2 * Copyright (c) 2012-2014 Apple Inc. All rights reserved.
4 * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
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.
15 * Please obtain a copy of the License at
16 * http://www.opensource.apple.com/apsl/ and read it before using this file.
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.
26 * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
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>
38 #include <sys/proc_internal.h>
39 #include <sys/resourcevar.h>
42 #include <netinet/in.h>
43 #include <netinet/in_var.h>
44 #include <netinet/tcp.h>
45 #include <netinet/tcp_fsm.h>
46 #include <netinet/tcp_seq.h>
47 #include <netinet/tcp_var.h>
48 #include <netinet/tcp_timer.h>
49 #include <netinet/mptcp_var.h>
50 #include <netinet/mptcp_timer.h>
54 static int mptcp_usr_attach(struct socket
*, int, struct proc
*);
55 static int mptcp_usr_detach(struct socket
*);
56 static int mptcp_attach(struct socket
*, struct proc
*);
57 static int mptcp_detach(struct socket
*, struct mppcb
*);
58 static int mptcp_connectx(struct mptses
*, struct sockaddr_list
**,
59 struct sockaddr_list
**, struct proc
*, uint32_t, associd_t
, connid_t
*,
60 uint32_t, void *, uint32_t);
61 static int mptcp_usr_connectx(struct socket
*, struct sockaddr_list
**,
62 struct sockaddr_list
**, struct proc
*, uint32_t, associd_t
, connid_t
*,
63 uint32_t, void *, uint32_t);
64 static int mptcp_getassocids(struct mptses
*, uint32_t *, user_addr_t
);
65 static int mptcp_getconnids(struct mptses
*, associd_t
, uint32_t *,
67 static int mptcp_getconninfo(struct mptses
*, 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
, caddr_t
, struct ifnet
*,
72 static int mptcp_disconnectx(struct mptses
*, associd_t
, connid_t
);
73 static int mptcp_usr_disconnect(struct socket
*);
74 static int mptcp_usr_disconnectx(struct socket
*, associd_t
, connid_t
);
75 static struct mptses
*mptcp_usrclosed(struct mptses
*);
76 static int mptcp_usr_peeloff(struct socket
*, associd_t
, struct socket
**);
77 static int mptcp_peeloff(struct mptses
*, associd_t
, struct socket
**);
78 static int mptcp_usr_rcvd(struct socket
*, int);
79 static int mptcp_usr_send(struct socket
*, int, struct mbuf
*,
80 struct sockaddr
*, struct mbuf
*, struct proc
*);
81 static int mptcp_usr_shutdown(struct socket
*);
82 static int mptcp_uiotombuf(struct uio
*, int, int, uint32_t, struct mbuf
**);
83 static int mptcp_usr_sosend(struct socket
*, struct sockaddr
*, struct uio
*,
84 struct mbuf
*, struct mbuf
*, int);
85 static int mptcp_usr_socheckopt(struct socket
*, struct sockopt
*);
86 static int mptcp_setopt_apply(struct mptses
*, struct mptopt
*);
87 static int mptcp_setopt(struct mptses
*, struct sockopt
*);
88 static int mptcp_getopt(struct mptses
*, struct sockopt
*);
89 static int mptcp_default_tcp_optval(struct mptses
*, struct sockopt
*, int *);
90 static void mptcp_connorder_helper(struct mptsub
*mpts
);
92 struct pr_usrreqs mptcp_usrreqs
= {
93 .pru_attach
= mptcp_usr_attach
,
94 .pru_connectx
= mptcp_usr_connectx
,
95 .pru_control
= mptcp_usr_control
,
96 .pru_detach
= mptcp_usr_detach
,
97 .pru_disconnect
= mptcp_usr_disconnect
,
98 .pru_disconnectx
= mptcp_usr_disconnectx
,
99 .pru_peeloff
= mptcp_usr_peeloff
,
100 .pru_rcvd
= mptcp_usr_rcvd
,
101 .pru_send
= mptcp_usr_send
,
102 .pru_shutdown
= mptcp_usr_shutdown
,
103 .pru_sosend
= mptcp_usr_sosend
,
104 .pru_soreceive
= soreceive
,
105 .pru_socheckopt
= mptcp_usr_socheckopt
,
109 * Attaches an MPTCP control block to a socket.
112 mptcp_usr_attach(struct socket
*mp_so
, int proto
, struct proc
*p
)
114 #pragma unused(proto)
117 VERIFY(sotomppcb(mp_so
) == NULL
);
119 error
= mptcp_attach(mp_so
, p
);
125 * Might want to use a different SO_LINGER timeout than TCP's?
127 if ((mp_so
->so_options
& SO_LINGER
) && mp_so
->so_linger
== 0)
128 mp_so
->so_linger
= TCP_LINGERTIME
* hz
;
134 * Detaches an MPTCP control block from a socket.
137 mptcp_usr_detach(struct socket
*mp_so
)
139 struct mppcb
*mpp
= sotomppcb(mp_so
);
143 VERIFY(mpp
->mpp_socket
!= NULL
);
145 error
= mptcp_detach(mp_so
, mpp
);
150 * Attach MPTCP protocol to socket, allocating MP control block,
151 * MPTCP session, control block, buffer space, etc.
154 mptcp_attach(struct socket
*mp_so
, struct proc
*p
)
162 if (mp_so
->so_snd
.sb_hiwat
== 0 || mp_so
->so_rcv
.sb_hiwat
== 0) {
163 error
= soreserve(mp_so
, tcp_sendspace
, MPTCP_RWIN_MAX
);
169 * MPTCP socket buffers cannot be compressed, due to the
170 * fact that each mbuf chained via m_next is a M_PKTHDR
171 * which carries some MPTCP metadata.
173 mp_so
->so_snd
.sb_flags
|= SB_NOCOMPRESS
;
174 mp_so
->so_rcv
.sb_flags
|= SB_NOCOMPRESS
;
176 /* Disable socket buffer auto-tuning. */
177 mp_so
->so_rcv
.sb_flags
&= ~SB_AUTOSIZE
;
178 mp_so
->so_snd
.sb_flags
&= ~SB_AUTOSIZE
;
180 if ((error
= mp_pcballoc(mp_so
, &mtcbinfo
)) != 0)
183 mpp
= sotomppcb(mp_so
);
186 mpte
= mptcp_sescreate(mp_so
, mpp
);
192 mp_tp
= mpte
->mpte_mptcb
;
193 VERIFY(mp_tp
!= NULL
);
196 mp_tp
->mpt_state
= MPTCPS_CLOSED
;
204 * Called when the socket layer loses its final reference to the socket;
205 * at this point, there is only one case in which we will keep things
209 mptcp_detach(struct socket
*mp_so
, struct mppcb
*mpp
)
212 struct mppcbinfo
*mppi
;
214 VERIFY(mp_so
->so_pcb
== mpp
);
215 VERIFY(mpp
->mpp_socket
== mp_so
);
217 mppi
= mpp
->mpp_pcbinfo
;
218 VERIFY(mppi
!= NULL
);
220 mpte
= &((struct mpp_mtp
*)mpp
)->mpp_ses
;
221 VERIFY(mpte
->mpte_mppcb
== mpp
);
223 MPTE_LOCK_ASSERT_HELD(mpte
); /* same as MP socket lock */
226 * We are done with this MPTCP socket (it has been closed);
227 * trigger all subflows to be disconnected, if not already,
228 * by initiating the PCB detach sequence (SOF_PCBCLEARING
233 (void) mptcp_disconnectx(mpte
, ASSOCID_ALL
, CONNID_ALL
);
238 * Here, we would want to handle time wait state.
