATE_DEAD) {
1076 error = ECONNRESET;
1077 goto out;
1078 }
1079 mpte = mptompte(mpp);
1080 VERIFY(mpte != NULL);
1081
1082 if (!(mp_so->so_state & SS_ISCONNECTED) &&
1083 !(mp_so->so_flags1 & SOF1_PRECONNECT_DATA)) {
1084 error = ENOTCONN;
1085 goto out;
1086 }
1087
1088 mptcp_insert_dsn(mpp, m);
1089 VERIFY(mp_so->so_snd.sb_flags & SB_NOCOMPRESS);
1090 sbappendstream(&mp_so->so_snd, m);
1091 m = NULL;
1092
1093 error = mptcp_output(mpte);
1094 if (error != 0) {
1095 goto out;
1096 }
1097
1098 if (mp_so->so_state & SS_ISCONNECTING) {
1099 if (mp_so->so_state & SS_NBIO) {
1100 error = EWOULDBLOCK;
1101 } else {
1102 error = sbwait(&mp_so->so_snd);
1103 }
1104 }
1105
1106 out:
1107 if (error) {
1108 if (m != NULL) {
1109 m_freem(m);
1110 }
1111 if (control != NULL) {
1112 m_freem(control);
1113 }
1114 }
1115 return error;
1116 }
1117
1118 /*
1119 * Mark the MPTCP connection as being incapable of further output.
1120 */
1121 static int
mptcp_usr_shutdown(struct socket * mp_so)1122 mptcp_usr_shutdown(struct socket *mp_so)
1123 {
1124 struct mppcb *mpp = mpsotomppcb(mp_so);
1125 struct mptses *mpte;
1126 int error = 0;
1127
1128 if (mpp == NULL || mpp->mpp_state == MPPCB_STATE_DEAD) {
1129 error = EINVAL;
1130 goto out;
1131 }
1132 mpte = mptompte(mpp);
1133 VERIFY(mpte != NULL);
1134
1135 socantsendmore(mp_so);
1136
1137 mpte = mptcp_usrclosed(mpte);
1138 if (mpte != NULL) {
1139 error = mptcp_output(mpte);
1140 }
1141 out:
1142 return error;
1143 }
1144
1145 /*
1146 * Copy the contents of uio into a properly sized mbuf chain.
1147 */
1148 static int
mptcp_uiotombuf(struct uio * uio,int how,user_ssize_t space,struct mbuf ** top)1149 mptcp_uiotombuf(struct uio *uio, int how, user_ssize_t space, struct mbuf **top)
1150 {
1151 struct mbuf *m, *mb, *nm = NULL, *mtail = NULL;
1152 int progress, len, error;
1153 user_ssize_t resid, tot;
1154
1155 VERIFY(top != NULL && *top == NULL);
1156
1157 /*
1158 * space can be zero or an arbitrary large value bound by
1159 * the total data supplied by the uio.
1160 */
1161 resid = uio_resid(uio);
1162 if (space > 0) {
1163 tot = MIN(resid, space);
1164 } else {
1165 tot = resid;
1166 }
1167
1168 if (tot < 0 || tot > INT_MAX) {
1169 return EINVAL;
1170 }
1171
1172 len = (int)tot;
1173 if (len == 0) {
1174 len = 1;
1175 }
1176
1177 /* Loop and append maximum sized mbufs to the chain tail. */
1178 while (len > 0) {
1179 uint32_t m_needed = 1;
1180
1181 if (len > MBIGCLBYTES) {
1182 mb = m_getpackets_internal(&m_needed, 1,
1183 how, 1, M16KCLBYTES);
1184 } else if (len > MCLBYTES) {
1185 mb = m_getpackets_internal(&m_needed, 1,
1186 how, 1, MBIGCLBYTES);
1187 } else if (len >= (signed)MINCLSIZE) {
1188 mb = m_getpackets_internal(&m_needed, 1,
1189 how, 1, MCLBYTES);
1190 } else {
1191 mb = m_gethdr(how, MT_DATA);
1192 }
1193
1194 /* Fail the whole operation if one mbuf can't be allocated. */
1195 if (mb == NULL) {
1196 if (nm != NULL) {
1197 m_freem(nm);
1198 }
1199 return ENOBUFS;
1200 }
1201
1202 /* Book keeping. */
1203 VERIFY(mb->m_flags & M_PKTHDR);
1204 len -= ((mb->m_flags & M_EXT) ? mb->m_ext.ext_size : MHLEN);
1205 if (mtail != NULL) {
1206 mtail->m_next = mb;
1207 } else {
1208 nm = mb;
1209 }
1210 mtail = mb;
1211 }
1212
1213 m = nm;
1214
1215 progress = 0;
1216 /* Fill all mbufs with uio data and update header information. */
1217 for (mb = m; mb != NULL; mb = mb->m_next) {
1218 /* tot >= 0 && tot <= INT_MAX (see above) */
1219 len = MIN((int)M_TRAILINGSPACE(mb), (int)(tot - progress));
1220
1221 error = uiomove(mtod(mb, char *), len, uio);
1222 if (error != 0) {
1223 m_freem(m);
1224 return error;
1225 }
1226
1227 /* each mbuf is M_PKTHDR chained via m_next */
1228 mb->m_len = len;
1229 mb->m_pkthdr.len = len;
1230
1231 progress += len;
1232 }
1233 VERIFY(progress == tot);
1234 *top = m;
1235 return 0;
1236 }
1237
1238 /*
1239 * MPTCP socket protocol-user socket send routine, derived from sosend().
1240 */
1241 static int
mptcp_usr_sosend(struct socket * mp_so,struct sockaddr * addr,struct uio * uio,struct mbuf * top,struct mbuf * control,int flags)1242 mptcp_usr_sosend(struct socket *mp_so, struct sockaddr *addr, struct uio *uio,
1243 struct mbuf *top, struct mbuf *control, int flags)
1244 {
1245 #pragma unused(addr)
1246 user_ssize_t resid, space;
1247 int error, sendflags;
1248 struct proc *p = current_proc();
1249 int sblocked = 0;
1250
1251 /* UIO is required for now, due to per-mbuf M_PKTHDR constrains */
1252 if (uio == NULL || top != NULL) {
1253 error = EINVAL;
1254 goto out;
1255 }
1256 resid = uio_resid(uio);
1257
1258 socket_lock(mp_so, 1);
1259 so_update_last_owner_locked(mp_so, p);
1260 so_update_policy(mp_so);
1261
1262 VERIFY(mp_so->so_type == SOCK_STREAM);
1263 VERIFY(!(mp_so->so_flags & SOF_MP_SUBFLOW));
1264
1265 if (flags & (MSG_OOB | MSG_DONTROUTE)) {
1266 error = EOPNOTSUPP;
1267 socket_unlock(mp_so, 1);
1268 goto out;
1269 }
1270
1271 /*
1272 * In theory resid should be unsigned. However, space must be
1273 * signed, as it might be less than 0 if we over-committed, and we
1274 * must use a signed comparison of space and resid. On the other
1275 * hand, a negative resid causes us to loop sending 0-length
1276 * segments to the protocol.
