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_stat->rxbytes;
537 mptcp_ci.mptcpci_init_txbytes = inp->inp_stat->txbytes;
538 initial_info_set = 1;
539 }
540
541 mptcpstats_update(mptcp_ci.mptcpci_itfstats, 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_stat->txbytes;
706 tcp_ci.tcpci_tcp_info.tcpi_rxbytes +=
707 mptsinp->inp_stat->rxbytes;
708
709 tcp_ci.tcpci_tcp_info.tcpi_wifi_txbytes +=
710 mptsinp->inp_wstat->txbytes;
711 tcp_ci.tcpci_tcp_info.tcpi_wifi_rxbytes +=
712 mptsinp->inp_wstat->rxbytes;
713
714 tcp_ci.tcpci_tcp_info.tcpi_wired_txbytes +=
715 mptsinp->inp_Wstat->txbytes;
716 tcp_ci.tcpci_tcp_info.tcpi_wired_rxbytes +=
717 mptsinp->inp_Wstat->rxbytes;
718
719 tcp_ci.tcpci_tcp_info.tcpi_cell_txbytes +=
720 mptsinp->inp_cstat->txbytes;
721 tcp_ci.tcpci_tcp_info.tcpi_cell_rxbytes +=
722 mptsinp->inp_cstat->rxbytes;
723 }
724 }
725
726 interface_info:
727 *aux_type = CIAUX_TCP;
728 if (*aux_len == 0) {
729 *aux_len = sizeof(tcp_ci);
730 } else if (aux_data != USER_ADDR_NULL) {
731 boolean_t create;
732
733 /*
734 * Finally, old subflows might have been closed - we
735 * want this data as well, so grab it from the interface
736 * stats.
737 */
738 create = orig_mpts != NULL;
739
740 /*
741 * When we found a subflow, we are willing to create a stats-index
742 * because we have some data to return. If there isn't a subflow,
743 * nor anything in the stats, return EINVAL. Because the
744 * ifindex belongs to something that doesn't exist.
745 */
746 index = mptcpstats_get_index_by_ifindex(mpte->mpte_itfstats, (u_short)(*cid), false);
747 if (index == -1) {
748 os_log_error(mptcp_log_handle,
749 "%s - %lx: Asking for too many ifindex: %u subcount %u, mpts? %s\n",
750 __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte),
751 *cid, mpte->mpte_numflows,
752 orig_mpts ? "yes" : "no");
753
754 if (orig_mpts == NULL) {
755 return EINVAL;
756 }
757 } else {
758 struct mptcp_itf_stats *stats;
759
760 stats = &mpte->mpte_itfstats[index];
761
762 /* Roll the itf-stats into the tcp_info */
763 tcp_ci.tcpci_tcp_info.tcpi_last_outif = *cid;
764 tcp_ci.tcpci_tcp_info.tcpi_txbytes +=
765 stats->mpis_txbytes;
766 tcp_ci.tcpci_tcp_info.tcpi_rxbytes +=
767 stats->mpis_rxbytes;
768
769 tcp_ci.tcpci_tcp_info.tcpi_wifi_txbytes +=
770 stats->mpis_wifi_txbytes;
771 tcp_ci.tcpci_tcp_info.tcpi_wifi_rxbytes +=
772 stats->mpis_wifi_rxbytes;
773
774 tcp_ci.tcpci_tcp_info.tcpi_wired_txbytes +=
775 stats->mpis_wired_txbytes;
776 tcp_ci.tcpci_tcp_info.tcpi_wired_rxbytes +=
777 stats->mpis_wired_rxbytes;
778
779 tcp_ci.tcpci_tcp_info.tcpi_cell_txbytes +=
780 stats->mpis_cell_txbytes;
781 tcp_ci.tcpci_tcp_info.tcpi_cell_rxbytes +=
782 stats->mpis_cell_rxbytes;
783 }
784
785 *aux_len = min(*aux_len, sizeof(tcp_ci));
786 error = copyout(&tcp_ci, aux_data, *aux_len);
787 if (error != 0) {
788 return error;
789 }
790 }
791 }
792
793 return 0;
794 }
795
796 /*
797 * User-protocol pru_control callback.
798 */
799 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)800 mptcp_usr_control(struct socket *mp_so, u_long cmd,
801 caddr_t __sized_by(IOCPARM_LEN(cmd)) data,
802 struct ifnet *ifp, struct proc *p)
803 {
804 #pragma unused(ifp, p)
805 struct mppcb *mpp = mpsotomppcb(mp_so);
806 struct mptses *mpte;
807 int error = 0;
808
809 if (mpp == NULL || mpp->mpp_state == MPPCB_STATE_DEAD) {
810 error = EINVAL;
811 goto out;
812 }
813 mpte = mptompte(mpp);
814
815 switch (cmd) {
816 case SIOCGASSOCIDS32: { /* struct so_aidreq32 */
817 struct so_aidreq32 aidr;
818 bcopy(data, &aidr, sizeof(aidr));
819 error = mptcp_getassocids(mpte, &aidr.sar_cnt,
820 aidr.sar_aidp);
821 if (error == 0) {
822 bcopy(&aidr, data, sizeof(aidr));
823 }
824 break;
825 }
826
827 case SIOCGASSOCIDS64: { /* struct so_aidreq64 */
828 struct so_aidreq64 aidr;
829 bcopy(data, &aidr, sizeof(aidr));
830 error = mptcp_getassocids(mpte, &aidr.sar_cnt,
831 (user_addr_t)aidr.sar_aidp);
832 if (error == 0) {
833 bcopy(&aidr, data, sizeof(aidr));
834 }
835 break;
836 }
837
838 case SIOCGCONNIDS32: { /* struct so_cidreq32 */
839 struct so_cidreq32 cidr;
840 bcopy(data, &cidr, sizeof(cidr));
841 error = mptcp_getconnids(mpte, cidr.scr_aid, &cidr.scr_cnt,
842 cidr.scr_cidp);
843 if (error == 0) {
844 bcopy(&cidr, data, sizeof(cidr));
845 }
846 break;
847 }
848
849 case SIOCGCONNIDS64: { /* struct so_cidreq64 */
850 struct so_cidreq64 cidr;
851 bcopy(data, &cidr, sizeof(cidr));
852 error = mptcp_getconnids(mpte, cidr.scr_aid, &cidr.scr_cnt,
853 (user_addr_t)cidr.scr_cidp);
854 if (error == 0) {
855 bcopy(&cidr, data, sizeof(cidr));
856 }
857 break;
858 }
859
860 case SIOCGCONNINFO32: { /* struct so_cinforeq32 */
861 struct so_cinforeq32 cifr;
862 bcopy(data, &cifr, sizeof(cifr));
863 error = mptcp_getconninfo(mpte, &cifr.scir_cid,
864 &cifr.scir_flags, &cifr.scir_ifindex, &cifr.scir_error,
865 cifr.scir_src, &cifr.scir_src_len, cifr.scir_dst,
866 &cifr.scir_dst_len, &cifr.scir_aux_type, cifr.scir_aux_data,
867 &cifr.scir_aux_len);
868 if (error == 0) {
869 bcopy(&cifr, data, sizeof(cifr));
870 }
871 break;
872 }
873
874 case SIOCGCONNINFO64: { /* struct so_cinforeq64 */
875 struct so_cinforeq64 cifr;
876 bcopy(data, &cifr, sizeof(cifr));
877 error = mptcp_getconninfo(mpte, &cifr.scir_cid,
878 &cifr.scir_flags, &cifr.scir_ifindex, &cifr.scir_error,
879 (user_addr_t)cifr.scir_src, &cifr.scir_src_len,
880 (user_addr_t)cifr.scir_dst, &cifr.scir_dst_len,
881 &cifr.scir_aux_type, (user_addr_t)cifr.scir_aux_data,
882 &cifr.scir_aux_len);
883 if (error == 0) {
884 bcopy(&cifr, data, sizeof(cifr));
885 }
886 break;
887 }
888
889 default:
890 error = EOPNOTSUPP;
891 break;
892 }
893 out:
894 return error;
895 }
896
897 static int
mptcp_disconnect(struct mptses * mpte)898 mptcp_disconnect(struct mptses *mpte)
899 {
900 struct socket *mp_so;
901 struct mptcb *mp_tp;
902 int error = 0;
903
904 mp_so = mptetoso(mpte);
905 mp_tp = mpte->mpte_mptcb;
906
907 /* if we're not detached, go thru socket state checks */
908 if (!(mp_so->so_flags & SOF_PCBCLEARING) && !(mp_so->so_flags & SOF_DEFUNCT)) {
909 if (!(mp_so->so_state & (SS_ISCONNECTED |
910 SS_ISCONNECTING))) {
911 error = ENOTCONN;
912 goto out;
913 }
914 if (mp_so->so_state & SS_ISDISCONNECTING) {
915 error = EALREADY;
916 goto out;
917 }
918 }
919
920 mptcp_cancel_all_timers(mp_tp);
921 if (mp_tp->mpt_state < MPTCPS_ESTABLISHED) {
922 mptcp_close(mpte, mp_tp);
923 } else if (((mp_so->so_options & SO_LINGER) &&
924 mp_so->so_linger == 0) ||
925 (mp_so->so_flags1 & SOF1_DEFUNCTINPROG)) {
926 mptcp_drop(mpte, mp_tp, 0);
927 } else {
928 soisdisconnecting(mp_so);
929 sbflush(&mp_so->so_rcv);
930 if (mptcp_usrclosed(mpte) != NULL) {
931 mptcp_output(mpte);
932 }
933 }
934
935 if (error == 0) {
936 mptcp_subflow_workloop(mpte);
937 }
938
939 out:
940 return error;
941 }
942
943 /*
944 * Wrapper function to support disconnect on socket
945 */
946 static int
mptcp_usr_disconnect(struct socket * mp_so)947 mptcp_usr_disconnect(struct socket *mp_so)
948 {
949 return mptcp_disconnect(mpsotompte(mp_so));
950 }
951
952 /*
953 * User-protocol pru_disconnectx callback.
