2 * Copyright (c) 1999-2017 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@
30 * Kernel Control domain - allows control connections to
31 * and to read/write data.
33 * Vincent Lubet, 040506
34 * Christophe Allie, 010928
35 * Justin C. Walker, 990319
38 #include <sys/types.h>
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/syslog.h>
42 #include <sys/socket.h>
43 #include <sys/socketvar.h>
44 #include <sys/protosw.h>
45 #include <sys/domain.h>
46 #include <sys/malloc.h>
48 #include <sys/sys_domain.h>
49 #include <sys/kern_event.h>
50 #include <sys/kern_control.h>
51 #include <sys/kauth.h>
52 #include <sys/sysctl.h>
53 #include <sys/proc_info.h>
54 #include <net/if_var.h>
56 #include <mach/vm_types.h>
58 #include <kern/thread.h>
61 TAILQ_ENTRY(kctl
) next
; /* controller chain */
64 /* controller information provided when registering */
65 char name
[MAX_KCTL_NAME
]; /* unique identifier */
69 /* misc communication information */
70 u_int32_t flags
; /* support flags */
71 u_int32_t recvbufsize
; /* request more than the default buffer size */
72 u_int32_t sendbufsize
; /* request more than the default buffer size */
74 /* Dispatch functions */
75 ctl_bind_func bind
; /* Prepare contact */
76 ctl_connect_func connect
; /* Make contact */
77 ctl_disconnect_func disconnect
; /* Break contact */
78 ctl_send_func send
; /* Send data to nke */
79 ctl_send_list_func send_list
; /* Send list of packets */
80 ctl_setopt_func setopt
; /* set kctl configuration */
81 ctl_getopt_func getopt
; /* get kctl configuration */
82 ctl_rcvd_func rcvd
; /* Notify nke when client reads data */
84 TAILQ_HEAD(, ctl_cb
) kcb_head
;
89 TAILQ_ENTRY(ctl_cb
) next
; /* controller chain */
91 struct socket
*so
; /* controlling socket */
92 struct kctl
*kctl
; /* back pointer to controller */
94 struct sockaddr_ctl sac
;
96 u_int32_t kcb_usecount
;
100 #define ROUNDUP64(x) P2ROUNDUP((x), sizeof (u_int64_t))
104 #define ADVANCE64(p, n) (void*)((char *)(p) + ROUNDUP64(n))
108 * Definitions and vars for we support
111 #define CTL_SENDSIZE (2 * 1024) /* default buffer size */
112 #define CTL_RECVSIZE (8 * 1024) /* default buffer size */
115 * Definitions and vars for we support
118 static u_int32_t ctl_maxunit
= 65536;
119 static lck_grp_attr_t
*ctl_lck_grp_attr
= 0;
120 static lck_attr_t
*ctl_lck_attr
= 0;
121 static lck_grp_t
*ctl_lck_grp
= 0;
122 static lck_mtx_t
*ctl_mtx
;
124 /* all the controllers are chained */
125 TAILQ_HEAD(kctl_list
, kctl
) ctl_head
;
127 static int ctl_attach(struct socket
*, int, struct proc
*);
128 static int ctl_detach(struct socket
*);
129 static int ctl_sofreelastref(struct socket
*so
);
130 static int ctl_bind(struct socket
*, struct sockaddr
*, struct proc
*);
131 static int ctl_connect(struct socket
*, struct sockaddr
*, struct proc
*);
132 static int ctl_disconnect(struct socket
*);
133 static int ctl_ioctl(struct socket
*so
, u_long cmd
, caddr_t data
,
134 struct ifnet
*ifp
, struct proc
*p
);
135 static int ctl_send(struct socket
*, int, struct mbuf
*,
136 struct sockaddr
*, struct mbuf
*, struct proc
*);
137 static int ctl_send_list(struct socket
*, int, struct mbuf
*,
138 struct sockaddr
*, struct mbuf
*, struct proc
*);
139 static int ctl_ctloutput(struct socket
*, struct sockopt
*);
140 static int ctl_peeraddr(struct socket
*so
, struct sockaddr
**nam
);
141 static int ctl_usr_rcvd(struct socket
*so
, int flags
);