245 * Common subroutine to open a MPTCP connection to one of the remote hosts
246 * specified by dst_sl. This includes allocating and establishing a
247 * subflow TCP connection, either initially to establish MPTCP connection,
248 * or to join an existing one. Returns a connection handle upon success.
251 mptcp_connectx(struct mptses
*mpte
, struct sockaddr_list
**src_sl
,
252 struct sockaddr_list
**dst_sl
, struct proc
*p
, uint32_t ifscope
,
253 associd_t aid
, connid_t
*pcid
, uint32_t flags
, void *arg
,
256 #pragma unused(p, aid, flags, arg, arglen)
258 struct socket
*mp_so
;
261 MPTE_LOCK_ASSERT_HELD(mpte
); /* same as MP socket lock */
262 mp_so
= mpte
->mpte_mppcb
->mpp_socket
;
264 VERIFY(dst_sl
!= NULL
&& *dst_sl
!= NULL
);
265 VERIFY(pcid
!= NULL
);
267 mptcplog((LOG_DEBUG
, "%s: mp_so 0x%llx\n", __func__
,
268 (u_int64_t
)VM_KERNEL_ADDRPERM(mp_so
)));
270 DTRACE_MPTCP3(connectx
, struct mptses
*, mpte
, associd_t
, aid
,
271 struct socket
*, mp_so
);
273 mpts
= mptcp_subflow_alloc(M_WAITOK
);
278 MPTS_ADDREF(mpts
); /* for this routine */
280 if (src_sl
!= NULL
) {
281 mpts
->mpts_src_sl
= *src_sl
;
284 mpts
->mpts_dst_sl
= *dst_sl
;
287 error
= mptcp_subflow_add(mpte
, mpts
, p
, ifscope
);
288 if (error
== 0 && pcid
!= NULL
)
289 *pcid
= mpts
->mpts_connid
;
293 if ((error
!= 0) && (error
!= EWOULDBLOCK
)) {
295 if (mpts
->mpts_flags
& MPTSF_ATTACHED
) {
298 mptcp_subflow_del(mpte
, mpts
, TRUE
);
310 * User-protocol pru_connectx callback.
313 mptcp_usr_connectx(struct socket
*mp_so
, struct sockaddr_list
**src_sl
,
314 struct sockaddr_list
**dst_sl
, struct proc
*p
, uint32_t ifscope
,
315 associd_t aid
, connid_t
*pcid
, uint32_t flags
, void *arg
,
318 #pragma unused(arg, arglen)
319 struct mppcb
*mpp
= sotomppcb(mp_so
);
323 if (mpp
== NULL
|| mpp
->mpp_state
== MPPCB_STATE_DEAD
) {
327 mpte
= mptompte(mpp
);
328 VERIFY(mpte
!= NULL
);
330 error
= mptcp_connectx(mpte
, src_sl
, dst_sl
, p
, ifscope
,
331 aid
, pcid
, flags
, arg
, arglen
);
337 * Handle SIOCGASSOCIDS ioctl for PF_MULTIPATH domain.
340 mptcp_getassocids(struct mptses
*mpte
, uint32_t *cnt
, user_addr_t aidp
)
342 MPTE_LOCK_ASSERT_HELD(mpte
); /* same as MP socket lock */
344 /* MPTCP has at most 1 association */
345 *cnt
= (mpte
->mpte_associd
!= ASSOCID_ANY
) ? 1 : 0;
347 /* just asking how many there are? */
348 if (aidp
== USER_ADDR_NULL
)
351 return (copyout(&mpte
->mpte_associd
, aidp
,
352 sizeof (mpte
->mpte_associd
)));
356 * Handle SIOCGCONNIDS ioctl for PF_MULTIPATH domain.
359 mptcp_getconnids(struct mptses
*mpte
, associd_t aid
, uint32_t *cnt
,
365 MPTE_LOCK_ASSERT_HELD(mpte
); /* same as MP socket lock */
367 if (aid
!= ASSOCID_ANY
&& aid
!= ASSOCID_ALL
&&
368 aid
!= mpte
->mpte_associd
)
371 *cnt
= mpte
->mpte_numflows
;
373 /* just asking how many there are? */
374 if (cidp
== USER_ADDR_NULL
)
377 TAILQ_FOREACH(mpts
, &mpte
->mpte_subflows
, mpts_entry
) {
378 if ((error
= copyout(&mpts
->mpts_connid
, cidp
,
379 sizeof (mpts
->mpts_connid
))) != 0)
382 cidp
+= sizeof (mpts
->mpts_connid
);
389 * Handle SIOCGCONNINFO ioctl for PF_MULTIPATH domain.
392 mptcp_getconninfo(struct mptses
*mpte
, connid_t
*cid
, uint32_t *flags
,
393 uint32_t *ifindex
, int32_t *soerror
, user_addr_t src
, socklen_t
*src_len
,
394 user_addr_t dst
, socklen_t
*dst_len
, uint32_t *aux_type
,
395 user_addr_t aux_data
, uint32_t *aux_len
)
397 #pragma unused(aux_data)
398 struct sockaddr_entry
*se
;
399 struct ifnet
*ifp
= NULL
;
403 MPTE_LOCK_ASSERT_HELD(mpte
); /* same as MP socket lock */
405 if (*cid
== CONNID_ALL
)
408 TAILQ_FOREACH(mpts
, &mpte
->mpte_subflows
, mpts_entry
) {
409 if (mpts
->mpts_connid
== *cid
|| *cid
== CONNID_ANY
)
413 return ((*cid
== CONNID_ANY
) ? ENXIO
: EINVAL
);
416 ifp
= mpts
->mpts_outif
;
417 *cid
= mpts
->mpts_connid
;
418 *ifindex
= ((ifp
!= NULL
) ? ifp
->if_index
: 0);
419 *soerror
= mpts
->mpts_soerror
;
421 if (mpts
->mpts_flags
& MPTSF_CONNECTING
)
422 *flags
|= CIF_CONNECTING
;
423 if (mpts
->mpts_flags
& MPTSF_CONNECTED
)
424 *flags
|= CIF_CONNECTED
;
425 if (mpts
->mpts_flags
& MPTSF_DISCONNECTING
)
426 *flags
|= CIF_DISCONNECTING
;
427 if (mpts
->mpts_flags
& MPTSF_DISCONNECTED
)
428 *flags
|= CIF_DISCONNECTED
;
429 if (mpts
->mpts_flags
& MPTSF_BOUND_IF
)
430 *flags
|= CIF_BOUND_IF
;
431 if (mpts
->mpts_flags
& MPTSF_BOUND_IP
)
432 *flags
|= CIF_BOUND_IP
;
433 if (mpts
->mpts_flags
& MPTSF_BOUND_PORT
)
434 *flags
|= CIF_BOUND_PORT
;
435 if (mpts
->mpts_flags
& MPTSF_PREFERRED
)
436 *flags
|= CIF_PREFERRED
;
437 if (mpts
->mpts_flags
& MPTSF_MP_CAPABLE
)
438 *flags
|= CIF_MP_CAPABLE
;
439 if (mpts
->mpts_flags
& MPTSF_MP_DEGRADED
)
440 *flags
|= CIF_MP_DEGRADED
;
441 if (mpts
->mpts_flags
& MPTSF_MP_READY
)
442 *flags
|= CIF_MP_READY
;
443 if (mpts
->mpts_flags
& MPTSF_ACTIVE
)
444 *flags
|= CIF_MP_ACTIVE
;
446 VERIFY(mpts
->mpts_src_sl
!= NULL
);
447 se
= TAILQ_FIRST(&mpts
->mpts_src_sl
->sl_head
);
448 VERIFY(se
!= NULL
&& se
->se_addr
!= NULL
);
449 *src_len
= se
->se_addr
->sa_len
;
450 if (src
!= USER_ADDR_NULL
) {
451 error
= copyout(se
->se_addr
, src
, se
->se_addr
->sa_len
);
456 VERIFY(mpts
->mpts_dst_sl
!= NULL
);
457 se
= TAILQ_FIRST(&mpts
->mpts_dst_sl
->sl_head
);
458 VERIFY(se
!= NULL
&& se
->se_addr
!= NULL
);
459 *dst_len
= se
->se_addr
->sa_len
;
460 if (dst
!= USER_ADDR_NULL
) {
461 error
= copyout(se
->se_addr
, dst
, se
->se_addr
->sa_len
);
468 if (mpts
->mpts_socket
!= NULL
) {
469 struct conninfo_tcp tcp_ci
;
471 *aux_type
= CIAUX_TCP
;
472 *aux_len
= sizeof (tcp_ci
);
474 if (aux_data
!= USER_ADDR_NULL
) {
475 struct socket
*so
= mpts