954 */
955 static int
mptcp_usr_disconnectx(struct socket * mp_so,sae_associd_t aid,sae_connid_t cid)956 mptcp_usr_disconnectx(struct socket *mp_so, sae_associd_t aid, sae_connid_t cid)
957 {
958 if (aid != SAE_ASSOCID_ANY && aid != SAE_ASSOCID_ALL) {
959 return EINVAL;
960 }
961
962 if (cid != SAE_CONNID_ANY && cid != SAE_CONNID_ALL) {
963 return EINVAL;
964 }
965
966 return mptcp_usr_disconnect(mp_so);
967 }
968
969 void
mptcp_finish_usrclosed(struct mptses * mpte)970 mptcp_finish_usrclosed(struct mptses *mpte)
971 {
972 struct mptcb *mp_tp = mpte->mpte_mptcb;
973 struct socket *mp_so = mptetoso(mpte);
974
975 if (mp_tp->mpt_state == MPTCPS_CLOSED || mp_tp->mpt_state == MPTCPS_TERMINATE) {
976 mpte = mptcp_close(mpte, mp_tp);
977 } else if (mp_tp->mpt_state >= MPTCPS_FIN_WAIT_2) {
978 soisdisconnected(mp_so);
979 } else {
980 struct mptsub *mpts;
981
982 TAILQ_FOREACH(mpts, &mpte->mpte_subflows, mpts_entry) {
983 if ((mp_so->so_state & (SS_CANTRCVMORE | SS_CANTSENDMORE)) ==
984 (SS_CANTRCVMORE | SS_CANTSENDMORE)) {
985 mptcp_subflow_disconnect(mpte, mpts);
986 } else {
987 mptcp_subflow_shutdown(mpte, mpts);
988 }
989 }
990 }
991 }
992
993 /*
994 * User issued close, and wish to trail thru shutdown states.
995 */
996 static struct mptses *
mptcp_usrclosed(struct mptses * mpte)997 mptcp_usrclosed(struct mptses *mpte)
998 {
999 struct mptcb *mp_tp = mpte->mpte_mptcb;
1000
1001 mptcp_close_fsm(mp_tp, MPCE_CLOSE);
1002
1003 /* Not everything has been acknowledged - don't close the subflows! */
1004 if (mp_tp->mpt_state != MPTCPS_TERMINATE &&
1005 mp_tp->mpt_sndnxt + 1 != mp_tp->mpt_sndmax) {
1006 return mpte;
1007 }
1008
1009 mptcp_finish_usrclosed(mpte);
1010
1011 return mpte;
1012 }
1013
1014 /*
1015 * After a receive, possible send some update to peer.
1016 */
1017 static int
mptcp_usr_rcvd(struct socket * mp_so,int flags)1018 mptcp_usr_rcvd(struct socket *mp_so, int flags)
1019 {
1020 #pragma unused(flags)
1021 struct mppcb *mpp = mpsotomppcb(mp_so);
1022 struct mptses *mpte;
1023 struct mptsub *mpts;
1024 int error = 0;
1025
1026 if (mpp == NULL || mpp->mpp_state == MPPCB_STATE_DEAD) {
1027 error = EINVAL;
1028 goto out;
1029 }
1030
1031 mpte = mptompte(mpp);
1032
1033 TAILQ_FOREACH(mpts, &mpte->mpte_subflows, mpts_entry) {
1034 struct socket *so = mpts->mpts_socket;
1035
1036 if (so->so_proto->pr_flags & PR_WANTRCVD && so->so_pcb != NULL) {
1037 (*so->so_proto->pr_usrreqs->pru_rcvd)(so, 0);
1038 }
1039 }
1040
1041 error = mptcp_output(mpte);
1042 out:
1043 return error;
1044 }
1045
1046 /*
1047 * Do a send by putting data in the output queue.
1048 */
1049 static int
mptcp_usr_send(struct socket * mp_so,int prus_flags,struct mbuf * m,struct sockaddr * nam,struct mbuf * control,struct proc * p)1050 mptcp_usr_send(struct socket *mp_so, int prus_flags, struct mbuf *m,
1051 struct sockaddr *nam, struct mbuf *control, struct proc *p)
1052 {
1053 #pragma unused(nam, p)
1054 struct mppcb *mpp = mpsotomppcb(mp_so);
1055 struct mptses *mpte;
1056 int error = 0;
1057
1058 if (prus_flags & (PRUS_OOB | PRUS_EOF)) {
1059 error = EOPNOTSUPP;
1060 goto out;
1061 }
1062
1063 if (nam != NULL) {
1064 error = EOPNOTSUPP;
1065 goto out;
1066 }
1067
1068 if (control != NULL && control->m_len != 0) {
1069 error = EOPNOTSUPP;
1070 goto out;
1071 }
1072
1073 if (mpp == NULL || mpp->mpp_state == MPPCB_STATE_DEAD) {
1074 error = ECONNRESET;
1075 goto out;
1076 }
1077 mpte = mptompte(mpp);
1078 VERIFY(mpte != NULL);
1079
1080 if (!(mp_so->so_state & SS_ISCONNECTED) &&
1081 !(mp_so->so_flags1 & SOF1_PRECONNECT_DATA)) {
1082 error = ENOTCONN;
1083 goto out;
1084 }
1085
1086 mptcp_insert_dsn(mpp, m);
1087 VERIFY(mp_so->so_snd.sb_flags & SB_NOCOMPRESS);
1088 sbappendstream(&mp_so->so_snd, m);
1089 m = NULL;
1090
1091 error = mptcp_output(mpte);
1092 if (error != 0) {
1093 goto out;
1094 }
1095
1096 if (mp_so->so_state & SS_ISCONNECTING) {
1097 if (mp_so->so_state & SS_NBIO) {
1098 error = EWOULDBLOCK;
1099 } else {
1100 error = sbwait(&mp_so->so_snd);
1101 }
1102 }
1103
1104 out:
1105 if (error) {
1106 if (m != NULL) {
1107 m_freem(m);
1108 }
1109 if (control != NULL) {
1110 m_freem(control);
1111 }
1112 }
1113 return error;
1114 }
1115
1116 /*
1117 * Mark the MPTCP connection as being incapable of further output.
1118 */
1119 static int
mptcp_usr_shutdown(struct socket * mp_so)1120 mptcp_usr_shutdown(struct socket *mp_so)
1121 {
1122 struct mppcb *mpp = mpsotomppcb(mp_so);
1123 struct mptses *mpte;
1124 int error = 0;
1125
1126 if (mpp == NULL || mpp->mpp_state == MPPCB_STATE_DEAD) {
1127 error = EINVAL;
1128 goto out;
1129 }
1130 mpte = mptompte(mpp);
1131 VERIFY(mpte != NULL);
1132
1133 socantsendmore(mp_so);
1134
1135 mpte = mptcp_usrclosed(mpte);
1136 if (mpte != NULL) {
1137 error = mptcp_output(mpte);
1138 }
1139 out:
1140 return error;
1141 }
1142
1143 /*
1144 * Copy the contents of uio into a properly sized mbuf chain.