143 static struct kctl
*ctl_find_by_name(const char *);
144 static struct kctl
*ctl_find_by_id_unit(u_int32_t id
, u_int32_t unit
);
146 static struct socket
*kcb_find_socket(kern_ctl_ref kctlref
, u_int32_t unit
,
148 static struct ctl_cb
*kcb_find(struct kctl
*, u_int32_t unit
);
149 static void ctl_post_msg(u_int32_t event_code
, u_int32_t id
);
151 static int ctl_lock(struct socket
*, int, void *);
152 static int ctl_unlock(struct socket
*, int, void *);
153 static lck_mtx_t
* ctl_getlock(struct socket
*, int);
155 static struct pr_usrreqs ctl_usrreqs
= {
156 .pru_attach
= ctl_attach
,
157 .pru_bind
= ctl_bind
,
158 .pru_connect
= ctl_connect
,
159 .pru_control
= ctl_ioctl
,
160 .pru_detach
= ctl_detach
,
161 .pru_disconnect
= ctl_disconnect
,
162 .pru_peeraddr
= ctl_peeraddr
,
163 .pru_rcvd
= ctl_usr_rcvd
,
164 .pru_send
= ctl_send
,
165 .pru_send_list
= ctl_send_list
,
166 .pru_sosend
= sosend
,
167 .pru_sosend_list
= sosend_list
,
168 .pru_soreceive
= soreceive
,
169 .pru_soreceive_list
= soreceive_list
,
172 static struct protosw kctlsw
[] = {
174 .pr_type
= SOCK_DGRAM
,
175 .pr_protocol
= SYSPROTO_CONTROL
,
176 .pr_flags
= PR_ATOMIC
| PR_CONNREQUIRED
| PR_PCBLOCK
| PR_WANTRCVD
,
177 .pr_ctloutput
= ctl_ctloutput
,
178 .pr_usrreqs
= &ctl_usrreqs
,
180 .pr_unlock
= ctl_unlock
,
181 .pr_getlock
= ctl_getlock
,
184 .pr_type
= SOCK_STREAM
,
185 .pr_protocol
= SYSPROTO_CONTROL
,
186 .pr_flags
= PR_CONNREQUIRED
| PR_PCBLOCK
| PR_WANTRCVD
,
187 .pr_ctloutput
= ctl_ctloutput
,
188 .pr_usrreqs
= &ctl_usrreqs
,
190 .pr_unlock
= ctl_unlock
,
191 .pr_getlock
= ctl_getlock
,
195 __private_extern__
int kctl_reg_list SYSCTL_HANDLER_ARGS
;
196 __private_extern__
int kctl_pcblist SYSCTL_HANDLER_ARGS
;
197 __private_extern__
int kctl_getstat SYSCTL_HANDLER_ARGS
;
200 SYSCTL_NODE(_net_systm
, OID_AUTO
, kctl
,
201 CTLFLAG_RW
| CTLFLAG_LOCKED
, 0, "Kernel control family");
203 struct kctlstat kctlstat
;
204 SYSCTL_PROC(_net_systm_kctl
, OID_AUTO
, stats
,
205 CTLTYPE_STRUCT
| CTLFLAG_RD
| CTLFLAG_LOCKED
, 0, 0,
206 kctl_getstat
, "S,kctlstat", "");
208 SYSCTL_PROC(_net_systm_kctl
, OID_AUTO
, reg_list
,
209 CTLTYPE_STRUCT
| CTLFLAG_RD
| CTLFLAG_LOCKED
, 0, 0,
210 kctl_reg_list
, "S,xkctl_reg", "");
212 SYSCTL_PROC(_net_systm_kctl
, OID_AUTO
, pcblist
,
213 CTLTYPE_STRUCT
| CTLFLAG_RD
| CTLFLAG_LOCKED
, 0, 0,
214 kctl_pcblist
, "S,xkctlpcb", "");
216 u_int32_t ctl_autorcvbuf_max
= 256 * 1024;
217 SYSCTL_INT(_net_systm_kctl
, OID_AUTO
, autorcvbufmax
,
218 CTLFLAG_RW
| CTLFLAG_LOCKED
, &ctl_autorcvbuf_max
, 0, "");
220 u_int32_t ctl_autorcvbuf_high
= 0;
221 SYSCTL_INT(_net_systm_kctl
, OID_AUTO
, autorcvbufhigh
,
222 CTLFLAG_RD
| CTLFLAG_LOCKED
, &ctl_autorcvbuf_high
, 0, "");
224 u_int32_t ctl_debug
= 0;
225 SYSCTL_INT(_net_systm_kctl
, OID_AUTO
, debug
,
226 CTLFLAG_RW
| CTLFLAG_LOCKED
, &ctl_debug
, 0, "");
228 #define KCTL_TBL_INC 16
230 static uintptr_t kctl_tbl_size
= 0;
231 static u_int32_t kctl_tbl_growing
= 0;
232 static u_int32_t kctl_tbl_growing_waiting
= 0;
233 static uintptr_t kctl_tbl_count
= 0;
234 static struct kctl
**kctl_table
= NULL
;
235 static uintptr_t kctl_ref_gencnt
= 0;
237 static void kctl_tbl_grow(void);
238 static kern_ctl_ref
kctl_make_ref(struct kctl
*kctl
);
239 static void kctl_delete_ref(kern_ctl_ref
);
240 static struct kctl
*kctl_from_ref(kern_ctl_ref
);
243 * Install the protosw's for the Kernel Control manager.