->mpts_socket
;
477 VERIFY(SOCK_PROTO(so
) == IPPROTO_TCP
);
478 bzero(&tcp_ci
, sizeof (tcp_ci
));
480 tcp_getconninfo(so
, &tcp_ci
);
481 socket_unlock(so
, 0);
482 error
= copyout(&tcp_ci
, aux_data
, sizeof (tcp_ci
));
487 mptcplog2((LOG_INFO
, "%s: cid %d flags %x \n",
488 __func__
, mpts
->mpts_connid
, mpts
->mpts_flags
));
495 * Handle SIOCSCONNORDER
498 mptcp_setconnorder(struct mptses
*mpte
, connid_t cid
, uint32_t rank
)
500 struct mptsub
*mpts
, *mpts1
;
503 MPTE_LOCK_ASSERT_HELD(mpte
); /* same as MP socket lock */
504 mptcplog((LOG_DEBUG
, "%s: cid %d rank %d \n", __func__
, cid
, rank
));
506 if (cid
== CONNID_ANY
|| cid
== CONNID_ALL
) {
511 TAILQ_FOREACH(mpts
, &mpte
->mpte_subflows
, mpts_entry
) {
512 if (mpts
->mpts_connid
== cid
)
520 if (rank
== 0 || rank
> 1) {
522 * If rank is 0, determine whether this should be the
523 * primary or backup subflow, depending on what we have.
525 * Otherwise, if greater than 0, make it a backup flow.
527 TAILQ_FOREACH(mpts1
, &mpte
->mpte_subflows
, mpts_entry
) {
529 if (mpts1
->mpts_flags
& MPTSF_PREFERRED
) {
537 mpts
->mpts_flags
&= ~MPTSF_PREFERRED
;
538 mpts
->mpts_rank
= rank
;
539 if (mpts1
!= NULL
&& mpts
!= mpts1
) {
540 /* preferred subflow found; set rank as necessary */
542 mpts
->mpts_rank
= (mpts1
->mpts_rank
+ 1);
543 } else if (rank
== 0) {
544 /* no preferred one found; promote this */
552 * If rank is 1, promote this subflow to be preferred.
554 TAILQ_FOREACH(mpts1
, &mpte
->mpte_subflows
, mpts_entry
) {
557 (mpts1
->mpts_flags
& MPTSF_PREFERRED
)) {
558 mpts1
->mpts_flags
&= ~MPTSF_PREFERRED
;
559 if (mpte
->mpte_nummpcapflows
> 1)
560 mptcp_connorder_helper(mpts1
);
561 } else if (mpts1
== mpts
) {
562 mpts1
->mpts_rank
= 1;
563 if (mpts1
->mpts_flags
& MPTSF_MP_CAPABLE
) {
564 mpts1
->mpts_flags
|= MPTSF_PREFERRED
;
565 if (mpte
->mpte_nummpcapflows
> 1)
566 mptcp_connorder_helper(mpts1
);
578 mptcp_connorder_helper(struct mptsub
*mpts
)
580 struct socket
*so
= mpts
->mpts_socket
;
581 struct tcpcb
*tp
= NULL
;
585 tp
= intotcpcb(sotoinpcb(so
));
586 tp
->t_mpflags
|= TMPF_SND_MPPRIO
;
587 if (mpts
->mpts_flags
& MPTSF_PREFERRED
)
588 tp
->t_mpflags
&= ~TMPF_BACKUP_PATH
;
590 tp
->t_mpflags
|= TMPF_BACKUP_PATH
;
591 mptcplog((LOG_DEBUG
, "%s cid %d flags %x", __func__
,
592 mpts
->mpts_connid
, mpts
->mpts_flags
));
593 socket_unlock(so
, 0);
598 * Handle SIOCSGONNORDER
601 mptcp_getconnorder(struct mptses
*mpte
, connid_t cid
, uint32_t *rank
)
606 MPTE_LOCK_ASSERT_HELD(mpte
); /* same as MP socket lock */
607 VERIFY(rank
!= NULL
);
610 if (cid
== CONNID_ANY
|| cid
== CONNID_ALL
) {
615 TAILQ_FOREACH(mpts
, &mpte
->mpte_subflows
, mpts_entry
) {
616 if (mpts
->mpts_connid
== cid
)
625 *rank
= mpts
->mpts_rank
;
632 * User-protocol pru_control callback.
635 mptcp_usr_control(struct socket
*mp_so
, u_long cmd
, caddr_t data
,
636 struct ifnet
*ifp
, struct proc
*p
)
638 #pragma unused(ifp, p)
639 struct mppcb
*mpp
= sotomppcb(mp_so
);
643 if (mpp
== NULL
|| mpp
->mpp_state
== MPPCB_STATE_DEAD
) {
647 mpte
= mptompte(mpp
);
648 VERIFY(mpte
!= NULL
);
650 MPTE_LOCK_ASSERT_HELD(mpte
); /* same as MP socket lock */
653 case SIOCGASSOCIDS32
: { /* struct so_aidreq32 */
654 struct so_aidreq32 aidr
;
655 bcopy(data
, &aidr
, sizeof (aidr
));
656 error
= mptcp_getassocids(mpte
, &aidr
.sar_cnt
,
659 bcopy(&aidr
, data
, sizeof (aidr
));
663 case SIOCGASSOCIDS64
: { /* struct so_aidreq64 */
664 struct so_aidreq64 aidr
;
665 bcopy(data
, &aidr
, sizeof (aidr
));
666 error
= mptcp_getassocids(mpte
, &aidr
.sar_cnt
,
669 bcopy(&aidr
, data
, sizeof (aidr
));
673 case SIOCGCONNIDS32
: { /* struct so_cidreq32 */
674 struct so_cidreq32 cidr
;
675 bcopy(data
, &cidr
, sizeof (cidr
));
676 error
= mptcp_getconnids(mpte
, cidr
.scr_aid
, &cidr
.scr_cnt
,
679 bcopy(&cidr
, data
, sizeof (cidr
));
683 case SIOCGCONNIDS64
: { /* struct so_cidreq64 */
684 struct so_cidreq64 cidr
;
685 bcopy(data
, &cidr
, sizeof (cidr
));
686 error
= mptcp_getconnids(mpte
, cidr
.scr_aid
, &cidr
.scr_cnt
,
689 bcopy(&cidr
, data
, sizeof (cidr
));
693 case SIOCGCONNINFO32
: { /* struct so_cinforeq32 */
694 struct so_cinforeq32 cifr
;
695 bcopy(data
, &cifr
, sizeof (cifr
));
696 error
= mptcp_getconninfo(mpte
, &cifr
.scir_cid
,
697 &cifr
.scir_flags
, &cifr
.scir_ifindex
, &cifr
.scir_error
,
698 cifr
.scir_src
, &cifr
.scir_src_len
, cifr
.scir_dst
,
699 &cifr
.scir_dst_len
, &cifr
.scir_aux_type
, cifr
.scir_aux_data
,
702 bcopy(&cifr
, data
, sizeof (cifr
));
706 case SIOCGCONNINFO64
: { /* struct so_cinforeq64 */
707 struct so_cinforeq64 cifr
;
708 bcopy(data
, &cifr
, sizeof (cifr
));
709 error
= mptcp_getconninfo(mpte
, &cifr
.scir_cid
,
710 &cifr
.scir_flags
, &cifr
.scir_ifindex
, &cifr
.scir_error
,
711 cifr
.scir_src
, &cifr
.scir_src_len
, cifr
.scir_dst
,
712 &cifr
.scir_dst_len
, &cifr
.scir_aux_type
, cifr
.scir_aux_data
,
715 bcopy(&cifr
, data
, sizeof (cifr
));
719 case SIOCSCONNORDER
: { /* struct so_cordreq */
720 struct so_cordreq cor
;
721 bcopy(data
, &cor
, sizeof (cor
));
722 error
= mptcp_setconnorder(mpte
, cor
.sco_cid
, cor
.sco_rank
);
724 bcopy(&cor
, data
, sizeof (cor
));
728 case SIOCGCONNORDER
: { /* struct so_cordreq */
729 struct so_cordreq cor
;
730 bcopy(data
, &cor
, sizeof (cor
));
731 error
= mptcp_getconnorder(mpte
, cor
.sco_cid
, &cor
.sco_rank
);
733 bcopy(&cor
, data
, sizeof (cor
));
746 * Initiate a disconnect. MPTCP-level disconnection is specified by
747 * CONNID_{ANY,ALL}. Otherwise, selectively disconnect a subflow
748 * connection while keeping the MPTCP-level connection (association).