1145 */
1146 static int
mptcp_uiotombuf(struct uio * uio,int how,user_ssize_t space,struct mbuf ** top)1147 mptcp_uiotombuf(struct uio *uio, int how, user_ssize_t space, struct mbuf **top)
1148 {
1149 struct mbuf *m, *mb, *nm = NULL, *mtail = NULL;
1150 int progress, len, error;
1151 user_ssize_t resid, tot;
1152
1153 VERIFY(top != NULL && *top == NULL);
1154
1155 /*
1156 * space can be zero or an arbitrary large value bound by
1157 * the total data supplied by the uio.
1158 */
1159 resid = uio_resid(uio);
1160 if (space > 0) {
1161 tot = MIN(resid, space);
1162 } else {
1163 tot = resid;
1164 }
1165
1166 if (tot < 0 || tot > INT_MAX) {
1167 return EINVAL;
1168 }
1169
1170 len = (int)tot;
1171 if (len == 0) {
1172 len = 1;
1173 }
1174
1175 /* Loop and append maximum sized mbufs to the chain tail. */
1176 while (len > 0) {
1177 uint32_t m_needed = 1;
1178
1179 if (njcl > 0 && len > MBIGCLBYTES) {
1180 mb = m_getpackets_internal(&m_needed, 1,
1181 how, 1, M16KCLBYTES);
1182 } else if (len > MCLBYTES) {
1183 mb = m_getpackets_internal(&m_needed, 1,
1184 how, 1, MBIGCLBYTES);
1185 } else if (len >= (signed)MINCLSIZE) {
1186 mb = m_getpackets_internal(&m_needed, 1,
1187 how, 1, MCLBYTES);
1188 } else {
1189 mb = m_gethdr(how, MT_DATA);
1190 }
1191
1192 /* Fail the whole operation if one mbuf can't be allocated. */
1193 if (mb == NULL) {
1194 if (nm != NULL) {
1195 m_freem(nm);
1196 }
1197 return ENOBUFS;
1198 }
1199
1200 /* Book keeping. */
1201 VERIFY(mb->m_flags & M_PKTHDR);
1202 len -= ((mb->m_flags & M_EXT) ? mb->m_ext.ext_size : MHLEN);
1203 if (mtail != NULL) {
1204 mtail->m_next = mb;
1205 } else {
1206 nm = mb;
1207 }
1208 mtail = mb;
1209 }
1210
1211 m = nm;
1212
1213 progress = 0;
1214 /* Fill all mbufs with uio data and update header information. */
1215 for (mb = m; mb != NULL; mb = mb->m_next) {
1216 /* tot >= 0 && tot <= INT_MAX (see above) */
1217 len = MIN((int)M_TRAILINGSPACE(mb), (int)(tot - progress));
1218
1219 error = uiomove(mtod(mb, char *), len, uio);
1220 if (error != 0) {
1221 m_freem(m);
1222 return error;
1223 }
1224
1225 /* each mbuf is M_PKTHDR chained via m_next */
1226 mb->m_len = len;
1227 mb->m_pkthdr.len = len;
1228
1229 progress += len;
1230 }
1231 VERIFY(progress == tot);
1232 *top = m;
1233 return 0;
1234 }
1235
1236 /*
1237 * MPTCP socket protocol-user socket send routine, derived from sosend().
1238 */
1239 static int
mptcp_usr_sosend(struct socket * mp_so,struct sockaddr * addr,struct uio * uio,struct mbuf * top,struct mbuf * control,int flags)1240 mptcp_usr_sosend(struct socket *mp_so, struct sockaddr *addr, struct uio *uio,
1241 struct mbuf *top, struct mbuf *control, int flags)
1242 {
1243 #pragma unused(addr)
1244 user_ssize_t resid, space;
1245 int error, sendflags;
1246 struct proc *p = current_proc();
1247 int sblocked = 0;
1248
1249 /* UIO is required for now, due to per-mbuf M_PKTHDR constrains */
1250 if (uio == NULL || top != NULL) {
1251 error = EINVAL;
1252 goto out;
1253 }
1254 resid = uio_resid(uio);
1255
1256 socket_lock(mp_so, 1);
1257 so_update_last_owner_locked(mp_so, p);
1258 so_update_policy(mp_so);
1259
1260 VERIFY(mp_so->so_type == SOCK_STREAM);
1261 VERIFY(!(mp_so->so_flags & SOF_MP_SUBFLOW));
1262
1263 if (flags & (MSG_OOB | MSG_DONTROUTE)) {
1264 error = EOPNOTSUPP;
1265 socket_unlock(mp_so, 1);
1266 goto out;
1267 }
1268
1269 /*
1270 * In theory resid should be unsigned. However, space must be
1271 * signed, as it might be less than 0 if we over-committed, and we
1272 * must use a signed comparison of space and resid. On the other
1273 * hand, a negative resid causes us to loop sending 0-length
1274 * segments to the protocol.
1275 */
1276 if (resid < 0 || resid > INT_MAX ||
1277 (flags & MSG_EOR) || control != NULL) {
1278 error = EINVAL;
1279 socket_unlock(mp_so, 1);
1280 goto out;
1281 }
1282
1283 OSIncrementAtomicLong(&p->p_stats->p_ru.ru_msgsnd);
1284
1285 do {
1286 error = sosendcheck(mp_so, NULL, resid, 0, 0, flags,
1287 &sblocked);
1288 if (error != 0) {
1289 goto release;
1290 }
1291
1292 space = sbspace(&mp_so->so_snd);
1293 do {
1294 socket_unlock(mp_so, 0);
1295 /*
1296 * Copy the data from userland into an mbuf chain.
1297 */
1298 error = mptcp_uiotombuf(uio, M_WAITOK, space, &top);
1299 if (error != 0) {
1300 socket_lock(mp_so, 0);
1301 goto release;
1302 }
1303 VERIFY(top != NULL);
1304 space -= resid - uio_resid(uio);
1305 resid = uio_resid(uio);
1306 socket_lock(mp_so, 0);
1307
1308 /*
1309 * Compute flags here, for pru_send and NKEs.
1310 */
1311 sendflags = (resid > 0 && space > 0) ?
1312 PRUS_MORETOCOME : 0;
1313
1314 /*
1315 * Socket filter processing
1316 */
1317 VERIFY(control == NULL);
1318 error = sflt_data_out(mp_so, NULL, &top, &control, 0);
1319 if (error != 0) {
1320 if (error == EJUSTRETURN) {
1321 error = 0;
1322 top = NULL;
1323 /* always free control if any */
1324 }
1325 goto release;
1326 }
1327 if (control != NULL) {
1328 m_freem(control);
1329 control = NULL;
1330 }
1331
1332 /*
1333 * Pass data to protocol.
1334 */
1335 error = (*mp_so->so_proto->pr_usrreqs->pru_send)
1336 (mp_so, sendflags, top, NULL, NULL, p);
1337
1338 top = NULL;
1339 if (error != 0) {
1340 goto release;
1341 }
1342 } while (resid != 0 && space > 0);
1343 } while (resid != 0);
1344
1345 release:
1346 if (sblocked) {
1347 sbunlock(&mp_so->so_snd, FALSE); /* will unlock socket */
1348 } else {
1349 socket_unlock(mp_so, 1);
1350 }
1351 out:
1352 if (top != NULL) {
1353 m_freem(top);
1354 }
1355 if (control != NULL) {
1356 m_freem(control);
1357 }
1358
1359 soclearfastopen(mp_so);
1360
1361 return error;
1362 }
1363
1364 /*
1365 * Called to filter SOPT_{SET,GET} for SOL_SOCKET level socket options.
1366 * This routine simply indicates to the caller whether or not to proceed
1367 * further with the given socket option. This is invoked by sosetoptlock()
1368 * and sogetoptlock().