245 __private_extern__
void
246 kern_control_init(struct domain
*dp
)
250 int kctl_proto_count
= (sizeof(kctlsw
) / sizeof(struct protosw
));
252 VERIFY(!(dp
->dom_flags
& DOM_INITIALIZED
));
253 VERIFY(dp
== systemdomain
);
255 ctl_lck_grp_attr
= lck_grp_attr_alloc_init();
256 if (ctl_lck_grp_attr
== NULL
) {
257 panic("%s: lck_grp_attr_alloc_init failed\n", __func__
);
261 ctl_lck_grp
= lck_grp_alloc_init("Kernel Control Protocol",
263 if (ctl_lck_grp
== NULL
) {
264 panic("%s: lck_grp_alloc_init failed\n", __func__
);
268 ctl_lck_attr
= lck_attr_alloc_init();
269 if (ctl_lck_attr
== NULL
) {
270 panic("%s: lck_attr_alloc_init failed\n", __func__
);
274 ctl_mtx
= lck_mtx_alloc_init(ctl_lck_grp
, ctl_lck_attr
);
275 if (ctl_mtx
== NULL
) {
276 panic("%s: lck_mtx_alloc_init failed\n", __func__
);
279 TAILQ_INIT(&ctl_head
);
281 for (i
= 0, pr
= &kctlsw
[0]; i
< kctl_proto_count
; i
++, pr
++) {
282 net_add_proto(pr
, dp
, 1);
287 kcb_delete(struct ctl_cb
*kcb
)
291 lck_mtx_free(kcb
->mtx
, ctl_lck_grp
);
298 * Kernel Controller user-request functions
299 * attach function must exist and succeed
300 * detach not necessary
301 * we need a pcb for the per socket mutex
304 ctl_attach(struct socket
*so
, int proto
, struct proc
*p
)
306 #pragma unused(proto, p)
308 struct ctl_cb
*kcb
= 0;
310 MALLOC(kcb
, struct ctl_cb
*, sizeof(struct ctl_cb
), M_TEMP
, M_WAITOK
);
315 bzero(kcb
, sizeof(struct ctl_cb
));
317 kcb
->mtx
= lck_mtx_alloc_init(ctl_lck_grp
, ctl_lck_attr
);
318 if (kcb
->mtx
== NULL
) {
323 so
->so_pcb
= (caddr_t
)kcb
;
334 ctl_sofreelastref(struct socket
*so
)
336 struct ctl_cb
*kcb
= (struct ctl_cb
*)so
->so_pcb
;
342 if ((kctl
= kcb
->kctl
) != 0) {
343 lck_mtx_lock(ctl_mtx
);
344 TAILQ_REMOVE(&kctl
->kcb_head
, kcb
, next
);
345 kctlstat
.kcs_pcbcount
--;
346 kctlstat
.kcs_gencnt
++;
347 lck_mtx_unlock(ctl_mtx
);
351 sofreelastref(so
, 1);
356 * Use this function to serialize calls into the kctl subsystem
359 ctl_kcb_increment_use_count(struct ctl_cb
*kcb
, lck_mtx_t
*mutex_held
)
361 LCK_MTX_ASSERT(mutex_held
, LCK_MTX_ASSERT_OWNED
);
362 while (kcb
->kcb_usecount
> 0) {
363 msleep(&kcb
->kcb_usecount
, mutex_held
, PSOCK
| PCATCH
, "kcb_usecount", NULL
);
369 clt_kcb_decrement_use_count(struct ctl_cb
*kcb
)
371 assert(kcb
->kcb_usecount
!= 0);
373 wakeup_one((caddr_t
)&kcb
->kcb_usecount
);
377 ctl_detach(struct socket
*so
)
379 struct ctl_cb
*kcb
= (struct ctl_cb
*)so
->so_pcb
;
385 lck_mtx_t
*mtx_held
= socket_getlock(so
, PR_F_WILLUNLOCK
);
386 ctl_kcb_increment_use_count(kcb
, mtx_held
);
388 if (kcb
->kctl
!= NULL
&& kcb
->kctl
->bind
!= NULL
&&
389 kcb
->userdata