751 mptcp_disconnectx(struct mptses
*mpte
, associd_t aid
, connid_t cid
)
754 struct socket
*mp_so
;
758 MPTE_LOCK_ASSERT_HELD(mpte
); /* same as MP socket lock */
760 mp_so
= mpte
->mpte_mppcb
->mpp_socket
;
761 mp_tp
= mpte
->mpte_mptcb
;
763 mptcplog((LOG_DEBUG
, "%s: mp_so 0x%llx aid %d cid %d %d\n", __func__
,
764 (u_int64_t
)VM_KERNEL_ADDRPERM(mp_so
), aid
, cid
, mp_so
->so_error
));
765 DTRACE_MPTCP5(disconnectx
, struct mptses
*, mpte
, associd_t
, aid
,
766 connid_t
, cid
, struct socket
*, mp_so
, struct mptcb
*, mp_tp
);
768 VERIFY(aid
== ASSOCID_ANY
|| aid
== ASSOCID_ALL
||
769 aid
== mpte
->mpte_associd
);
771 /* terminate the association? */
772 if (cid
== CONNID_ANY
|| cid
== CONNID_ALL
) {
773 /* if we're not detached, go thru socket state checks */
774 if (!(mp_so
->so_flags
& SOF_PCBCLEARING
)) {
775 if (!(mp_so
->so_state
& (SS_ISCONNECTED
|
780 if (mp_so
->so_state
& SS_ISDISCONNECTING
) {
786 mptcp_cancel_all_timers(mp_tp
);
787 if (mp_tp
->mpt_state
< MPTCPS_ESTABLISHED
) {
788 (void) mptcp_close(mpte
, mp_tp
);
790 } else if ((mp_so
->so_options
& SO_LINGER
) &&
791 mp_so
->so_linger
== 0) {
792 (void) mptcp_drop(mpte
, mp_tp
, 0);
796 soisdisconnecting(mp_so
);
797 sbflush(&mp_so
->so_rcv
);
798 if (mptcp_usrclosed(mpte
) != NULL
)
799 (void) mptcp_output(mpte
);
802 TAILQ_FOREACH(mpts
, &mpte
->mpte_subflows
, mpts_entry
) {
803 if (mpts
->mpts_connid
!= cid
)
806 mpts
->mpts_flags
|= MPTSF_USER_DISCONNECT
;
807 mptcp_subflow_disconnect(mpte
, mpts
, FALSE
);
819 mptcp_thread_signal(mpte
);
821 if ((mp_so
->so_state
& (SS_CANTRCVMORE
| SS_CANTSENDMORE
)) ==
822 (SS_CANTRCVMORE
| SS_CANTSENDMORE
)) {
823 /* the socket has been shutdown, no more sockopt's */
824 mptcp_flush_sopts(mpte
);
832 * Wrapper function to support disconnect on socket
835 mptcp_usr_disconnect(struct socket
*mp_so
)
839 error
= mptcp_usr_disconnectx(mp_so
, ASSOCID_ALL
, CONNID_ALL
);
844 * User-protocol pru_disconnectx callback.
847 mptcp_usr_disconnectx(struct socket
*mp_so
, associd_t aid
, connid_t cid
)
849 struct mppcb
*mpp
= sotomppcb(mp_so
);
853 if (mpp
== NULL
|| mpp
->mpp_state
== MPPCB_STATE_DEAD
) {
857 mpte
= mptompte(mpp
);
858 VERIFY(mpte
!= NULL
);
859 MPTE_LOCK_ASSERT_HELD(mpte
); /* same as MP socket lock */
861 if (aid
!= ASSOCID_ANY
&& aid
!= ASSOCID_ALL
&&
862 aid
!= mpte
->mpte_associd
) {
867 error
= mptcp_disconnectx(mpte
, aid
, cid
);
873 * User issued close, and wish to trail thru shutdown states.
875 static struct mptses
*
876 mptcp_usrclosed(struct mptses
*mpte
)
878 struct socket
*mp_so
;
882 MPTE_LOCK_ASSERT_HELD(mpte
); /* same as MP socket lock */
883 mp_so
= mpte
->mpte_mppcb
->mpp_socket
;
884 mp_tp
= mpte
->mpte_mptcb
;
887 mptcp_close_fsm(mp_tp
, MPCE_CLOSE
);
889 if (mp_tp
->mpt_state
== MPTCPS_CLOSED
) {
890 mpte
= mptcp_close(mpte
, mp_tp
);
892 } else if (mp_tp
->mpt_state
>= MPTCPS_FIN_WAIT_2
) {
894 soisdisconnected(mp_so
);
895 TAILQ_FOREACH(mpts
, &mpte
->mpte_subflows
, mpts_entry
) {
897 mpts
->mpts_flags
|= MPTSF_USER_DISCONNECT
;
903 TAILQ_FOREACH(mpts
, &mpte
->mpte_subflows
, mpts_entry
) {
905 mpts
->mpts_flags
|= MPTSF_USER_DISCONNECT
;
906 mptcp_subflow_disconnect(mpte
, mpts
, FALSE
);
915 * User-protocol pru_peeloff callback.
918 mptcp_usr_peeloff(struct socket
*mp_so
, associd_t aid
, struct socket
**psop
)
920 struct mppcb
*mpp
= sotomppcb(mp_so
);
924 VERIFY(psop
!= NULL
);
926 if (mpp
== NULL
|| mpp
->mpp_state
== MPPCB_STATE_DEAD
) {
930 mpte
= mptompte(mpp
);
931 VERIFY(mpte
!= NULL
);
933 error
= mptcp_peeloff(mpte
, aid
, psop
);
939 * Transform a previously connected TCP subflow connection which has
940 * failed to negotiate MPTCP to its own socket which can be externalized
941 * with a file descriptor. Valid only when the MPTCP socket is not
942 * yet associated (MPTCP-level connection has not been established.)