1277 */
1278 if (resid < 0 || resid > INT_MAX ||
1279 (flags & MSG_EOR) || control != NULL) {
1280 error = EINVAL;
1281 socket_unlock(mp_so, 1);
1282 goto out;
1283 }
1284
1285 OSIncrementAtomicLong(&p->p_stats->p_ru.ru_msgsnd);
1286
1287 do {
1288 error = sosendcheck(mp_so, NULL, resid, 0, 0, flags,
1289 &sblocked);
1290 if (error != 0) {
1291 goto release;
1292 }
1293
1294 space = sbspace(&mp_so->so_snd);
1295 do {
1296 socket_unlock(mp_so, 0);
1297 /*
1298 * Copy the data from userland into an mbuf chain.
1299 */
1300 error = mptcp_uiotombuf(uio, M_WAITOK, space, &top);
1301 if (error != 0) {
1302 socket_lock(mp_so, 0);
1303 goto release;
1304 }
1305 VERIFY(top != NULL);
1306 space -= resid - uio_resid(uio);
1307 resid = uio_resid(uio);
1308 socket_lock(mp_so, 0);
1309
1310 /*
1311 * Compute flags here, for pru_send and NKEs.
1312 */
1313 sendflags = (resid > 0 && space > 0) ?
1314 PRUS_MORETOCOME : 0;
1315
1316 /*
1317 * Socket filter processing
1318 */
1319 VERIFY(control == NULL);
1320 error = sflt_data_out(mp_so, NULL, &top, &control, 0);
1321 if (error != 0) {
1322 if (error == EJUSTRETURN) {
1323 error = 0;
1324 top = NULL;
1325 /* always free control if any */
1326 }
1327 goto release;
1328 }
1329 if (control != NULL) {
1330 m_freem(control);
1331 control = NULL;
1332 }
1333
1334 /*
1335 * Pass data to protocol.
1336 */
1337 error = (*mp_so->so_proto->pr_usrreqs->pru_send)
1338 (mp_so, sendflags, top, NULL, NULL, p);
1339
1340 top = NULL;
1341 if (error != 0) {
1342 goto release;
1343 }
1344 } while (resid != 0 && space > 0);
1345 } while (resid != 0);
1346
1347 release:
1348 if (sblocked) {
1349 sbunlock(&mp_so->so_snd, FALSE); /* will unlock socket */
1350 } else {
1351 socket_unlock(mp_so, 1);
1352 }
1353 out:
1354 if (top != NULL) {
1355 m_freem(top);
1356 }
1357 if (control != NULL) {
1358 m_freem(control);
1359 }
1360
1361 soclearfastopen(mp_so);
1362
1363 return error;
1364 }
1365
1366 /*
1367 * Called to filter SOPT_{SET,GET} for SOL_SOCKET level socket options.
1368 * This routine simply indicates to the caller whether or not to proceed
1369 * further with the given socket option. This is invoked by sosetoptlock()
1370 * and sogetoptlock().
1371 */
1372 static int
mptcp_usr_socheckopt(struct socket * mp_so,struct sockopt * sopt)1373 mptcp_usr_socheckopt(struct socket *mp_so, struct sockopt *sopt)
1374 {
1375 #pragma unused(mp_so)
1376 int error = 0;
1377
1378 VERIFY(sopt->sopt_level == SOL_SOCKET);
1379
1380 /*
1381 * We could check for sopt_dir (set/get) here, but we'll just
1382 * let the caller deal with it as appropriate; therefore the
1383 * following is a superset of the socket options which we
1384 * allow for set/get.
1385 *
1386 * XXX: [email protected]
1387 *
1388 * Need to consider the following cases:
1389 *
1390 * a. Certain socket options don't have a clear definition
1391 * on the expected behavior post connect(2). At the time
1392 * those options are issued on the MP socket, there may
1393 * be existing subflow sockets that are already connected.
1394 */
1395 switch (sopt->sopt_name) {
1396 case SO_LINGER: /* MP */
1397 case SO_LINGER_SEC: /* MP */
1398 case SO_TYPE: /* MP */
1399 case SO_NREAD: /* MP */
1400 case SO_NWRITE: /* MP */
1401 case SO_ERROR: /* MP */
1402 case SO_SNDBUF: /* MP */
1403 case SO_RCVBUF: /* MP */
1404 case SO_SNDLOWAT: /* MP */
1405 case SO_RCVLOWAT: /* MP */
1406 case SO_SNDTIMEO: /* MP */
1407 case SO_RCVTIMEO: /* MP */
1408 case SO_NKE: /* MP */
1409 case SO_NOSIGPIPE: /* MP */
1410 case SO_NOADDRERR: /* MP */
1411 case SO_LABEL: /* MP */
1412 case SO_PEERLABEL: /* MP */
1413 case SO_DEFUNCTIT: /* MP */
1414 case SO_DEFUNCTOK: /* MP */
1415 case SO_ISDEFUNCT: /* MP */
1416 case SO_TRAFFIC_CLASS_DBG: /* MP */
1417 case SO_DELEGATED: /* MP */
1418 case SO_DELEGATED_UUID: /* MP */
1419 #if NECP
1420 case SO_NECP_ATTRIBUTES:
1421 case SO_NECP_CLIENTUUID:
1422 #endif /* NECP */
1423 case SO_MPKL_SEND_INFO:
1424 /*
1425 * Tell the caller that these options are to be processed.