1369 */
1370 static int
mptcp_usr_socheckopt(struct socket * mp_so,struct sockopt * sopt)1371 mptcp_usr_socheckopt(struct socket *mp_so, struct sockopt *sopt)
1372 {
1373 #pragma unused(mp_so)
1374 int error = 0;
1375
1376 VERIFY(sopt->sopt_level == SOL_SOCKET);
1377
1378 /*
1379 * We could check for sopt_dir (set/get) here, but we'll just
1380 * let the caller deal with it as appropriate; therefore the
1381 * following is a superset of the socket options which we
1382 * allow for set/get.
1383 *
1384 * XXX: [email protected]
1385 *
1386 * Need to consider the following cases:
1387 *
1388 * a. Certain socket options don't have a clear definition
1389 * on the expected behavior post connect(2). At the time
1390 * those options are issued on the MP socket, there may
1391 * be existing subflow sockets that are already connected.
1392 */
1393 switch (sopt->sopt_name) {
1394 case SO_LINGER: /* MP */
1395 case SO_LINGER_SEC: /* MP */
1396 case SO_TYPE: /* MP */
1397 case SO_NREAD: /* MP */
1398 case SO_NWRITE: /* MP */
1399 case SO_ERROR: /* MP */
1400 case SO_SNDBUF: /* MP */
1401 case SO_RCVBUF: /* MP */
1402 case SO_SNDLOWAT: /* MP */
1403 case SO_RCVLOWAT: /* MP */
1404 case SO_SNDTIMEO: /* MP */
1405 case SO_RCVTIMEO: /* MP */
1406 case SO_NKE: /* MP */
1407 case SO_NOSIGPIPE: /* MP */
1408 case SO_NOADDRERR: /* MP */
1409 case SO_LABEL: /* MP */
1410 case SO_PEERLABEL: /* MP */
1411 case SO_DEFUNCTIT: /* MP */
1412 case SO_DEFUNCTOK: /* MP */
1413 case SO_ISDEFUNCT: /* MP */
1414 case SO_TRAFFIC_CLASS_DBG: /* MP */
1415 case SO_DELEGATED: /* MP */
1416 case SO_DELEGATED_UUID: /* MP */
1417 #if NECP
1418 case SO_NECP_ATTRIBUTES:
1419 case SO_NECP_CLIENTUUID:
1420 #endif /* NECP */
1421 case SO_MPKL_SEND_INFO:
1422 /*
1423 * Tell the caller that these options are to be processed.
1424 */
1425 break;
1426
1427 case SO_DEBUG: /* MP + subflow */
1428 case SO_KEEPALIVE: /* MP + subflow */
1429 case SO_USELOOPBACK: /* MP + subflow */
1430 case SO_RANDOMPORT: /* MP + subflow */
1431 case SO_TRAFFIC_CLASS: /* MP + subflow */
1432 case SO_RECV_TRAFFIC_CLASS: /* MP + subflow */
1433 case SO_PRIVILEGED_TRAFFIC_CLASS: /* MP + subflow */
1434 case SO_RECV_ANYIF: /* MP + subflow */
1435 case SO_RESTRICTIONS: /* MP + subflow */
1436 case SO_FLUSH: /* MP + subflow */
1437 case SO_NOWAKEFROMSLEEP:
1438 case SO_NOAPNFALLBK:
1439 case SO_MARK_CELLFALLBACK:
1440 case SO_MARK_CELLFALLBACK_UUID:
1441 case SO_MARK_KNOWN_TRACKER:
1442 case SO_MARK_KNOWN_TRACKER_NON_APP_INITIATED:
1443 case SO_MARK_APPROVED_APP_DOMAIN:
1444 case SO_FALLBACK_MODE:
1445 /*
1446 * Tell the caller that these options are to be processed;
1447 * these will also be recorded later by mptcp_setopt().
1448 *
1449 * NOTE: Only support integer option value for now.
1450 */
1451 if (sopt->sopt_valsize != sizeof(int)) {
1452 error = EINVAL;
1453 }
1454 break;
1455
1456 default:
1457 /*
1458 * Tell the caller to stop immediately and return an error.
1459 */
1460 error = ENOPROTOOPT;
1461 break;
1462 }
1463
1464 return error;
1465 }
1466
1467 /*
1468 * Issue SOPT_SET for all MPTCP subflows (for integer option values.)
1469 */
1470 static int
mptcp_setopt_apply(struct mptses * mpte,struct mptopt * mpo)1471 mptcp_setopt_apply(struct mptses *mpte, struct mptopt *mpo)
1472 {
1473 struct socket *mp_so;
1474 struct mptsub *mpts;
1475 struct mptopt smpo;
1476 int error = 0;
1477
1478 /* just bail now if this isn't applicable to subflow sockets */
1479 if (!(mpo->mpo_flags & MPOF_SUBFLOW_OK)) {
1480 error = ENOPROTOOPT;
1481 goto out;
1482 }
1483
1484 /*
1485 * Skip those that are handled internally; these options
1486 * should not have been recorded and marked with the
1487 * MPOF_SUBFLOW_OK by mptcp_setopt(), but just in case.
1488 */
1489 if (mpo->mpo_level == SOL_SOCKET &&
1490 (mpo->mpo_name == SO_NOSIGPIPE || mpo->mpo_name == SO_NOADDRERR)) {
1491 error = ENOPROTOOPT;
1492 goto out;
1493 }
1494
1495 mp_so = mptetoso(mpte);
1496
1497 /*
1498 * Don't bother going further if there's no subflow; mark the option
1499 * with MPOF_INTERIM so that we know whether or not to remove this
1500 * option upon encountering an error while issuing it during subflow
1501 * socket creation.
1502 */
1503 if (mpte->mpte_numflows == 0) {
1504 VERIFY(TAILQ_EMPTY(&mpte->mpte_subflows));
1505 mpo->mpo_flags |= MPOF_INTERIM;
1506 /* return success */
1507 goto out;
1508 }
1509
1510 bzero(&smpo, sizeof(smpo));
1511 smpo.mpo_flags |= MPOF_SUBFLOW_OK;
1512 smpo.mpo_level = mpo->mpo_level;
1513 smpo.mpo_name = mpo->mpo_name;
1514
1515 /* grab exisiting values in case we need to rollback */
1516 TAILQ_FOREACH(mpts, &mpte->mpte_subflows, mpts_entry) {
1517 struct socket *so;
1518
1519 mpts->mpts_flags &= ~(MPTSF_SOPT_OLDVAL | MPTSF_SOPT_INPROG);
1520 mpts->mpts_oldintval = 0;
1521 smpo.mpo_intval = 0;
1522 VERIFY(mpts->mpts_socket != NULL);
1523 so = mpts->mpts_socket;
1524 if (mptcp_subflow_sogetopt(mpte, so, &smpo) == 0) {
1525 mpts->mpts_flags |= MPTSF_SOPT_OLDVAL;
1526 mpts->mpts_oldintval = smpo.mpo_intval;
1527 }
1528 }
1529
1530 /* apply socket option */
1531 TAILQ_FOREACH(mpts, &mpte->mpte_subflows, mpts_entry) {
1532 struct socket *so;
1533
1534 mpts->mpts_flags |= MPTSF_SOPT_INPROG;
1535 VERIFY(mpts->mpts_socket != NULL);
1536 so = mpts->mpts_socket;
1537 error = mptcp_subflow_sosetopt(mpte, mpts, mpo);
1538 if (error != 0) {
1539 break;
1540 }
1541 }
1542
1543 /* cleanup, and rollback if needed */
1544 TAILQ_FOREACH(mpts, &mpte->mpte_subflows, mpts_entry) {
1545 struct socket *so;
1546
1547 if (!(mpts->mpts_flags & MPTSF_SOPT_INPROG)) {
1548 /* clear in case it's set */
1549 mpts->mpts_flags &= ~MPTSF_SOPT_OLDVAL;
1550 mpts->mpts_oldintval = 0;
1551 continue;
1552 }
1553 if (!(mpts->mpts_flags & MPTSF_SOPT_OLDVAL)) {
1554 mpts->mpts_flags &= ~MPTSF_SOPT_INPROG;
1555 VERIFY(mpts->mpts_oldintval == 0);
1556 continue;
1557 }
1558 /* error during sosetopt, so roll it back */
1559 if (error != 0) {
1560 VERIFY(mpts->mpts_socket != NULL);
1561 so = mpts->mpts_socket;
1562 smpo.mpo_intval = mpts->mpts_oldintval;
1563 mptcp_subflow_sosetopt(mpte, mpts, &smpo);
1564 }
1565 mpts->mpts_oldintval = 0;
1566 mpts->mpts_flags &= ~(MPTSF_SOPT_OLDVAL | MPTSF_SOPT_INPROG);
1567 }
1568
1569 out:
1570 return error;
1571 }
1572
1573 /*
1574 * Handle SOPT_SET for socket options issued on MP socket.