945 mptcp_peeloff(struct mptses
*mpte
, associd_t aid
, struct socket
**psop
)
947 struct socket
*so
= NULL
, *mp_so
;
951 MPTE_LOCK_ASSERT_HELD(mpte
); /* same as MP socket lock */
952 mp_so
= mpte
->mpte_mppcb
->mpp_socket
;
954 VERIFY(psop
!= NULL
);
957 DTRACE_MPTCP3(peeloff
, struct mptses
*, mpte
, associd_t
, aid
,
958 struct socket
*, mp_so
);
960 /* peeloff cannot happen after an association is established */
961 if (mpte
->mpte_associd
!= ASSOCID_ANY
) {
966 if (aid
!= ASSOCID_ANY
&& aid
!= ASSOCID_ALL
) {
971 TAILQ_FOREACH(mpts
, &mpte
->mpte_subflows
, mpts_entry
) {
973 if (mpts
->mpts_flags
& MPTSF_MP_CAPABLE
) {
974 panic("%s: so %p is MPTCP capable but mp_so %p "
975 "aid is %d\n", __func__
, so
, mp_so
,
979 MPTS_ADDREF_LOCKED(mpts
); /* for us */
980 so
= mpts
->mpts_socket
;
983 * This subflow socket is about to be externalized; make it
984 * appear as if it has the same properties as the MPTCP socket,
985 * undo what's done earlier in mptcp_subflow_add().
987 mptcp_subflow_sopeeloff(mpte
, mpts
, so
);
990 mptcp_subflow_del(mpte
, mpts
, FALSE
);
991 MPTS_REMREF(mpts
); /* ours */
995 * Here we need to make sure the subflow socket is not
996 * flow controlled; need to clear both INP_FLOW_CONTROLLED
997 * and INP_FLOW_SUSPENDED on the subflow socket, since
998 * we will no longer be monitoring its events.
1009 mptcplog((LOG_DEBUG
, "%s: mp_so 0x%llx\n", __func__
,
1010 (u_int64_t
)VM_KERNEL_ADDRPERM(mp_so
)));
1016 * After a receive, possible send some update to peer.
1019 mptcp_usr_rcvd(struct socket
*mp_so
, int flags
)
1021 #pragma unused(flags)
1022 struct mppcb
*mpp
= sotomppcb(mp_so
);
1023 struct mptses
*mpte
;
1026 if (mpp
== NULL
|| mpp
->mpp_state
== MPPCB_STATE_DEAD
) {
1030 mpte
= mptompte(mpp
);
1031 VERIFY(mpte
!= NULL
);
1033 error
= mptcp_output(mpte
);
1039 * Do a send by putting data in the output queue.
1042 mptcp_usr_send(struct socket
*mp_so
, int prus_flags
, struct mbuf
*m
,
1043 struct sockaddr
*nam
, struct mbuf
*control
, struct proc
*p
)
1045 #pragma unused(nam, p)
1046 struct mppcb
*mpp
= sotomppcb(mp_so
);
1047 struct mptses
*mpte
;
1050 if (prus_flags
& (PRUS_OOB
|PRUS_EOF
)) {
1060 if (control
!= NULL
&& control
->m_len
!= 0) {
1065 if (mpp
== NULL
|| mpp
->mpp_state
== MPPCB_STATE_DEAD
) {
1069 mpte
= mptompte(mpp
);
1070 VERIFY(mpte
!= NULL
);
1072 if (!(mp_so
->so_state
& SS_ISCONNECTED
)) {
1077 mptcp_insert_dsn(mpp
, m
);
1078 VERIFY(mp_so
->so_snd
.sb_flags
& SB_NOCOMPRESS
);
1079 (void) sbappendstream(&mp_so
->so_snd
, m
);
1084 * XXX: adi@apple.com
1086 * PRUS_MORETOCOME could be set, but we don't check it now.
1088 error
= mptcp_output(mpte
);
1095 if (control
!= NULL
)
1102 * Mark the MPTCP connection as being incapable of further output.
1105 mptcp_usr_shutdown(struct socket
*mp_so
)
1107 struct mppcb
*mpp
= sotomppcb(mp_so
);
1108 struct mptses
*mpte
;
1111 if (mpp
== NULL
|| mpp
->mpp_state
== MPPCB_STATE_DEAD
) {
1115 mpte
= mptompte(mpp
);
1116 VERIFY(mpte
!= NULL
);
1118 socantsendmore(mp_so
);
1120 mpte
= mptcp_usrclosed(mpte
);
1122 error
= mptcp_output(mpte
);
1128 * Copy the contents of uio into a properly sized mbuf chain.
1131 mptcp_uiotombuf(struct uio
*uio
, int how
, int space
, uint32_t align
,
1134 struct mbuf
*m
, *mb
, *nm
= NULL
, *mtail
= NULL
;
1135 user_ssize_t resid
, tot
, len
, progress
; /* must be user_ssize_t */
1138 VERIFY(top
!= NULL
&& *top
== NULL
);
1141 * space can be zero or an arbitrary large value bound by
1142 * the total data supplied by the uio.
1144 resid
= uio_resid(uio
);
1146 tot
= imin(resid
, space
);
1151 * The smallest unit is a single mbuf with pkthdr.
1152 * We can't align past it.
1158 * Give us the full allocation or nothing.
1159 * If space is zero return the smallest empty mbuf.
1161 if ((len
= tot
+ align
) == 0)
1164 /* Loop and append maximum sized mbufs to the chain tail. */
1166 uint32_t m_needed
= 1;
1168 if (njcl
> 0 && len
> MBIGCLBYTES
)
1169 mb
= m_getpackets_internal(&m_needed
, 1,
1170 how
, 1, M16KCLBYTES
);
1171 else if (len
> MCLBYTES
)
1172 mb
= m_getpackets_internal(&m_needed
, 1,
1173 how
, 1, MBIGCLBYTES
);
1174 else if (len
>= (signed)MINCLSIZE
)
1175 mb
= m_getpackets_internal(&m_needed
, 1,
1178 mb
= m_gethdr(how
, MT_DATA
);
1180 /* Fail the whole operation if one mbuf can't be allocated. */
1188 VERIFY(mb
->m_flags
& M_PKTHDR
);
1189 len
-= ((mb
->m_flags
& M_EXT
) ? mb
->m_ext
.ext_size
: MHLEN
);
1201 /* Fill all mbufs with uio data and update header information. */
1202 for (mb
= m
; mb
!= NULL
; mb
= mb
->m_next
) {
1203 len
= imin(M_TRAILINGSPACE(mb
), tot
- progress
);
1205 error
= uiomove(mtod(mb
, char *), len
, uio
);
1211 /* each mbuf is M_PKTHDR chained via m_next */
1213 mb
->m_pkthdr
.len
= len
;
1217 VERIFY(progress
== tot
);
1223 * MPTCP socket protocol-user socket send routine, derived from sosend().