1426 */
1427 break;
1428
1429 case SO_DEBUG: /* MP + subflow */
1430 case SO_KEEPALIVE: /* MP + subflow */
1431 case SO_USELOOPBACK: /* MP + subflow */
1432 case SO_RANDOMPORT: /* MP + subflow */
1433 case SO_TRAFFIC_CLASS: /* MP + subflow */
1434 case SO_RECV_TRAFFIC_CLASS: /* MP + subflow */
1435 case SO_PRIVILEGED_TRAFFIC_CLASS: /* MP + subflow */
1436 case SO_RECV_ANYIF: /* MP + subflow */
1437 case SO_RESTRICTIONS: /* MP + subflow */
1438 case SO_FLUSH: /* MP + subflow */
1439 case SO_NOWAKEFROMSLEEP:
1440 case SO_NOAPNFALLBK:
1441 case SO_MARK_CELLFALLBACK:
1442 case SO_MARK_CELLFALLBACK_UUID:
1443 case SO_MARK_KNOWN_TRACKER:
1444 case SO_MARK_KNOWN_TRACKER_NON_APP_INITIATED:
1445 case SO_MARK_APPROVED_APP_DOMAIN:
1446 case SO_FALLBACK_MODE:
1447 /*
1448 * Tell the caller that these options are to be processed;
1449 * these will also be recorded later by mptcp_setopt().
1450 *
1451 * NOTE: Only support integer option value for now.
1452 */
1453 if (sopt->sopt_valsize != sizeof(int)) {
1454 error = EINVAL;
1455 }
1456 break;
1457
1458 default:
1459 /*
1460 * Tell the caller to stop immediately and return an error.
1461 */
1462 error = ENOPROTOOPT;
1463 break;
1464 }
1465
1466 return error;
1467 }
1468
1469 /*
1470 * Issue SOPT_SET for all MPTCP subflows (for integer option values.)
1471 */
1472 static int
mptcp_setopt_apply(struct mptses * mpte,struct mptopt * mpo)1473 mptcp_setopt_apply(struct mptses *mpte, struct mptopt *mpo)
1474 {
1475 struct socket *mp_so;
1476 struct mptsub *mpts;
1477 struct mptopt smpo;
1478 int error = 0;
1479
1480 /* just bail now if this isn't applicable to subflow sockets */
1481 if (!(mpo->mpo_flags & MPOF_SUBFLOW_OK)) {
1482 error = ENOPROTOOPT;
1483 goto out;
1484 }
1485
1486 /*
1487 * Skip those that are handled internally; these options
1488 * should not have been recorded and marked with the
1489 * MPOF_SUBFLOW_OK by mptcp_setopt(), but just in case.
1490 */
1491 if (mpo->mpo_level == SOL_SOCKET &&
1492 (mpo->mpo_name == SO_NOSIGPIPE || mpo->mpo_name == SO_NOADDRERR)) {
1493 error = ENOPROTOOPT;
1494 goto out;
1495 }
1496
1497 mp_so = mptetoso(mpte);
1498
1499 /*
1500 * Don't bother going further if there's no subflow; mark the option
1501 * with MPOF_INTERIM so that we know whether or not to remove this
1502 * option upon encountering an error while issuing it during subflow
1503 * socket creation.
1504 */
1505 if (mpte->mpte_numflows == 0) {
1506 VERIFY(TAILQ_EMPTY(&mpte->mpte_subflows));
1507 mpo->mpo_flags |= MPOF_INTERIM;
1508 /* return success */
1509 goto out;
1510 }
1511
1512 bzero(&smpo, sizeof(smpo));
1513 smpo.mpo_flags |= MPOF_SUBFLOW_OK;
1514 smpo.mpo_level = mpo->mpo_level;
1515 smpo.mpo_name = mpo->mpo_name;
1516
1517 /* grab exisiting values in case we need to rollback */
1518 TAILQ_FOREACH(mpts, &mpte->mpte_subflows, mpts_entry) {
1519 struct socket *so;
1520
1521 mpts->mpts_flags &= ~(MPTSF_SOPT_OLDVAL | MPTSF_SOPT_INPROG);
1522 mpts->mpts_oldintval = 0;
1523 smpo.mpo_intval = 0;
1524 VERIFY(mpts->mpts_socket != NULL);
1525 so = mpts->mpts_socket;
1526 if (mptcp_subflow_sogetopt(mpte, so, &smpo) == 0) {
1527 mpts->mpts_flags |= MPTSF_SOPT_OLDVAL;
1528 mpts->mpts_oldintval = smpo.mpo_intval;
1529 }
1530 }
1531
1532 /* apply socket option */
1533 TAILQ_FOREACH(mpts, &mpte->mpte_subflows, mpts_entry) {
1534 struct socket *so;
1535
1536 mpts->mpts_flags |= MPTSF_SOPT_INPROG;
1537 VERIFY(mpts->mpts_socket != NULL);
1538 so = mpts->mpts_socket;
1539 error = mptcp_subflow_sosetopt(mpte, mpts, mpo);
1540 if (error != 0) {
1541 break;
1542 }
1543 }
1544
1545 /* cleanup, and rollback if needed */
1546 TAILQ_FOREACH(mpts, &mpte->mpte_subflows, mpts_entry) {
1547 struct socket *so;
1548
1549 if (!(mpts->mpts_flags & MPTSF_SOPT_INPROG)) {
1550 /* clear in case it's set */
1551 mpts->mpts_flags &= ~MPTSF_SOPT_OLDVAL;
1552 mpts->mpts_oldintval = 0;
1553 continue;
1554 }
1555 if (!(mpts->mpts_flags & MPTSF_SOPT_OLDVAL)) {
1556 mpts->mpts_flags &= ~MPTSF_SOPT_INPROG;
1557 VERIFY(mpts->mpts_oldintval == 0);
1558 continue;
1559 }
1560 /* error during sosetopt, so roll it back */
1561 if (error != 0) {
1562 VERIFY(mpts->mpts_socket != NULL);
1563 so = mpts->mpts_socket;
1564 smpo.mpo_intval = mpts->mpts_oldintval;
1565 mptcp_subflow_sosetopt(mpte, mpts, &smpo);
1566 }
1567 mpts->mpts_oldintval = 0;
1568 mpts->mpts_flags &= ~(MPTSF_SOPT_OLDVAL | MPTSF_SOPT_INPROG);
1569 }
1570
1571 out:
1572 return error;
1573 }
1574
1575 /*
1576 * Handle SOPT_SET for socket options issued on MP socket.
1577 */
1578 static int
mptcp_setopt(struct mptses * mpte,struct sockopt * sopt)1579 mptcp_setopt(struct mptses *mpte, struct sockopt *sopt)
1580 {
1581 int error = 0, optval = 0, level, optname, rec = 1;
1582 struct mptopt smpo, *mpo = NULL;
1583 struct socket *mp_so;
1584
1585 level = sopt->sopt_level;
1586 optname = sopt->sopt_name;
1587
1588 mp_so = mptetoso(mpte);
1589
1590 VERIFY(!(mpsotomppcb(mp_so)->mpp_flags & MPP_INSIDE_SETGETOPT));
1591 mpsotomppcb(mp_so)->mpp_flags |= MPP_INSIDE_SETGETOPT;
1592
1593 /*
1594 * Record socket options which are applicable to subflow sockets so
1595 * that we can replay them for new ones; see mptcp_usr_socheckopt()
1596 * for the list of eligible socket-level options.