1575 */
1576 static int
mptcp_setopt(struct mptses * mpte,struct sockopt * sopt)1577 mptcp_setopt(struct mptses *mpte, struct sockopt *sopt)
1578 {
1579 int error = 0, optval = 0, level, optname, rec = 1;
1580 struct mptopt smpo, *mpo = NULL;
1581 struct socket *mp_so;
1582
1583 level = sopt->sopt_level;
1584 optname = sopt->sopt_name;
1585
1586 mp_so = mptetoso(mpte);
1587
1588 VERIFY(!(mpsotomppcb(mp_so)->mpp_flags & MPP_INSIDE_SETGETOPT));
1589 mpsotomppcb(mp_so)->mpp_flags |= MPP_INSIDE_SETGETOPT;
1590
1591 /*
1592 * Record socket options which are applicable to subflow sockets so
1593 * that we can replay them for new ones; see mptcp_usr_socheckopt()
1594 * for the list of eligible socket-level options.
1595 */
1596 if (level == SOL_SOCKET) {
1597 switch (optname) {
1598 case SO_DEBUG:
1599 case SO_KEEPALIVE:
1600 case SO_USELOOPBACK:
1601 case SO_RANDOMPORT:
1602 case SO_TRAFFIC_CLASS:
1603 case SO_RECV_TRAFFIC_CLASS:
1604 case SO_PRIVILEGED_TRAFFIC_CLASS:
1605 case SO_RECV_ANYIF:
1606 case SO_RESTRICTIONS:
1607 case SO_NOWAKEFROMSLEEP:
1608 case SO_NOAPNFALLBK:
1609 case SO_MARK_CELLFALLBACK:
1610 case SO_MARK_KNOWN_TRACKER:
1611 case SO_MARK_KNOWN_TRACKER_NON_APP_INITIATED:
1612 case SO_MARK_APPROVED_APP_DOMAIN:
1613 case SO_FALLBACK_MODE:
1614 /* record it */
1615 break;
1616 case SO_FLUSH:
1617 /* don't record it */
1618 rec = 0;
1619 break;
1620
1621 /* Next ones, record at MPTCP-level */
1622 case SO_DELEGATED:
1623 error = sooptcopyin(sopt, &mpte->mpte_epid,
1624 sizeof(int), sizeof(int));
1625 if (error != 0) {
1626 goto err_out;
1627 }
1628
1629 goto out;
1630 case SO_DELEGATED_UUID:
1631 error = sooptcopyin(sopt, &mpte->mpte_euuid,
1632 sizeof(uuid_t), sizeof(uuid_t));
1633 if (error != 0) {
1634 goto err_out;
1635 }
1636
1637 goto out;
1638 #if NECP
1639 case SO_NECP_CLIENTUUID:
1640 if (!uuid_is_null(mpsotomppcb(mp_so)->necp_client_uuid)) {
1641 error = EINVAL;
1642 goto err_out;
1643 }
1644
1645 error = sooptcopyin(sopt, &mpsotomppcb(mp_so)->necp_client_uuid,
1646 sizeof(uuid_t), sizeof(uuid_t));
1647 if (error != 0) {
1648 goto err_out;
1649 }
1650
1651 mpsotomppcb(mp_so)->necp_cb = mptcp_session_necp_cb;
1652 error = necp_client_register_multipath_cb(mp_so->last_pid,
1653 mpsotomppcb(mp_so)->necp_client_uuid,
1654 mpsotomppcb(mp_so));
1655 if (error) {
1656 goto err_out;
1657 }
1658
1659 if (uuid_is_null(mpsotomppcb(mp_so)->necp_client_uuid)) {
1660 error = EINVAL;
1661 goto err_out;
1662 }
1663
1664 goto out;
1665 case SO_NECP_ATTRIBUTES:
1666 error = necp_set_socket_attributes(&mpsotomppcb(mp_so)->inp_necp_attributes, sopt);
1667 if (error) {
1668 goto err_out;
1669 }
1670
1671 goto out;
1672 #endif /* NECP */
1673 default:
1674 /* nothing to do; just return */
1675 goto out;
1676 }
1677 } else if (sopt->sopt_level == IPPROTO_IP) {
1678 switch (optname) {
1679 case IP_TOS:
1680 /* eligible; record it */
1681 break;
1682 default:
1683 /* not eligible */
1684 error = ENOPROTOOPT;
1685 goto err_out;
1686 }
1687 } else if (sopt->sopt_level == IPPROTO_IPV6) {
1688 switch (optname) {
1689 case IPV6_TCLASS:
1690 /* eligible; record it */
1691 break;
1692 default:
1693 /* not eligible */
1694 error = ENOPROTOOPT;
1695 goto err_out;
1696 }
1697 } else {
1698 switch (optname) {
1699 case TCP_NODELAY:
1700 case TCP_RXT_FINDROP:
1701 case TCP_KEEPALIVE:
1702 case TCP_KEEPINTVL:
1703 case TCP_KEEPCNT:
1704 case TCP_CONNECTIONTIMEOUT:
1705 case TCP_RXT_CONNDROPTIME:
1706 case PERSIST_TIMEOUT:
1707 case TCP_ADAPTIVE_READ_TIMEOUT:
1708 case TCP_ADAPTIVE_WRITE_TIMEOUT:
1709 case TCP_FASTOPEN_FORCE_ENABLE:
1710 /* eligible; record it */
1711 break;
1712 case TCP_NOTSENT_LOWAT:
1713 /* record at MPTCP level */
1714 error = sooptcopyin(sopt, &optval, sizeof(optval),
1715 sizeof(optval));
1716 if (error) {
1717 goto err_out;
1718 }
1719 if (optval < 0) {
1720 error = EINVAL;
1721 goto err_out;
1722 } else {
1723 if (optval == 0) {
1724 mp_so->so_flags &= ~SOF_NOTSENT_LOWAT;
1725 error = mptcp_set_notsent_lowat(mpte, 0);
1726 } else {
1727 mp_so->so_flags |= SOF_NOTSENT_LOWAT;
1728 error = mptcp_set_notsent_lowat(mpte,
1729 optval);
1730 }
1731
1732 if (error) {
1733 goto err_out;
1734 }
1735 }
1736 goto out;
1737 case MPTCP_SERVICE_TYPE:
1738 /* record at MPTCP level */
1739 error = sooptcopyin(sopt, &optval, sizeof(optval),
1740 sizeof(optval));
1741 if (error) {
1742 goto err_out;
1743 }
1744 if (optval < 0 || optval >= MPTCP_SVCTYPE_MAX) {
1745 error = EINVAL;
1746 goto err_out;
1747 }
1748
1749 if (mptcp_entitlement_check(mp_so, (uint8_t)optval) < 0) {
1750 error = EACCES;
1751 goto err_out;
1752 }
1753
1754 mpte->mpte_svctype = (uint8_t)optval;
1755 mpte->mpte_flags |= MPTE_SVCTYPE_CHECKED;
1756
1757 goto out;
1758 case MPTCP_ALTERNATE_PORT:
1759 /* record at MPTCP level */
1760 error = sooptcopyin(sopt, &optval, sizeof(optval),
1761 sizeof(optval));
1762 if (error) {
1763 goto err_out;
1764 }
1765
1766 if (optval < 0 || optval > UINT16_MAX) {
1767 error = EINVAL;
1768 goto err_out;
1769 }
1770
1771 mpte->mpte_alternate_port = (uint16_t)optval;
1772
1773 goto out;
1774 case MPTCP_FORCE_ENABLE:
1775 /* record at MPTCP level */
1776 error = sooptcopyin(sopt, &optval, sizeof(optval),
1777 sizeof(optval));
1778 if (error) {
1779 goto err_out;
1780 }
1781
1782 if (optval < 0 || optval > 1) {
1783 error = EINVAL;
1784 goto err_out;
1785 }
1786
1787 if (optval) {
1788 mpte->mpte_flags |= MPTE_FORCE_ENABLE;
1789 } else {
1790 mpte->mpte_flags &= ~MPTE_FORCE_ENABLE;
1791 }
1792
1793 goto out;
1794 case MPTCP_FORCE_VERSION:
1795 error = sooptcopyin(sopt, &optval, sizeof(optval),
1796 sizeof(optval));
1797 if (error) {
1798 goto err_out;
1799 }
1800
1801 if (optval != 0 && optval != 1) {
1802 error = EINVAL;
1803 goto err_out;
1804 }
1805
1806 if (optval == 0) {
1807 mpte->mpte_flags |= MPTE_FORCE_V0;
1808 mpte->mpte_flags &= ~MPTE_FORCE_V1;
1809 } else {
1810 mpte->mpte_flags |= MPTE_FORCE_V1;
1811 mpte->mpte_flags &= ~MPTE_FORCE_V0;
1812 }
1813
1814 goto out;
1815 case MPTCP_EXPECTED_PROGRESS_TARGET:
1816 {
1817 struct mptcb *mp_tp = mpte->mpte_mptcb;
1818 uint64_t mach_time_target;
1819 uint64_t nanoseconds;
1820
1821 if (mpte->mpte_svctype != MPTCP_SVCTYPE_TARGET_BASED) {
1822 os_log(mptcp_log_handle, "%s - %lx: Can't set urgent activity when svctype is %u\n",
1823 __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte), mpte->mpte_svctype);
1824 error = EINVAL;
1825 goto err_out;
1826 }
1827
1828 error = sooptcopyin(sopt, &mach_time_target, sizeof(mach_time_target), sizeof(mach_time_target));
1829 if (error) {
1830 goto err_out;
1831 }
1832
1833 if (!mptcp_ok_to_create_subflows(mp_tp)) {
1834 os_log(mptcp_log_handle, "%s - %lx: Not ok to create subflows, state %u flags %#x\n",
1835 __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte), mp_tp->mpt_state, mp_tp->mpt_flags);
1836 error = EINVAL;
1837 goto err_out;
1838 }
1839
1840 if (mach_time_target) {
1841 uint64_t time_now = 0;
1842 uint64_t time_now_nanoseconds;
1843
1844 absolutetime_to_nanoseconds(mach_time_target, &nanoseconds);
1845 nanoseconds = nanoseconds - (mptcp_expected_progress_headstart * NSEC_PER_MSEC);
1846
1847 time_now = mach_continuous_time();
1848 absolutetime_to_nanoseconds(time_now, &time_now_nanoseconds);
1849
1850 nanoseconds_to_absolutetime(nanoseconds, &mach_time_target);
1851 /* If the timer is already running and it would
1852 * fire in less than mptcp_expected_progress_headstart
1853 * seconds, then it's not worth canceling it.