1226 mptcp_usr_sosend(struct socket
*mp_so
, struct sockaddr
*addr
, struct uio
*uio
,
1227 struct mbuf
*top
, struct mbuf
*control
, int flags
)
1229 #pragma unused(addr)
1232 int error
, sendflags
;
1233 struct proc
*p
= current_proc();
1236 /* UIO is required for now, due to per-mbuf M_PKTHDR constrains */
1237 if (uio
== NULL
|| top
!= NULL
) {
1241 resid
= uio_resid(uio
);
1243 socket_lock(mp_so
, 1);
1244 so_update_last_owner_locked(mp_so
, p
);
1245 so_update_policy(mp_so
);
1247 VERIFY(mp_so
->so_type
== SOCK_STREAM
);
1248 VERIFY(!(mp_so
->so_flags
& SOF_MP_SUBFLOW
));
1250 if ((flags
& (MSG_OOB
|MSG_DONTROUTE
|MSG_HOLD
|MSG_SEND
|MSG_FLUSH
)) ||
1251 (mp_so
->so_flags
& SOF_ENABLE_MSGS
)) {
1253 socket_unlock(mp_so
, 1);
1258 * In theory resid should be unsigned. However, space must be
1259 * signed, as it might be less than 0 if we over-committed, and we
1260 * must use a signed comparison of space and resid. On the other
1261 * hand, a negative resid causes us to loop sending 0-length
1262 * segments to the protocol.
1264 if (resid
< 0 || (flags
& MSG_EOR
) || control
!= NULL
) {
1266 socket_unlock(mp_so
, 1);
1270 OSIncrementAtomicLong(&p
->p_stats
->p_ru
.ru_msgsnd
);
1273 error
= sosendcheck(mp_so
, NULL
, resid
, 0, 0, flags
,
1278 space
= sbspace(&mp_so
->so_snd
);
1280 socket_unlock(mp_so
, 0);
1282 * Copy the data from userland into an mbuf chain.
1284 error
= mptcp_uiotombuf(uio
, M_WAITOK
, space
, 0, &top
);
1286 socket_lock(mp_so
, 0);
1289 VERIFY(top
!= NULL
);
1290 space
-= resid
- uio_resid(uio
);
1291 resid
= uio_resid(uio
);
1292 socket_lock(mp_so
, 0);
1295 * Compute flags here, for pru_send and NKEs.
1297 sendflags
= (resid
> 0 && space
> 0) ?
1298 PRUS_MORETOCOME
: 0;
1301 * Socket filter processing
1303 VERIFY(control
== NULL
);
1304 error
= sflt_data_out(mp_so
, NULL
, &top
, &control
, 0);
1306 if (error
== EJUSTRETURN
) {
1309 /* always free control if any */
1313 if (control
!= NULL
) {
1319 * Pass data to protocol.
1321 error
= (*mp_so
->so_proto
->pr_usrreqs
->pru_send
)
1322 (mp_so
, sendflags
, top
, NULL
, NULL
, p
);
1327 } while (resid
!= 0 && space
> 0);
1328 } while (resid
!= 0);
1332 sbunlock(&mp_so
->so_snd
, FALSE
); /* will unlock socket */
1334 socket_unlock(mp_so
, 1);
1338 if (control
!= NULL
)
1345 * Called to filter SOPT_{SET,GET} for SOL_SOCKET level socket options.
1346 * This routine simply indicates to the caller whether or not to proceed
1347 * further with the given socket option. This is invoked by sosetoptlock()
1348 * and sogetoptlock().
1351 mptcp_usr_socheckopt(struct socket
*mp_so
, struct sockopt
*sopt
)
1353 #pragma unused(mp_so)
1356 VERIFY(sopt
->sopt_level
== SOL_SOCKET
);
1359 * We could check for sopt_dir (set/get) here, but we'll just
1360 * let the caller deal with it as appropriate; therefore the
1361 * following is a superset of the socket options which we
1362 * allow for set/get.
1364 * XXX: adi@apple.com
1366 * Need to consider the following cases:
1368 * a. In the event peeloff(2) occurs on the subflow socket,
1369 * we may want to issue those options which are now
1370 * handled at the MP socket. In that case, we will need
1371 * to record them in mptcp_setopt() so that they can
1372 * be replayed during peeloff.
1374 * b. Certain socket options don't have a clear definition
1375 * on the expected behavior post connect(2). At the time
1376 * those options are issued on the MP socket, there may
1377 * be existing subflow sockets that are already connected.
1379 switch (sopt
->sopt_name
) {
1380 case SO_LINGER
: /* MP */
1381 case SO_LINGER_SEC
: /* MP */
1382 case SO_TYPE
: /* MP */
1383 case SO_NREAD
: /* MP */
1384 case SO_NWRITE
: /* MP */
1385 case SO_ERROR
: /* MP */
1386 case SO_SNDBUF
: /* MP */
1387 case SO_RCVBUF
: /* MP */
1388 case SO_SNDLOWAT
: /* MP */
1389 case SO_RCVLOWAT
: /* MP */
1390 case SO_SNDTIMEO
: /* MP */
1391 case SO_RCVTIMEO
: /* MP */
1392 case SO_NKE
: /* MP */
1393 case SO_NOSIGPIPE
: /* MP */
1394 case SO_NOADDRERR
: /* MP */
1395 case SO_LABEL
: /* MP */
1396 case SO_PEERLABEL
: /* MP */
1397 case SO_DEFUNCTOK
: /* MP */
1398 case SO_ISDEFUNCT
: /* MP */
1399 case SO_TRAFFIC_CLASS_DBG
: /* MP */
1401 * Tell the caller that these options are to be processed.
1405 case SO_DEBUG
: /* MP + subflow */
1406 case SO_KEEPALIVE
: /* MP + subflow */
1407 case SO_USELOOPBACK
: /* MP + subflow */
1408 case SO_RANDOMPORT
: /* MP + subflow */
1409 case SO_TRAFFIC_CLASS
: /* MP + subflow */
1410 case SO_RECV_TRAFFIC_CLASS
: /* MP + subflow */
1411 case SO_PRIVILEGED_TRAFFIC_CLASS
: /* MP + subflow */
1412 case SO_RECV_ANYIF
: /* MP + subflow */
1413 case SO_RESTRICTIONS
: /* MP + subflow */
1414 case SO_FLUSH
: /* MP + subflow */
1415 case SO_MPTCP_FASTJOIN
: /* MP + subflow */
1416 case SO_NOWAKEFROMSLEEP
:
1418 * Tell the caller that these options are to be processed;
1419 * these will also be recorded later by mptcp_setopt().
1421 * NOTE: Only support integer option value for now.
1423 if (sopt
->sopt_valsize
!= sizeof (int))
1429 * Tell the caller to stop immediately and return an error.
1431 error
= ENOPROTOOPT
;
1439 * Issue SOPT_SET for all MPTCP subflows (for integer option values.)
1442 mptcp_setopt_apply(struct mptses
*mpte
, struct mptopt
*mpo
)
1444 struct socket
*mp_so
;
1445 struct mptsub
*mpts
;
1449 /* just bail now if this isn't applicable to subflow sockets */
1450 if (!(mpo
->mpo_flags
& MPOF_SUBFLOW_OK
)) {
1451 error
= ENOPROTOOPT
;
1456 * Skip those that are handled internally; these options
1457 * should not have been recorded and marked with the
1458 * MPOF_SUBFLOW_OK by mptcp_setopt(), but just in case.