1597 */
1598 if (level == SOL_SOCKET) {
1599 switch (optname) {
1600 case SO_DEBUG:
1601 case SO_KEEPALIVE:
1602 case SO_USELOOPBACK:
1603 case SO_RANDOMPORT:
1604 case SO_TRAFFIC_CLASS:
1605 case SO_RECV_TRAFFIC_CLASS:
1606 case SO_PRIVILEGED_TRAFFIC_CLASS:
1607 case SO_RECV_ANYIF:
1608 case SO_RESTRICTIONS:
1609 case SO_NOWAKEFROMSLEEP:
1610 case SO_NOAPNFALLBK:
1611 case SO_MARK_CELLFALLBACK:
1612 case SO_MARK_KNOWN_TRACKER:
1613 case SO_MARK_KNOWN_TRACKER_NON_APP_INITIATED:
1614 case SO_MARK_APPROVED_APP_DOMAIN:
1615 case SO_FALLBACK_MODE:
1616 /* record it */
1617 break;
1618 case SO_FLUSH:
1619 /* don't record it */
1620 rec = 0;
1621 break;
1622
1623 /* Next ones, record at MPTCP-level */
1624 case SO_DELEGATED:
1625 error = sooptcopyin(sopt, &mpte->mpte_epid,
1626 sizeof(int), sizeof(int));
1627 if (error != 0) {
1628 goto err_out;
1629 }
1630
1631 goto out;
1632 case SO_DELEGATED_UUID:
1633 error = sooptcopyin(sopt, &mpte->mpte_euuid,
1634 sizeof(uuid_t), sizeof(uuid_t));
1635 if (error != 0) {
1636 goto err_out;
1637 }
1638
1639 goto out;
1640 #if NECP
1641 case SO_NECP_CLIENTUUID:
1642 if (!uuid_is_null(mpsotomppcb(mp_so)->necp_client_uuid)) {
1643 error = EINVAL;
1644 goto err_out;
1645 }
1646
1647 error = sooptcopyin(sopt, &mpsotomppcb(mp_so)->necp_client_uuid,
1648 sizeof(uuid_t), sizeof(uuid_t));
1649 if (error != 0) {
1650 goto err_out;
1651 }
1652
1653 mpsotomppcb(mp_so)->necp_cb = mptcp_session_necp_cb;
1654 error = necp_client_register_multipath_cb(mp_so->last_pid,
1655 mpsotomppcb(mp_so)->necp_client_uuid,
1656 mpsotomppcb(mp_so));
1657 if (error) {
1658 goto err_out;
1659 }
1660
1661 if (uuid_is_null(mpsotomppcb(mp_so)->necp_client_uuid)) {
1662 error = EINVAL;
1663 goto err_out;
1664 }
1665
1666 goto out;
1667 case SO_NECP_ATTRIBUTES:
1668 error = necp_set_socket_attributes(&mpsotomppcb(mp_so)->inp_necp_attributes, sopt);
1669 if (error) {
1670 goto err_out;
1671 }
1672
1673 goto out;
1674 #endif /* NECP */
1675 default:
1676 /* nothing to do; just return */
1677 goto out;
1678 }
1679 } else if (sopt->sopt_level == IPPROTO_IP) {
1680 switch (optname) {
1681 case IP_TOS:
1682 /* eligible; record it */
1683 break;
1684 default:
1685 /* not eligible */
1686 error = ENOPROTOOPT;
1687 goto err_out;
1688 }
1689 } else if (sopt->sopt_level == IPPROTO_IPV6) {
1690 switch (optname) {
1691 case IPV6_TCLASS:
1692 /* eligible; record it */
1693 break;
1694 default:
1695 /* not eligible */
1696 error = ENOPROTOOPT;
1697 goto err_out;
1698 }
1699 } else {
1700 switch (optname) {
1701 case TCP_NODELAY:
1702 case TCP_RXT_FINDROP:
1703 case TCP_KEEPALIVE:
1704 case TCP_KEEPINTVL:
1705 case TCP_KEEPCNT:
1706 case TCP_CONNECTIONTIMEOUT:
1707 case TCP_RXT_CONNDROPTIME:
1708 case PERSIST_TIMEOUT:
1709 case TCP_ADAPTIVE_READ_TIMEOUT:
1710 case TCP_ADAPTIVE_WRITE_TIMEOUT:
1711 case TCP_FASTOPEN_FORCE_ENABLE:
1712 /* eligible; record it */
1713 break;
1714 case TCP_NOTSENT_LOWAT:
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) {
1722 error = EINVAL;
1723 goto err_out;
1724 } else {
1725 if (optval == 0) {
1726 mp_so->so_flags &= ~SOF_NOTSENT_LOWAT;
1727 error = mptcp_set_notsent_lowat(mpte, 0);
1728 } else {
1729 mp_so->so_flags |= SOF_NOTSENT_LOWAT;
1730 error = mptcp_set_notsent_lowat(mpte,
1731 optval);
1732 }
1733
1734 if (error) {
1735 goto err_out;
1736 }
1737 }
1738 goto out;
1739 case MPTCP_SERVICE_TYPE:
1740 /* record at MPTCP level */
1741 error = sooptcopyin(sopt, &optval, sizeof(optval),
1742 sizeof(optval));
1743 if (error) {
1744 goto err_out;
1745 }
1746 if (optval < 0 || optval >= MPTCP_SVCTYPE_MAX) {
1747 error = EINVAL;
1748 goto err_out;
1749 }
1750
1751 if (mptcp_entitlement_check(mp_so, (uint8_t)optval) < 0) {
1752 error = EACCES;
1753 goto err_out;
1754 }
1755
1756 mpte->mpte_svctype = (uint8_t)optval;
1757 mpte->mpte_flags |= MPTE_SVCTYPE_CHECKED;
1758
1759 goto out;
1760 case MPTCP_ALTERNATE_PORT:
1761 /* record at MPTCP level */
1762 error = sooptcopyin(sopt, &optval, sizeof(optval),
1763 sizeof(optval));
1764 if (error) {
1765 goto err_out;
1766 }
1767
1768 if (optval < 0 || optval > UINT16_MAX) {
1769 error = EINVAL;
1770 goto err_out;
1771 }
1772
1773 mpte->mpte_alternate_port = (uint16_t)optval;
1774
1775 goto out;
1776 case MPTCP_FORCE_ENABLE:
1777 /* record at MPTCP level */
1778 error = sooptcopyin(sopt, &optval, sizeof(optval),