1854 */
1855 if (mpte->mpte_time_target &&
1856 mpte->mpte_time_target < time_now &&
1857 time_now_nanoseconds > nanoseconds - (mptcp_expected_progress_headstart * NSEC_PER_MSEC)) {
1858 os_log(mptcp_log_handle, "%s - %lx: Not rescheduling timer %llu now %llu target %llu\n",
1859 __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte),
1860 mpte->mpte_time_target,
1861 time_now,
1862 mach_time_target);
1863 goto out;
1864 }
1865 }
1866
1867 mpte->mpte_time_target = mach_time_target;
1868 mptcp_set_urgency_timer(mpte);
1869
1870 goto out;
1871 }
1872 default:
1873 /* not eligible */
1874 error = ENOPROTOOPT;
1875 goto err_out;
1876 }
1877 }
1878
1879 if ((error = sooptcopyin(sopt, &optval, sizeof(optval),
1880 sizeof(optval))) != 0) {
1881 goto err_out;
1882 }
1883
1884 if (rec) {
1885 /* search for an existing one; if not found, allocate */
1886 if ((mpo = mptcp_sopt_find(mpte, sopt)) == NULL) {
1887 mpo = mptcp_sopt_alloc();
1888 }
1889
1890 /* initialize or update, as needed */
1891 mpo->mpo_intval = optval;
1892 if (!(mpo->mpo_flags & MPOF_ATTACHED)) {
1893 mpo->mpo_level = level;
1894 mpo->mpo_name = optname;
1895 mptcp_sopt_insert(mpte, mpo);
1896 }
1897 /* this can be issued on the subflow socket */
1898 mpo->mpo_flags |= MPOF_SUBFLOW_OK;
1899 } else {
1900 bzero(&smpo, sizeof(smpo));
1901 mpo = &smpo;
1902 mpo->mpo_flags |= MPOF_SUBFLOW_OK;
1903 mpo->mpo_level = level;
1904 mpo->mpo_name = optname;
1905 mpo->mpo_intval = optval;
1906 }
1907
1908 /* issue this socket option on existing subflows */
1909 error = mptcp_setopt_apply(mpte, mpo);
1910 if (error != 0 && (mpo->mpo_flags & MPOF_ATTACHED)) {
1911 VERIFY(mpo != &smpo);
1912 mptcp_sopt_remove(mpte, mpo);
1913 mptcp_sopt_free(mpo);
1914 }
1915 if (mpo == &smpo) {
1916 mpo->mpo_flags &= ~MPOF_INTERIM;
1917 }
1918
1919 if (error) {
1920 goto err_out;
1921 }
1922
1923 out:
1924
1925 mpsotomppcb(mp_so)->mpp_flags &= ~MPP_INSIDE_SETGETOPT;
1926 return 0;
1927
1928 err_out:
1929 os_log_error(mptcp_log_handle, "%s - %lx: sopt %s (%d, %d) val %d can't be issued error %d\n",
1930 __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte),
1931 mptcp_sopt2str(level, optname), level, optname, optval, error);
1932 mpsotomppcb(mp_so)->mpp_flags &= ~MPP_INSIDE_SETGETOPT;
1933 return error;
1934 }
1935
1936 static void
mptcp_fill_info_bytestats(struct tcp_info * ti,struct mptses * mpte)1937 mptcp_fill_info_bytestats(struct tcp_info *ti, struct mptses *mpte)
1938 {
1939 struct mptsub *mpts;
1940 int i;
1941
1942 TAILQ_FOREACH(mpts, &mpte->mpte_subflows, mpts_entry) {
1943 const struct inpcb *inp = sotoinpcb(mpts->mpts_socket);
1944
1945 if (inp == NULL) {
1946 continue;
1947 }
1948
1949 ti->tcpi_txbytes += inp->inp_stat->txbytes;
1950 ti->tcpi_rxbytes += inp->inp_stat->rxbytes;
1951 ti->tcpi_cell_txbytes += inp->inp_cstat->txbytes;
1952 ti->tcpi_cell_rxbytes += inp->inp_cstat->rxbytes;
1953 ti->tcpi_wifi_txbytes += inp->inp_wstat->txbytes;
1954 ti->tcpi_wifi_rxbytes += inp->inp_wstat->rxbytes;
1955 ti->tcpi_wired_txbytes += inp->inp_Wstat->txbytes;
1956 ti->tcpi_wired_rxbytes += inp->inp_Wstat->rxbytes;
1957 }
1958
1959 for (i = 0; i < MPTCP_ITFSTATS_SIZE; i++) {
1960 struct mptcp_itf_stats *stats = &mpte->mpte_itfstats[i];
1961
1962 ti->tcpi_txbytes += stats->mpis_txbytes;
1963 ti->tcpi_rxbytes += stats->mpis_rxbytes;
1964
1965 ti->tcpi_wifi_txbytes += stats->mpis_wifi_txbytes;
1966 ti->tcpi_wifi_rxbytes += stats->mpis_wifi_rxbytes;
1967
1968 ti->tcpi_wired_txbytes += stats->mpis_wired_txbytes;
1969 ti->tcpi_wired_rxbytes += stats->mpis_wired_rxbytes;
1970
1971 ti->tcpi_cell_txbytes += stats->mpis_cell_txbytes;
1972 ti->tcpi_cell_rxbytes += stats->mpis_cell_rxbytes;
1973 }
1974 }
1975
1976 static void
mptcp_fill_info(struct mptses * mpte,struct tcp_info * ti)1977 mptcp_fill_info(struct mptses *mpte, struct tcp_info *ti)
1978 {
1979 struct mptsub *actsub = mpte->mpte_active_sub;
1980 struct mptcb *mp_tp = mpte->mpte_mptcb;
1981 struct tcpcb *acttp = NULL;
1982
1983 if (actsub) {
1984 acttp = sototcpcb(actsub->mpts_socket);
1985 }
1986
1987 bzero(ti, sizeof(*ti));
1988
1989 ti->tcpi_state = (uint8_t)mp_tp->mpt_state;
1990 /* tcpi_options */
1991 /* tcpi_snd_wscale */
1992 /* tcpi_rcv_wscale */
1993 /* tcpi_flags */
1994 if (acttp) {
1995 ti->tcpi_rto = acttp->t_timer[TCPT_REXMT] ? acttp->t_rxtcur : 0;
1996 }
1997
1998 /* tcpi_snd_mss */
1999 /* tcpi_rcv_mss */
2000 if (acttp) {
2001 ti->tcpi_rttcur = acttp->t_rttcur;
2002 ti->tcpi_srtt = acttp->t_srtt >> TCP_RTT_SHIFT;