1460 if (mpo
->mpo_level
== SOL_SOCKET
&&
1461 (mpo
->mpo_name
== SO_NOSIGPIPE
|| mpo
->mpo_name
== SO_NOADDRERR
)) {
1462 error
= ENOPROTOOPT
;
1466 MPTE_LOCK_ASSERT_HELD(mpte
); /* same as MP socket lock */
1467 mp_so
= mpte
->mpte_mppcb
->mpp_socket
;
1470 * Don't bother going further if there's no subflow; mark the option
1471 * with MPOF_INTERIM so that we know whether or not to remove this
1472 * option upon encountering an error while issuing it during subflow
1475 if (mpte
->mpte_numflows
== 0) {
1476 VERIFY(TAILQ_EMPTY(&mpte
->mpte_subflows
));
1477 mpo
->mpo_flags
|= MPOF_INTERIM
;
1478 /* return success */
1482 bzero(&smpo
, sizeof (smpo
));
1483 smpo
.mpo_flags
|= MPOF_SUBFLOW_OK
;
1484 smpo
.mpo_level
= mpo
->mpo_level
;
1485 smpo
.mpo_name
= mpo
->mpo_name
;
1487 /* grab exisiting values in case we need to rollback */
1488 TAILQ_FOREACH(mpts
, &mpte
->mpte_subflows
, mpts_entry
) {
1492 mpts
->mpts_flags
&= ~(MPTSF_SOPT_OLDVAL
|MPTSF_SOPT_INPROG
);
1493 mpts
->mpts_oldintval
= 0;
1494 smpo
.mpo_intval
= 0;
1495 VERIFY(mpts
->mpts_socket
!= NULL
);
1496 so
= mpts
->mpts_socket
;
1498 if (mptcp_subflow_sogetopt(mpte
, so
, &smpo
) == 0) {
1499 mpts
->mpts_flags
|= MPTSF_SOPT_OLDVAL
;
1500 mpts
->mpts_oldintval
= smpo
.mpo_intval
;
1502 socket_unlock(so
, 0);
1506 /* apply socket option */
1507 TAILQ_FOREACH(mpts
, &mpte
->mpte_subflows
, mpts_entry
) {
1511 mpts
->mpts_flags
|= MPTSF_SOPT_INPROG
;
1512 VERIFY(mpts
->mpts_socket
!= NULL
);
1513 so
= mpts
->mpts_socket
;
1515 error
= mptcp_subflow_sosetopt(mpte
, so
, mpo
);
1516 socket_unlock(so
, 0);
1522 /* cleanup, and rollback if needed */
1523 TAILQ_FOREACH(mpts
, &mpte
->mpte_subflows
, mpts_entry
) {
1527 if (!(mpts
->mpts_flags
& MPTSF_SOPT_INPROG
)) {
1528 /* clear in case it's set */
1529 mpts
->mpts_flags
&= ~MPTSF_SOPT_OLDVAL
;
1530 mpts
->mpts_oldintval
= 0;
1534 if (!(mpts
->mpts_flags
& MPTSF_SOPT_OLDVAL
)) {
1535 mpts
->mpts_flags
&= ~MPTSF_SOPT_INPROG
;
1536 VERIFY(mpts
->mpts_oldintval
== 0);
1540 /* error during sosetopt, so roll it back */
1542 VERIFY(mpts
->mpts_socket
!= NULL
);
1543 so
= mpts
->mpts_socket
;
1545 smpo
.mpo_intval
= mpts
->mpts_oldintval
;
1546 (void) mptcp_subflow_sosetopt(mpte
, so
, &smpo
);
1547 socket_unlock(so
, 0);
1549 mpts
->mpts_oldintval
= 0;
1550 mpts
->mpts_flags
&= ~(MPTSF_SOPT_OLDVAL
|MPTSF_SOPT_INPROG
);
1559 * Handle SOPT_SET for socket options issued on MP socket.
1562 mptcp_setopt(struct mptses
*mpte
, struct sockopt
*sopt
)
1564 int error
= 0, optval
, level
, optname
, rec
= 1;
1565 struct mptopt smpo
, *mpo
= NULL
;
1566 struct socket
*mp_so
;
1569 level
= sopt
->sopt_level
;
1570 optname
= sopt
->sopt_name
;
1572 VERIFY(sopt
->sopt_dir
== SOPT_SET
);
1573 VERIFY(level
== SOL_SOCKET
|| level
== IPPROTO_TCP
);
1574 MPTE_LOCK_ASSERT_HELD(mpte
); /* same as MP socket lock */
1575 mp_so
= mpte
->mpte_mppcb
->mpp_socket
;
1578 * Record socket options which are applicable to subflow sockets so
1579 * that we can replay them for new ones; see mptcp_usr_socheckopt()
1580 * for the list of eligible socket-level options.
1582 if (level
== SOL_SOCKET
) {
1586 case SO_USELOOPBACK
:
1588 case SO_TRAFFIC_CLASS
:
1589 case SO_RECV_TRAFFIC_CLASS
:
1590 case SO_PRIVILEGED_TRAFFIC_CLASS
:
1592 case SO_RESTRICTIONS
:
1593 case SO_NOWAKEFROMSLEEP
:
1594 case SO_MPTCP_FASTJOIN
:
1598 /* don't record it */
1602 /* nothing to do; just return success */
1608 case TCP_RXT_FINDROP
:
1612 case TCP_CONNECTIONTIMEOUT
:
1613 case TCP_RXT_CONNDROPTIME
:
1614 case PERSIST_TIMEOUT
:
1615 /* eligible; record it */
1617 case TCP_NOTSENT_LOWAT
:
1618 /* record at MPTCP level */
1619 error
= sooptcopyin(sopt
, &optval
, sizeof(optval
),
1628 mp_so
->so_flags
&= ~SOF_NOTSENT_LOWAT
;
1629 error
= mptcp_set_notsent_lowat(mpte
,0);
1631 mp_so
->so_flags
|= SOF_NOTSENT_LOWAT
;
1632 error
= mptcp_set_notsent_lowat(mpte
,
1639 error
= ENOPROTOOPT
;
1644 if ((error
= sooptcopyin(sopt
, &optval
, sizeof (optval
),
1645 sizeof (optval
))) != 0)
1649 /* search for an existing one; if not found, allocate */
1650 if ((mpo
= mptcp_sopt_find(mpte
, sopt
)) == NULL
)
1651 mpo
= mptcp_sopt_alloc(M_WAITOK
);
1656 mptcplog((LOG_DEBUG
, "%s: mp_so 0x%llx sopt %s "
1657 "val %d %s\n", __func__
,
1658 (u_int64_t
)VM_KERNEL_ADDRPERM(mp_so
),
1659 mptcp_sopt2str(level
, optname
, buf
,
1660 sizeof (buf
)), optval
,
1661 (mpo
->mpo_flags
& MPOF_ATTACHED
) ?
1662 "updated" : "recorded"));
1664 /* initialize or update, as needed */
1665 mpo
->mpo_intval
= optval
;
1666 if (!(mpo
->mpo_flags
& MPOF_ATTACHED
)) {
1667 mpo
->mpo_level
= level
;
1668 mpo
->mpo_name
= optname
;
1669 mptcp_sopt_insert(mpte
, mpo
);
1671 VERIFY(mpo
->mpo_flags
& MPOF_ATTACHED
);
1672 /* this can be issued on the subflow socket */
1673 mpo
->mpo_flags
|= MPOF_SUBFLOW_OK
;
1676 bzero(&smpo
, sizeof (smpo
));
1678 mpo
->mpo_flags
|= MPOF_SUBFLOW_OK
;
1679 mpo
->mpo_level
= level
;
1680 mpo
->mpo_name
= optname
;
1681 mpo
->mpo_intval
= optval
;
1683 VERIFY(mpo
== NULL
|| error
== 0);
1685 /* issue this socket option on existing subflows */
1687 error
= mptcp_setopt_apply(mpte
, mpo
);
1688 if (error
!= 0 && (mpo
->mpo_flags
& MPOF_ATTACHED
)) {
1689 VERIFY(mpo
!= &smpo
);
1690 mptcp_sopt_remove(mpte
, mpo
);
1691 mptcp_sopt_free(mpo
);
1694 mpo
->mpo_flags
&= ~MPOF_INTERIM
;
1697 if (error
== 0 && mpo
!= NULL
) {
1698 mptcplog((LOG_ERR
, "%s: mp_so 0x%llx sopt %s val %d set %s\n",
1699 __func__
, (u_int64_t
)VM_KERNEL_ADDRPERM(mp_so
),
1700 mptcp_sopt2str(level
, optname
, buf
,
1701 sizeof (buf
)), optval
, (mpo
->mpo_flags
& MPOF_INTERIM
) ?