1779 sizeof(optval));
1780 if (error) {
1781 goto err_out;
1782 }
1783
1784 if (optval < 0 || optval > 1) {
1785 error = EINVAL;
1786 goto err_out;
1787 }
1788
1789 if (optval) {
1790 mpte->mpte_flags |= MPTE_FORCE_ENABLE;
1791 } else {
1792 mpte->mpte_flags &= ~MPTE_FORCE_ENABLE;
1793 }
1794
1795 goto out;
1796 case MPTCP_FORCE_VERSION:
1797 error = sooptcopyin(sopt, &optval, sizeof(optval),
1798 sizeof(optval));
1799 if (error) {
1800 goto err_out;
1801 }
1802
1803 if (optval != 0 && optval != 1) {
1804 error = EINVAL;
1805 goto err_out;
1806 }
1807
1808 if (optval == 0) {
1809 mpte->mpte_flags |= MPTE_FORCE_V0;
1810 mpte->mpte_flags &= ~MPTE_FORCE_V1;
1811 } else {
1812 mpte->mpte_flags |= MPTE_FORCE_V1;
1813 mpte->mpte_flags &= ~MPTE_FORCE_V0;
1814 }
1815
1816 goto out;
1817 case MPTCP_EXPECTED_PROGRESS_TARGET:
1818 {
1819 struct mptcb *mp_tp = mpte->mpte_mptcb;
1820 uint64_t mach_time_target;
1821 uint64_t nanoseconds;
1822
1823 if (mpte->mpte_svctype != MPTCP_SVCTYPE_TARGET_BASED) {
1824 os_log(mptcp_log_handle, "%s - %lx: Can't set urgent activity when svctype is %u\n",
1825 __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte), mpte->mpte_svctype);
1826 error = EINVAL;
1827 goto err_out;
1828 }
1829
1830 error = sooptcopyin(sopt, &mach_time_target, sizeof(mach_time_target), sizeof(mach_time_target));
1831 if (error) {
1832 goto err_out;
1833 }
1834
1835 if (!mptcp_ok_to_create_subflows(mp_tp)) {
1836 os_log(mptcp_log_handle, "%s - %lx: Not ok to create subflows, state %u flags %#x\n",
1837 __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte), mp_tp->mpt_state, mp_tp->mpt_flags);
1838 error = EINVAL;
1839 goto err_out;
1840 }
1841
1842 if (mach_time_target) {
1843 uint64_t time_now = 0;
1844 uint64_t time_now_nanoseconds;
1845
1846 absolutetime_to_nanoseconds(mach_time_target, &nanoseconds);
1847 nanoseconds = nanoseconds - (mptcp_expected_progress_headstart * NSEC_PER_MSEC);
1848
1849 time_now = mach_continuous_time();
1850 absolutetime_to_nanoseconds(time_now, &time_now_nanoseconds);
1851
1852 nanoseconds_to_absolutetime(nanoseconds, &mach_time_target);
1853 /* If the timer is already running and it would
1854 * fire in less than mptcp_expected_progress_headstart
1855 * seconds, then it's not worth canceling it.
1856 */
1857 if (mpte->mpte_time_target &&
1858 mpte->mpte_time_target < time_now &&
1859 time_now_nanoseconds > nanoseconds - (mptcp_expected_progress_headstart * NSEC_PER_MSEC)) {
1860 os_log(mptcp_log_handle, "%s - %lx: Not rescheduling timer %llu now %llu target %llu\n",
1861 __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte),
1862 mpte->mpte_time_target,
1863 time_now,
1864 mach_time_target);
1865 goto out;
1866 }
1867 }
1868
1869 mpte->mpte_time_target = mach_time_target;
1870 mptcp_set_urgency_timer(mpte);
1871
1872 goto out;
1873 }
1874 default:
1875 /* not eligible */
1876 error = ENOPROTOOPT;
1877 goto err_out;
1878 }
1879 }
1880
1881 if ((error = sooptcopyin(sopt, &optval, sizeof(optval),
1882 sizeof(optval))) != 0) {
1883 goto err_out;
1884 }
1885
1886 if (rec) {
1887 /* search for an existing one; if not found, allocate */
1888 if ((mpo = mptcp_sopt_find(mpte, sopt)) == NULL) {
1889 mpo = mptcp_sopt_alloc();
1890 }
1891
1892 /* initialize or update, as needed */
1893 mpo->mpo_intval = optval;
1894 if (!(mpo->mpo_flags & MPOF_ATTACHED)) {
1895 mpo->mpo_level = level;
1896 mpo->mpo_name = optname;
1897 mptcp_sopt_insert(mpte, mpo);
1898 }
1899 /* this can be issued on the subflow socket */
1900 mpo->mpo_flags |= MPOF_SUBFLOW_OK;
1901 } else {
1902 bzero(&smpo, sizeof(smpo));
1903 mpo = &smpo;
1904 mpo->mpo_flags |= MPOF_SUBFLOW_OK;
1905 mpo->mpo_level = level;
1906 mpo->mpo_name = optname;
1907 mpo->mpo_intval = optval;
1908 }
1909
1910 /* issue this socket option on existing subflows */
1911 error = mptcp_setopt_apply(mpte, mpo);
1912 if (error != 0 && (mpo->mpo_flags & MPOF_ATTACHED)) {
1913 VERIFY(mpo != &smpo);
1914 mptcp_sopt_remove(mpte, mpo);
1915 mptcp_sopt_free(mpo);
1916 }
1917 if (mpo == &smpo) {
1918 mpo->mpo_flags &= ~MPOF_INTERIM;
1919 }
1920
1921 if (error) {
1922 goto err_out;
1923 }
1924
1925 out:
1926
1927 mpsotomppcb(mp_so)->mpp_flags &= ~MPP_INSIDE_SETGETOPT;
1928 return 0;
1929
1930 err_out:
1931 os_log_error(mptcp_log_handle, "%s - %lx: sopt %s (%d, %d) val %d can't be issued error %d\n",