2003 ti->tcpi_rttvar = acttp->t_rttvar >> TCP_RTTVAR_SHIFT;
2004 ti->tcpi_rttbest = acttp->t_rttbest >> TCP_RTT_SHIFT;
2005 ti->tcpi_rcv_srtt = acttp->rcv_srtt >> TCP_RTT_SHIFT;
2006 }
2007 /* tcpi_snd_ssthresh */
2008 /* tcpi_snd_cwnd */
2009 /* tcpi_rcv_space */
2010 ti->tcpi_snd_wnd = mp_tp->mpt_sndwnd;
2011 ti->tcpi_snd_nxt = (uint32_t)mp_tp->mpt_sndnxt;
2012 ti->tcpi_rcv_nxt = (uint32_t)mp_tp->mpt_rcvnxt;
2013 if (acttp) {
2014 ti->tcpi_last_outif = (acttp->t_inpcb->inp_last_outifp == NULL) ? 0 :
2015 acttp->t_inpcb->inp_last_outifp->if_index;
2016 }
2017
2018 mptcp_fill_info_bytestats(ti, mpte);
2019 /* tcpi_txpackets */
2020
2021 /* tcpi_txretransmitbytes */
2022 /* tcpi_txunacked */
2023 /* tcpi_rxpackets */
2024
2025 /* tcpi_rxduplicatebytes */
2026 /* tcpi_rxoutoforderbytes */
2027 /* tcpi_snd_bw */
2028 /* tcpi_synrexmits */
2029 /* tcpi_unused1 */
2030 /* tcpi_unused2 */
2031 /* tcpi_cell_rxpackets */
2032
2033 /* tcpi_cell_txpackets */
2034
2035 /* tcpi_wifi_rxpackets */
2036
2037 /* tcpi_wifi_txpackets */
2038
2039 /* tcpi_wired_rxpackets */
2040 /* tcpi_wired_txpackets */
2041 /* tcpi_connstatus */
2042 /* TFO-stuff */
2043 /* ECN stuff */
2044 /* tcpi_ecn_recv_ce */
2045 /* tcpi_ecn_recv_cwr */
2046 if (acttp) {
2047 ti->tcpi_rcvoopack = acttp->t_rcvoopack;
2048 }
2049 /* tcpi_pawsdrop */
2050 /* tcpi_sack_recovery_episode */
2051 /* tcpi_reordered_pkts */
2052 /* tcpi_dsack_sent */
2053 /* tcpi_dsack_recvd */
2054 /* tcpi_flowhash */
2055 if (acttp) {
2056 ti->tcpi_txretransmitpackets = acttp->t_stat.rxmitpkts;
2057 }
2058 }
2059
2060 /*
2061 * Handle SOPT_GET for socket options issued on MP socket.
2062 */
2063 static int
mptcp_getopt(struct mptses * mpte,struct sockopt * sopt)2064 mptcp_getopt(struct mptses *mpte, struct sockopt *sopt)
2065 {
2066 int error = 0, optval = 0;
2067 struct socket *__single mp_so;
2068
2069 mp_so = mptetoso(mpte);
2070
2071 VERIFY(!(mpsotomppcb(mp_so)->mpp_flags & MPP_INSIDE_SETGETOPT));
2072 mpsotomppcb(mp_so)->mpp_flags |= MPP_INSIDE_SETGETOPT;
2073
2074 /*
2075 * We only handle SOPT_GET for TCP level socket options; we should
2076 * not get here for socket level options since they are already
2077 * handled at the socket layer.
2078 */
2079 if (sopt->sopt_level != IPPROTO_TCP) {
2080 error = ENOPROTOOPT;
2081 goto out;
2082 }
2083
2084 switch (sopt->sopt_name) {
2085 case PERSIST_TIMEOUT:
2086 /* Only case for which we have a non-zero default */
2087 optval = tcp_max_persist_timeout;
2088 OS_FALLTHROUGH;
2089 case TCP_NODELAY:
2090 case TCP_RXT_FINDROP:
2091 case TCP_KEEPALIVE:
2092 case TCP_KEEPINTVL:
2093 case TCP_KEEPCNT:
2094 case TCP_CONNECTIONTIMEOUT:
2095 case TCP_RXT_CONNDROPTIME:
2096 case TCP_ADAPTIVE_READ_TIMEOUT:
2097 case TCP_ADAPTIVE_WRITE_TIMEOUT:
2098 case TCP_FASTOPEN_FORCE_ENABLE:
2099 {
2100 struct mptopt *__single mpo = mptcp_sopt_find(mpte, sopt);
2101
2102 if (mpo != NULL) {
2103 optval = mpo->mpo_intval;
2104 }
2105 break;
2106 }
2107
2108 /* The next ones are stored at the MPTCP-level */
2109 case TCP_NOTSENT_LOWAT:
2110 if (mptetoso(mpte)->so_flags & SOF_NOTSENT_LOWAT) {
2111 optval = mptcp_get_notsent_lowat(mpte);
2112 } else {
2113 optval = 0;
2114 }
2115 break;
2116 case TCP_INFO:
2117 {
2118 struct tcp_info ti;
2119
2120 mptcp_fill_info(mpte, &ti);
2121 error = sooptcopyout(sopt, &ti, sizeof(struct tcp_info));
2122
2123 goto out;
2124 }
2125 case MPTCP_SERVICE_TYPE:
2126 optval = mpte->mpte_svctype;
2127 break;
2128 case MPTCP_ALTERNATE_PORT:
2129 optval = mpte->mpte_alternate_port;
2130 break;
2131 case MPTCP_FORCE_ENABLE:
2132 optval = !!(mpte->mpte_flags & MPTE_FORCE_ENABLE);
2133 break;
2134 case MPTCP_FORCE_VERSION:
2135 if (mpte->mpte_flags & MPTE_FORCE_V0) {
2136 optval = 0;
2137 } else if (mpte->mpte_flags & MPTE_FORCE_V1) {
2138 optval = 1;
2139 } else {
2140 optval = -1;
2141 }
2142 break;
2143 case MPTCP_EXPECTED_PROGRESS_TARGET:
2144 error = sooptcopyout(sopt, &mpte->mpte_time_target, sizeof(mpte->mpte_time_target));
2145
2146 goto out;
2147 default:
2148 /* not eligible */
2149 error = ENOPROTOOPT;
2150 break;
2151 }
2152
2153 if (error == 0) {
2154 error = sooptcopyout(sopt, &optval, sizeof(int));
2155 }
2156
2157 out:
2158 mpsotomppcb(mp_so)->mpp_flags &= ~MPP_INSIDE_SETGETOPT;
2159 return error;
2160 }
2161
2162 /*
2163 * MPTCP SOPT_{SET,GET} socket option handler, for options issued on the MP
2164 * socket, at SOL_SOCKET and IPPROTO_TCP levels. The former is restricted
2165 * to those that are allowed by mptcp_usr_socheckopt().