1702 "pending" : "successful"));
1703 } else if (error
!= 0) {
1704 mptcplog((LOG_ERR
, "%s: mp_so 0x%llx sopt %s can't be issued "
1705 "error %d\n", __func__
,
1706 (u_int64_t
)VM_KERNEL_ADDRPERM(mp_so
), mptcp_sopt2str(level
,
1707 optname
, buf
, sizeof (buf
)), error
));
1713 * Handle SOPT_GET for socket options issued on MP socket.
1716 mptcp_getopt(struct mptses
*mpte
, struct sockopt
*sopt
)
1718 int error
= 0, optval
;
1720 VERIFY(sopt
->sopt_dir
== SOPT_GET
);
1721 MPTE_LOCK_ASSERT_HELD(mpte
); /* same as MP socket lock */
1724 * We only handle SOPT_GET for TCP level socket options; we should
1725 * not get here for socket level options since they are already
1726 * handled at the socket layer.
1728 if (sopt
->sopt_level
!= IPPROTO_TCP
) {
1729 error
= ENOPROTOOPT
;
1733 switch (sopt
->sopt_name
) {
1735 case TCP_RXT_FINDROP
:
1739 case TCP_CONNECTIONTIMEOUT
:
1740 case TCP_RXT_CONNDROPTIME
:
1741 case PERSIST_TIMEOUT
:
1742 case TCP_NOTSENT_LOWAT
:
1743 /* eligible; get the default value just in case */
1744 error
= mptcp_default_tcp_optval(mpte
, sopt
, &optval
);
1748 error
= ENOPROTOOPT
;
1752 switch (sopt
->sopt_name
) {
1753 case TCP_NOTSENT_LOWAT
:
1754 if (mpte
->mpte_mppcb
->mpp_socket
->so_flags
& SOF_NOTSENT_LOWAT
)
1755 optval
= mptcp_get_notsent_lowat(mpte
);
1762 * Search for a previously-issued TCP level socket option and
1763 * return the recorded option value. This assumes that the
1764 * value did not get modified by the lower layer after it was
1765 * issued at setsockopt(2) time. If not found, we'll return
1766 * the default value obtained ealier.
1771 if ((mpo
= mptcp_sopt_find(mpte
, sopt
)) != NULL
)
1772 optval
= mpo
->mpo_intval
;
1774 error
= sooptcopyout(sopt
, &optval
, sizeof (int));
1781 * Return default values for TCP socket options. Ideally we would query the
1782 * subflow TCP socket, but that requires creating a subflow socket before
1783 * connectx(2) time. To simplify things, just return the default values
1787 mptcp_default_tcp_optval(struct mptses
*mpte
, struct sockopt
*sopt
, int *optval
)
1791 VERIFY(sopt
->sopt_level
== IPPROTO_TCP
);
1792 VERIFY(sopt
->sopt_dir
== SOPT_GET
);
1793 MPTE_LOCK_ASSERT_HELD(mpte
); /* same as MP socket lock */
1795 /* try to do what tcp_newtcpcb() does */
1796 switch (sopt
->sopt_name
) {
1798 case TCP_RXT_FINDROP
:
1801 case TCP_CONNECTIONTIMEOUT
:
1802 case TCP_RXT_CONNDROPTIME
:
1803 case TCP_NOTSENT_LOWAT
:
1808 *optval
= mptcp_subflow_keeptime
;
1811 case PERSIST_TIMEOUT
:
1812 *optval
= tcp_max_persist_timeout
;
1816 error
= ENOPROTOOPT
;
1823 * MPTCP SOPT_{SET,GET} socket option handler, for options issued on the MP
1824 * socket, at SOL_SOCKET and IPPROTO_TCP levels. The former is restricted
1825 * to those that are allowed by mptcp_usr_socheckopt().
1828 mptcp_ctloutput(struct socket
*mp_so
, struct sockopt
*sopt
)
1830 struct mppcb
*mpp
= sotomppcb(mp_so
);
1831 struct mptses
*mpte
;
1834 if (mpp
== NULL
|| mpp
->mpp_state
== MPPCB_STATE_DEAD
) {
1838 mpte
= mptompte(mpp
);
1839 MPTE_LOCK_ASSERT_HELD(mpte
); /* same as MP socket lock */
1841 /* we only handle socket and TCP-level socket options for MPTCP */
1842 if (sopt
->sopt_level
!= SOL_SOCKET
&& sopt
->sopt_level
!= IPPROTO_TCP
) {
1844 mptcplog((LOG_DEBUG
, "%s: mp_so 0x%llx sopt %s level not "
1845 "handled\n", __func__
, (u_int64_t
)VM_KERNEL_ADDRPERM(mp_so
),
1846 mptcp_sopt2str(sopt
->sopt_level
,
1847 sopt
->sopt_name
, buf
, sizeof (buf
))));
1852 switch (sopt
->sopt_dir
) {
1854 error
= mptcp_setopt(mpte
, sopt
);
1858 error
= mptcp_getopt(mpte
, sopt
);
1866 * Return a string representation of <sopt_level,sopt_name>
1869 mptcp_sopt2str(int level
, int optname
, char *dst
, int size
)
1871 char lbuf
[32], obuf
[32];
1872 const char *l
= lbuf
, *o
= obuf
;
1874 (void) snprintf(lbuf
, sizeof (lbuf
), "0x%x", level
);
1875 (void) snprintf(obuf
, sizeof (obuf
), "0x%x", optname
);
1885 o
= "SO_LINGER_SEC";
1893 case SO_USELOOPBACK
:
1894 o
= "SO_USELOOPBACK";
1935 case SO_RESTRICTIONS
:
1936 o
= "SO_RESTRICTIONS";
1945 o
= "SO_RANDOMPORT";
1947 case SO_TRAFFIC_CLASS
:
1948 o
= "SO_TRAFFIC_CLASS";
1950 case SO_RECV_TRAFFIC_CLASS
:
1951 o
= "SO_RECV_TRAFFIC_CLASS";
1953 case SO_TRAFFIC_CLASS_DBG
:
1954 o
= "SO_TRAFFIC_CLASS_DBG";
1956 case SO_PRIVILEGED_TRAFFIC_CLASS
:
1957 o
= "SO_PRIVILEGED_TRAFFIC_CLASS";
1965 case SO_OPPORTUNISTIC
:
1966 o
= "SO_OPPORTUNISTIC";
1972 o
= "SO_RECV_ANYIF";
1974 case SO_NOWAKEFROMSLEEP
:
1975 o
= "SO_NOWAKEFROMSLEEP";
1977 case SO_MPTCP_FASTJOIN
:
1978 o
= "SO_MPTCP_FASTJOIN";
1986 o
= "TCP_KEEPALIVE";
1989 o
= "TCP_KEEPINTVL";
1994 case TCP_CONNECTIONTIMEOUT
:
1995 o
= "TCP_CONNECTIONTIMEOUT";
1997 case TCP_RXT_CONNDROPTIME
:
1998 o
= "TCP_RXT_CONNDROPTIME";
2000 case PERSIST_TIMEOUT
:
2001 o
= "PERSIST_TIMEOUT";
2007 (void) snprintf(dst
, size
, "<%s,%s>", l
, o
);