1932 __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte),
1933 mptcp_sopt2str(level, optname), level, optname, optval, error);
1934 mpsotomppcb(mp_so)->mpp_flags &= ~MPP_INSIDE_SETGETOPT;
1935 return error;
1936 }
1937
1938 static void
mptcp_fill_info_bytestats(struct tcp_info * ti,struct mptses * mpte)1939 mptcp_fill_info_bytestats(struct tcp_info *ti, struct mptses *mpte)
1940 {
1941 struct mptsub *mpts;
1942 int i;
1943
1944 TAILQ_FOREACH(mpts, &mpte->mpte_subflows, mpts_entry) {
1945 const struct inpcb *inp = sotoinpcb(mpts->mpts_socket);
1946
1947 if (inp == NULL) {
1948 continue;
1949 }
1950
1951 ti->tcpi_txbytes += inp->inp_mstat.ms_total.ts_txbytes;
1952 ti->tcpi_rxbytes += inp->inp_mstat.ms_total.ts_rxbytes;
1953 ti->tcpi_cell_txbytes += inp->inp_mstat.ms_cellular.ts_txbytes;
1954 ti->tcpi_cell_rxbytes += inp->inp_mstat.ms_cellular.ts_rxbytes;
1955 ti->tcpi_wifi_txbytes += inp->inp_mstat.ms_wifi_infra.ts_txbytes +
1956 inp->inp_mstat.ms_wifi_non_infra.ts_txbytes;
1957 ti->tcpi_wifi_rxbytes += inp->inp_mstat.ms_wifi_infra.ts_rxbytes +
1958 inp->inp_mstat.ms_wifi_non_infra.ts_rxbytes;
1959 ti->tcpi_wired_txbytes += inp->inp_mstat.ms_wired.ts_txbytes;
1960 ti->tcpi_wired_rxbytes += inp->inp_mstat.ms_wired.ts_rxbytes;
1961 }
1962
1963 for (i = 0; i < MPTCP_ITFSTATS_SIZE; i++) {
1964 struct mptcp_itf_stats *stats = &mpte->mpte_itfstats[i];
1965
1966 ti->tcpi_txbytes += stats->mpis_txbytes;
1967 ti->tcpi_rxbytes += stats->mpis_rxbytes;
1968
1969 ti->tcpi_wifi_txbytes += stats->mpis_wifi_txbytes;
1970 ti->tcpi_wifi_rxbytes += stats->mpis_wifi_rxbytes;
1971
1972 ti->tcpi_wired_txbytes += stats->mpis_wired_txbytes;
1973 ti->tcpi_wired_rxbytes += stats->mpis_wired_rxbytes;
1974
1975 ti->tcpi_cell_txbytes += stats->mpis_cell_txbytes;
1976 ti->tcpi_cell_rxbytes += stats->mpis_cell_rxbytes;
1977 }
1978 }
1979
1980 static void
mptcp_fill_info(struct mptses * mpte,struct tcp_info * ti)1981 mptcp_fill_info(struct mptses *mpte, struct tcp_info *ti)
1982 {
1983 struct mptsub *actsub = mpte->mpte_active_sub;
1984 struct mptcb *mp_tp = mpte->mpte_mptcb;
1985 struct tcpcb *acttp = NULL;
1986
1987 if (actsub) {
1988 acttp = sototcpcb(actsub->mpts_socket);
1989 }
1990
1991 bzero(ti, sizeof(*ti));
1992
1993 ti->tcpi_state = (uint8_t)mp_tp->mpt_state;
1994 /* tcpi_options */
1995 /* tcpi_snd_wscale */
1996 /* tcpi_rcv_wscale */
1997 /* tcpi_flags */
1998 if (acttp) {
1999 ti->tcpi_rto = acttp->t_timer[TCPT_REXMT] ? acttp->t_rxtcur : 0;
2000 }
2001
2002 /* tcpi_snd_mss */
2003 /* tcpi_rcv_mss */
2004 if (acttp) {
2005 ti->tcpi_rttcur = acttp->t_rttcur;
2006 ti->tcpi_srtt = acttp->t_srtt >> TCP_RTT_SHIFT;
2007 ti->tcpi_rttvar = acttp->t_rttvar >> TCP_RTTVAR_SHIFT;
2008 ti->tcpi_rttbest = acttp->t_rttbest >> TCP_RTT_SHIFT;
2009 ti->tcpi_rcv_srtt = acttp->rcv_srtt >> TCP_RTT_SHIFT;
2010 }
2011 /* tcpi_snd_ssthresh */
2012 /* tcpi_snd_cwnd */
2013 /* tcpi_rcv_space */
2014 ti->tcpi_snd_wnd = mp_tp->mpt_sndwnd;
2015 ti->tcpi_snd_nxt = (uint32_t)mp_tp->mpt_sndnxt;
2016 ti->tcpi_rcv_nxt = (uint32_t)mp_tp->mpt_rcvnxt;
2017 if (acttp) {
2018 ti->tcpi_last_outif = (acttp->t_inpcb->inp_last_outifp == NULL) ? 0 :
2019 acttp->t_inpcb->inp_last_outifp->if_index;
2020 }
2021
2022 mptcp_fill_info_bytestats(ti, mpte);
2023 /* tcpi_txpackets */
2024
2025 /* tcpi_txretransmitbytes */
2026 /* tcpi_txunacked */
2027 /* tcpi_rxpackets */
2028
2029 /* tcpi_rxduplicatebytes */
2030 /* tcpi_rxoutoforderbytes */
2031 /* tcpi_snd_bw */
2032 /* tcpi_synrexmits */
2033 /* tcpi_unused1 */
2034 /* tcpi_unused2 */
2035 /* tcpi_cell_rxpackets */
2036
2037 /* tcpi_cell_txpackets */
2038
2039 /* tcpi_wifi_rxpackets */
2040
2041 /* tcpi_wifi_txpackets */
2042
2043 /* tcpi_wired_rxpackets */
2044 /* tcpi_wired_txpackets */
2045 /* tcpi_connstatus */
2046 /* TFO-stuff */
2047 /* ECN stuff */
2048 /* tcpi_ecn_recv_ce */
2049 /* tcpi_ecn_recv_cwr */
2050 if (acttp) {
2051 ti->tcpi_rcvoopack = acttp->t_rcvoopack;
2052 }
2053 /* tcpi_pawsdrop */
2054 /* tcpi_sack_recovery_episode */
2055 /* tcpi_reordered_pkts */
2056 /* tcpi_dsack_sent */
2057 /* tcpi_dsack_recvd */
2058 /* tcpi_flowhash */
2059 if (acttp) {
2060 ti->tcpi_txretransmitpackets = acttp->t_stat.rxmitpkts;
2061 }
2062 }
2063
2064 /*
2065 * Handle SOPT_GET for socket options issued on MP socket.