2166 */
2167 int
mptcp_ctloutput(struct socket * mp_so,struct sockopt * sopt)2168 mptcp_ctloutput(struct socket *mp_so, struct sockopt *sopt)
2169 {
2170 struct mppcb *__single mpp = mpsotomppcb(mp_so);
2171 struct mptses *__single mpte;
2172 int error = 0;
2173
2174 if (mpp == NULL || mpp->mpp_state == MPPCB_STATE_DEAD) {
2175 error = EINVAL;
2176 goto out;
2177 }
2178 mpte = mptompte(mpp);
2179 socket_lock_assert_owned(mp_so);
2180
2181 /* we only handle socket and TCP-level socket options for MPTCP */
2182 if (sopt->sopt_level != SOL_SOCKET && sopt->sopt_level != IPPROTO_TCP &&
2183 sopt->sopt_level != IPPROTO_IP && sopt->sopt_level != IPPROTO_IPV6) {
2184 error = EINVAL;
2185 goto out;
2186 }
2187
2188 switch (sopt->sopt_dir) {
2189 case SOPT_SET:
2190 error = mptcp_setopt(mpte, sopt);
2191 break;
2192
2193 case SOPT_GET:
2194 error = mptcp_getopt(mpte, sopt);
2195 break;
2196 }
2197 out:
2198 return error;
2199 }
2200
2201 const char *
mptcp_sopt2str(int level,int optname)2202 mptcp_sopt2str(int level, int optname)
2203 {
2204 switch (level) {
2205 case SOL_SOCKET:
2206 switch (optname) {
2207 case SO_LINGER:
2208 return "SO_LINGER";
2209 case SO_LINGER_SEC:
2210 return "SO_LINGER_SEC";
2211 case SO_DEBUG:
2212 return "SO_DEBUG";
2213 case SO_KEEPALIVE:
2214 return "SO_KEEPALIVE";
2215 case SO_USELOOPBACK:
2216 return "SO_USELOOPBACK";
2217 case SO_TYPE:
2218 return "SO_TYPE";
2219 case SO_NREAD:
2220 return "SO_NREAD";
2221 case SO_NWRITE:
2222 return "SO_NWRITE";
2223 case SO_ERROR:
2224 return "SO_ERROR";
2225 case SO_SNDBUF:
2226 return "SO_SNDBUF";
2227 case SO_RCVBUF:
2228 return "SO_RCVBUF";
2229 case SO_SNDLOWAT:
2230 return "SO_SNDLOWAT";
2231 case SO_RCVLOWAT:
2232 return "SO_RCVLOWAT";
2233 case SO_SNDTIMEO:
2234 return "SO_SNDTIMEO";
2235 case SO_RCVTIMEO:
2236 return "SO_RCVTIMEO";
2237 case SO_NKE:
2238 return "SO_NKE";
2239 case SO_NOSIGPIPE:
2240 return "SO_NOSIGPIPE";
2241 case SO_NOADDRERR:
2242 return "SO_NOADDRERR";
2243 case SO_RESTRICTIONS:
2244 return "SO_RESTRICTIONS";
2245 case SO_LABEL:
2246 return "SO_LABEL";
2247 case SO_PEERLABEL:
2248 return "SO_PEERLABEL";
2249 case SO_RANDOMPORT:
2250 return "SO_RANDOMPORT";
2251 case SO_TRAFFIC_CLASS:
2252 return "SO_TRAFFIC_CLASS";
2253 case SO_RECV_TRAFFIC_CLASS:
2254 return "SO_RECV_TRAFFIC_CLASS";
2255 case SO_TRAFFIC_CLASS_DBG:
2256 return "SO_TRAFFIC_CLASS_DBG";
2257 case SO_PRIVILEGED_TRAFFIC_CLASS:
2258 return "SO_PRIVILEGED_TRAFFIC_CLASS";
2259 case SO_DEFUNCTIT:
2260 return "SO_DEFUNCTIT";
2261 case SO_DEFUNCTOK:
2262 return "SO_DEFUNCTOK";
2263 case SO_ISDEFUNCT:
2264 return "SO_ISDEFUNCT";
2265 case SO_OPPORTUNISTIC:
2266 return "SO_OPPORTUNISTIC";
2267 case SO_FLUSH:
2268 return "SO_FLUSH";
2269 case SO_RECV_ANYIF:
2270 return "SO_RECV_ANYIF";
2271 case SO_NOWAKEFROMSLEEP:
2272 return "SO_NOWAKEFROMSLEEP";
2273 case SO_NOAPNFALLBK:
2274 return "SO_NOAPNFALLBK";
2275 case SO_MARK_CELLFALLBACK:
2276 return "SO_CELLFALLBACK";
2277 case SO_FALLBACK_MODE:
2278 return "SO_FALLBACK_MODE";
2279 case SO_MARK_KNOWN_TRACKER:
2280 return "SO_MARK_KNOWN_TRACKER";
2281 case SO_MARK_KNOWN_TRACKER_NON_APP_INITIATED:
2282 return "SO_MARK_KNOWN_TRACKER_NON_APP_INITIATED";
2283 case SO_MARK_APPROVED_APP_DOMAIN:
2284 return "SO_MARK_APPROVED_APP_DOMAIN";
2285 case SO_DELEGATED:
2286 return "SO_DELEGATED";
2287 case SO_DELEGATED_UUID:
2288 return "SO_DELEGATED_UUID";
2289 #if NECP
2290 case SO_NECP_ATTRIBUTES:
2291 return "SO_NECP_ATTRIBUTES";
2292 case SO_NECP_CLIENTUUID:
2293 return "SO_NECP_CLIENTUUID";
2294 #endif /* NECP */
2295 }
2296
2297 break;
2298 case IPPROTO_IP:
2299 switch (optname) {
2300 case IP_TOS:
2301 return "IP_TOS";
2302 }
2303
2304 break;
2305 case IPPROTO_IPV6:
2306 switch (optname) {
2307 case IPV6_TCLASS:
2308 return "IPV6_TCLASS";
2309 }
2310
2311 break;
2312 case IPPROTO_TCP:
2313 switch (optname) {
2314 case TCP_NODELAY:
2315 return "TCP_NODELAY";
2316 case TCP_KEEPALIVE:
2317 return "TCP_KEEPALIVE";
2318 case TCP_KEEPINTVL:
2319 return "TCP_KEEPINTVL";
2320 case TCP_KEEPCNT:
2321 return "TCP_KEEPCNT";
2322 case TCP_CONNECTIONTIMEOUT:
2323 return "TCP_CONNECTIONTIMEOUT";
2324 case TCP_RXT_CONNDROPTIME:
2325 return "TCP_RXT_CONNDROPTIME";
2326 case PERSIST_TIMEOUT:
2327 return "PERSIST_TIMEOUT";
2328 case TCP_NOTSENT_LOWAT:
2329 return "NOTSENT_LOWAT";
2330 case TCP_ADAPTIVE_READ_TIMEOUT:
2331 return "ADAPTIVE_READ_TIMEOUT";
2332 case TCP_ADAPTIVE_WRITE_TIMEOUT:
2333 return "ADAPTIVE_WRITE_TIMEOUT";
2334 case TCP_FASTOPEN_FORCE_ENABLE:
2335 return "TCP_FASTOPEN_FORCE_ENABLE";
2336 case MPTCP_SERVICE_TYPE:
2337 return "MPTCP_SERVICE_TYPE";
2338 case MPTCP_ALTERNATE_PORT:
2339 return "MPTCP_ALTERNATE_PORT";
2340 case MPTCP_FORCE_ENABLE:
2341 return "MPTCP_FORCE_ENABLE";
2342 case MPTCP_FORCE_VERSION:
2343 return "MPTCP_FORCE_VERSION";
2344 case MPTCP_EXPECTED_PROGRESS_TARGET:
2345 return "MPTCP_EXPECTED_PROGRESS_TARGET";
2346 }
2347
2348 break;
2349 }
2350
2351 return "unknown";
2352 }
2353
2354 static int
mptcp_usr_preconnect(struct socket * mp_so)2355 mptcp_usr_preconnect(struct socket *mp_so)
2356 {
2357 struct mptsub *__single mpts = NULL;
2358 struct mppcb *__single mpp = mpsotomppcb(mp_so);
2359 struct mptses *__single mpte;
2360 struct socket *__single so;
2361 struct tcpcb *__single tp = NULL;
2362 int error;
2363
2364 mpte = mptompte(mpp);
2365
2366 mpts = mptcp_get_subflow(mpte, NULL);
2367 if (mpts == NULL) {
2368 os_log_error(mptcp_log_handle, "%s - %lx: invalid preconnect ",
2369 __func__, (unsigned long)VM_KERNEL_ADDRPERM(mpte));
2370 return EINVAL;
2371 }
2372 mpts->mpts_flags &= ~MPTSF_TFO_REQD;
2373 so = mpts->mpts_socket;
2374 tp = intotcpcb(sotoinpcb(so));
2375 tp->t_mpflags &= ~TMPF_TFO_REQUEST;
2376 error = tcp_output(sototcpcb(so));
2377
2378 soclearfastopen(mp_so);
2379
2380 return error;
2381 }
2382