2066 */
2067 static int
mptcp_getopt(struct mptses * mpte,struct sockopt * sopt)2068 mptcp_getopt(struct mptses *mpte, struct sockopt *sopt)
2069 {
2070 int error = 0, optval = 0;
2071 struct socket *__single mp_so;
2072
2073 mp_so = mptetoso(mpte);
2074
2075 VERIFY(!(mpsotomppcb(mp_so)->mpp_flags & MPP_INSIDE_SETGETOPT));
2076 mpsotomppcb(mp_so)->mpp_flags |= MPP_INSIDE_SETGETOPT;
2077
2078 /*
2079 * We only handle SOPT_GET for TCP level socket options; we should
2080 * not get here for socket level options since they are already
2081 * handled at the socket layer.
2082 */
2083 if (sopt->sopt_level != IPPROTO_TCP) {
2084 error = ENOPROTOOPT;
2085 goto out;
2086 }
2087
2088 switch (sopt->sopt_name) {
2089 case PERSIST_TIMEOUT:
2090 /* Only case for which we have a non-zero default */
2091 optval = tcp_max_persist_timeout;
2092 OS_FALLTHROUGH;
2093 case TCP_NODELAY:
2094 case TCP_RXT_FINDROP:
2095 case TCP_KEEPALIVE:
2096 case TCP_KEEPINTVL:
2097 case TCP_KEEPCNT:
2098 case TCP_CONNECTIONTIMEOUT:
2099 case TCP_RXT_CONNDROPTIME:
2100 case TCP_ADAPTIVE_READ_TIMEOUT:
2101 case TCP_ADAPTIVE_WRITE_TIMEOUT:
2102 case TCP_FASTOPEN_FORCE_ENABLE:
2103 {
2104 struct mptopt *__single mpo = mptcp_sopt_find(mpte, sopt);
2105
2106 if (mpo != NULL) {
2107 optval = mpo->mpo_intval;
2108 }
2109 break;
2110 }
2111
2112 /* The next ones are stored at the MPTCP-level */
2113 case TCP_NOTSENT_LOWAT:
2114 if (mptetoso(mpte)->so_flags & SOF_NOTSENT_LOWAT) {
2115 optval = mptcp_get_notsent_lowat(mpte);
2116 } else {
2117 optval = 0;
2118 }
2119 break;
2120 case TCP_INFO:
2121 {
2122 struct tcp_info ti;
2123
2124 mptcp_fill_info(mpte, &ti);
2125 error = sooptcopyout(sopt, &ti, sizeof(struct tcp_info));
2126
2127 goto out;
2128 }
2129 case MPTCP_SERVICE_TYPE:
2130 optval = mpte->mpte_svctype;
2131 break;
2132 case MPTCP_ALTERNATE_PORT:
2133 optval = mpte->mpte_alternate_port;
2134 break;
2135 case MPTCP_FORCE_ENABLE:
2136 optval = !!(mpte->mpte_flags & MPTE_FORCE_ENABLE);
2137 break;
2138 case MPTCP_FORCE_VERSION:
2139 if (mpte->mpte_flags & MPTE_FORCE_V0) {
2140 optval = 0;
2141 } else if (mpte->mpte_flags & MPTE_FORCE_V1) {
2142 optval = 1;
2143 } else {
2144 optval = -1;
2145 }
2146 break;
2147 case MPTCP_EXPECTED_PROGRESS_TARGET:
2148 error = sooptcopyout(sopt, &mpte->mpte_time_target, sizeof(mpte->mpte_time_target));
2149
2150 goto out;
2151 default:
2152 /* not eligible */
2153 error = ENOPROTOOPT;
2154 break;
2155 }
2156
2157 if (error == 0) {
2158 error = sooptcopyout(sopt, &optval, sizeof(int));
2159 }
2160
2161 out:
2162 mpsotomppcb(mp_so)->mpp_flags &= ~MPP_INSIDE_SETGETOPT;
2163 return error;
2164 }
2165
2166 /*
2167 * MPTCP SOPT_{SET,GET} socket option handler, for options issued on the MP
2168 * socket, at SOL_SOCKET and IPPROTO_TCP levels. The former is restricted
2169 * to those that are allowed by mptcp_usr_socheckopt().
2170 */
2171 int
mptcp_ctloutput(struct socket * mp_so,struct sockopt * sopt)2172 mptcp_ctloutput(struct socket *mp_so, struct sockopt *sopt)
2173 {
2174 struct mppcb *__single mpp = mpsotomppcb(mp_so);
2175 struct mptses *__single mpte;
2176 int error = 0;
2177
2178 if (mpp == NULL || mpp->mpp_state == MPPCB_STATE_DEAD) {
2179 error = EINVAL;
2180 goto out;
2181 }
2182 mpte = mptompte(mpp);
2183 socket_lock_assert_owned(mp_so);
2184
2185 /* we only handle socket and TCP-level socket options for MPTCP */
2186 if (sopt->sopt_level != SOL_SOCKET && sopt->sopt_level != IPPROTO_TCP &&
2187 sopt->sopt_level != IPPROTO_IP && sopt->sopt_level != IPPROTO_IPV6) {
2188 error = EINVAL;
2189 goto out;
2190 }
2191
2192 switch (sopt->sopt_dir) {
2193 case SOPT_SET:
2194 error = mptcp_setopt(mpte, sopt);
2195 break;
2196
2197 case SOPT_GET:
2198 error = mptcp_getopt(mpte, sopt);
2199 break;
2200 }
2201 out:
2202 return error;
2203 }
2204
2205 const char *
mptcp_sopt2str(int level,int optname)2206 mptcp_sopt2str(int level, int optname)
2207 {
2208 switch (level) {
2209 case SOL_SOCKET:
2210 switch (optname) {
2211 case SO_LINGER:
2212 return "SO_LINGER";
2213 case SO_LINGER_SEC:
2214 return "SO_LINGER_SEC";
2215 case SO_DEBUG:
2216 return "SO_DEBUG";
2217 case SO_KEEPALIVE:
2218 return "SO_KEEPALIVE";
2219 case SO_USELOOPBACK:
2220 return "SO_USELOOPBACK";
2221 case SO_TYPE:
2222 return "SO_TYPE";
2223 case SO_NREAD:
2224 return "SO_NREAD";
2225 case SO_NWRITE:
2226 return "SO_NWRITE";
2227 case SO_ERROR:
2228 return "SO_ERROR";
2229 case SO_SNDBUF:
2230 return "SO_SNDBUF";
2231 case SO_RCVBUF:
2232 return "SO_RCVBUF";
2233 case SO_SNDLOWAT:
2234 return "SO_SNDLOWAT";
2235 case SO_RCVLOWAT:
2236 return "SO_RCVLOWAT";
2237 case SO_SNDTIMEO:
2238 return "SO_SNDTIMEO";
2239 case SO_RCVTIMEO:
2240 return "SO_RCVTIMEO";
2241 case SO_NKE:
2242 return "SO_NKE";
2243 case SO_NOSIGPIPE:
2244 return "SO_NOSIGPIPE";
2245 case SO_NOADDRERR:
2246 return "SO_NOADDRERR";
2247 case SO_RESTRICTIONS:
2248 return "SO_RESTRICTIONS";
2249 case SO_LABEL:
2250 return "SO_LABEL";
2251 case SO_PEERLABEL:
2252 return "SO_PEERLABEL";
2253 case SO_RANDOMPORT:
2254 return "SO_RANDOMPORT";
2255 case SO_TRAFFIC_CLASS:
2256 return "SO_TRAFFIC_CLASS";
2257 case SO_RECV_TRAFFIC_CLASS:
2258 return "SO_RECV_TRAFFIC_CLASS";
2259 case SO_TRAFFIC_CLASS_DBG:
2260 return "SO_TRAFFIC_CLASS_DBG";
2261 case SO_PRIVILEGED_TRAFFIC_CLASS:
2262 return "SO_PRIVILEGED_TRAFFIC_CLASS";
2263 case SO_DEFUNCTIT:
2264 return "SO_DEFUNCTIT";
2265 case SO_DEFUNCTOK:
2266 return "SO_DEFUNCTOK";
2267 case SO_ISDEFUNCT:
2268 return "SO_ISDEFUNCT";
2269 case SO_OPPORTUNISTIC:
2270 return "SO_OPPORTUNISTIC";
2271 case SO_FLUSH:
2272 return "SO_FLUSH";
2273 case SO_RECV_ANYIF:
2274 return "SO_RECV_ANYIF";
2275 case SO_NOWAKEFROMSLEEP:
2276 return "SO_NOWAKEFROMSLEEP";
2277 case SO_NOAPNFALLBK:
2278 return "SO_NOAPNFALLBK";
2279 case SO_MARK_CELLFALLBACK:
2280 return "SO_CELLFALLBACK";
2281 case SO_FALLBACK_MODE:
2282 return "SO_FALLBACK_MODE";
2283 case SO_MARK_KNOWN_TRACKER:
2284 return "SO_MARK_KNOWN_TRACKER";
2285 case SO_MARK_KNOWN_TRACKER_NON_APP_INITIATED:
2286 return "SO_MARK_KNOWN_TRACKER_NON_APP_INITIATED";
2287 case SO_MARK_APPROVED_APP_DOMAIN:
2288 return "SO_MARK_APPROVED_APP_DOMAIN";
2289 case SO_DELEGATED:
2290 return "SO_DELEGATED";
2291 case SO_DELEGATED_UUID:
2292 return "SO_DELEGATED_UUID";
2293 #if NECP
2294 case SO_NECP_ATTRIBUTES:
2295 return "SO_NECP_ATTRIBUTES";
2296 case SO_NECP_CLIENTUUID:
2297 return "SO_NECP_CLIENTUUID";
2298 #endif /* NECP */
2299 }
2300
2301 break;
2302 case IPPROTO_IP:
2303 switch (optname) {
2304 case IP_TOS:
2305 return "IP_TOS";
2306 }
2307
2308 break;
2309 case IPPROTO_IPV6:
2310 switch (optname) {
2311 case IPV6_TCLASS:
2312 return "IPV6_TCLASS";
2313 }
2314
2315 break;
2316 case IPPROTO_TCP:
2317 switch (optname) {
2318 case TCP_NODELAY:
2319 return "TCP_NODELAY";
2320 case TCP_KEEPALIVE:
2321 return "TCP_KEEPALIVE";
2322 case TCP_KEEPINTVL:
2323 return "TCP_KEEPINTVL";
2324 case TCP_KEEPCNT:
2325 return "TCP_KEEPCNT";
2326 case TCP_CONNECTIONTIMEOUT:
2327 return "TCP_CONNECTIONTIMEOUT";
2328 case TCP_RXT_CONNDROPTIME:
2329 return "TCP_RXT_CONNDROPTIME";
2330 case PERSIST_TIMEOUT:
2331 return "PERSIST_TIMEOUT";
2332 case TCP_NOTSENT_LOWAT:
2333 return "NOTSENT_LOWAT";
2334 case TCP_ADAPTIVE_READ_TIMEOUT:
2335 return "ADAPTIVE_READ_TIMEOUT";
2336 case TCP_ADAPTIVE_WRITE_TIMEOUT:
2337 return "ADAPTIVE_WRITE_TIMEOUT";
2338 case TCP_FASTOPEN_FORCE_ENABLE:
2339 return "TCP_FASTOPEN_FORCE_ENABLE";
2340 case MPTCP_SERVICE_TYPE:
2341 return "MPTCP_SERVICE_TYPE";
2342 case MPTCP_ALTERNATE_PORT:
2343 return "MPTCP_ALTERNATE_PORT";
2344 case MPTCP_FORCE_ENABLE:
2345 return "MPTCP_FORCE_ENABLE";
2346 case MPTCP_FORCE_VERSION:
2347 return "MPTCP_FORCE_VERSION";
2348 case MPTCP_EXPECTED_PROGRESS_TARGET:
2349 return "MPTCP_EXPECTED_PROGRESS_TARGET";
2350 }
2351
2352 break;
2353 }
2354
2355 return "unknown";
2356 }
2357
2358 static int
mptcp_usr_preconnect(struct socket * mp_so)2359 mptcp_usr_preconnect(struct socket *mp_so)
2360 {
2361 struct mptsub *__single mpts = NULL;
2362 struct mppcb *__single mpp = mpsotomppcb(mp_so);
2363 struct mptses *__single mpte;
2364 struct socket *__single so;
2365 struct tcpcb *__single tp = NULL;
2366 int error;
2367
2368 mpte = mptompte(mpp);
2369
2370 mpts = mptcp_get_subflow(mpte, NULL);
2371 if (mpts == NULL) {
2372 os_log_error(mptcp_log_handle, "%s - %lx: invalid preconnect ",
2373 __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte));
2374 return EINVAL;
2375 }
2376 mpts->mpts_flags &= ~MPTSF_TFO_REQD;
2377 so = mpts->mpts_socket;
2378 tp = intotcpcb(sotoinpcb(so));
2379 tp->t_mpflags &= ~TMPF_TFO_REQUEST;
2380 error = tcp_output(sototcpcb(so));
2381
2382 soclearfastopen(mp_so);
2383
2384 return error;
2385 }
2386