!= NULL
&& !(so
->so_state
& SS_ISCONNECTED
)) {
390 // The unit was bound, but not connected
391 // Invoke the disconnected call to cleanup
392 if (kcb
->kctl
->disconnect
!= NULL
) {
393 socket_unlock(so
, 0);
394 (*kcb
->kctl
->disconnect
)(kcb
->kctl
->kctlref
,
395 kcb
->sac
.sc_unit
, kcb
->userdata
);
400 soisdisconnected(so
);
401 so
->so_flags
|= SOF_PCBCLEARING
;
402 clt_kcb_decrement_use_count(kcb
);
407 ctl_setup_kctl(struct socket
*so
, struct sockaddr
*nam
, struct proc
*p
)
409 struct kctl
*kctl
= NULL
;
411 struct sockaddr_ctl sa
;
412 struct ctl_cb
*kcb
= (struct ctl_cb
*)so
->so_pcb
;
413 struct ctl_cb
*kcb_next
= NULL
;
415 u_int32_t recvbufsize
, sendbufsize
;
418 panic("ctl_setup_kctl so_pcb null\n");
421 if (kcb
->kctl
!= NULL
) {
422 // Already set up, skip
426 if (nam
->sa_len
!= sizeof(struct sockaddr_ctl
)) {
430 bcopy(nam
, &sa
, sizeof(struct sockaddr_ctl
));
432 lck_mtx_lock(ctl_mtx
);
433 kctl
= ctl_find_by_id_unit(sa
.sc_id
, sa
.sc_unit
);
435 lck_mtx_unlock(ctl_mtx
);
439 if (((kctl
->flags
& CTL_FLAG_REG_SOCK_STREAM
) &&
440 (so
->so_type
!= SOCK_STREAM
)) ||
441 (!(kctl
->flags
& CTL_FLAG_REG_SOCK_STREAM
) &&
442 (so
->so_type
!= SOCK_DGRAM
))) {
443 lck_mtx_unlock(ctl_mtx
);
447 if (kctl
->flags
& CTL_FLAG_PRIVILEGED
) {
449 lck_mtx_unlock(ctl_mtx
);
452 if (kauth_cred_issuser(kauth_cred_get()) == 0) {
453 lck_mtx_unlock(ctl_mtx
);
458 if ((kctl
->flags
& CTL_FLAG_REG_ID_UNIT
) || sa
.sc_unit
!= 0) {
459 if (kcb_find(kctl
, sa
.sc_unit
) != NULL
) {
460 lck_mtx_unlock(ctl_mtx
);
464 /* Find an unused ID, assumes control IDs are in order */
467 TAILQ_FOREACH(kcb_next
, &kctl
->kcb_head
, next
) {
468 if (kcb_next
->sac
.sc_unit
> unit
) {
469 /* Found a gap, lets fill it in */
472 unit
= kcb_next
->sac
.sc_unit
+ 1;
473 if (unit
== ctl_maxunit
) {
478 if (unit
== ctl_maxunit
) {
479 lck_mtx_unlock(ctl_mtx
);
486 bcopy(&sa
, &kcb
->sac
, sizeof(struct sockaddr_ctl
));
488 if (kcb_next
!= NULL
) {
489 TAILQ_INSERT_BEFORE(kcb_next
, kcb
, next
);
491 TAILQ_INSERT_TAIL(&kctl
->kcb_head
, kcb
, next
);
493 kctlstat
.kcs_pcbcount
++;
494 kctlstat
.kcs_gencnt
++;
495 kctlstat
.kcs_connections
++;
496 lck_mtx_unlock(ctl_mtx
);
499 * rdar://15526688: Limit the send and receive sizes to sb_max
500 * by using the same scaling as sbreserve()
502 sbmaxsize
= (u_quad_t
)sb_max
* MCLBYTES
/ (MSIZE
+ MCLBYTES
);
504 if (kctl
->sendbufsize
> sbmaxsize
) {
505 sendbufsize
= sbmaxsize
;
507 sendbufsize
= kctl
->sendbufsize
;
510 if (kctl
->recvbufsize
> sbmaxsize
) {
511 recvbufsize
= sbmaxsize
;
513 recvbufsize
= kctl
->recvbufsize
;
516 error
= soreserve(so
, sendbufsize
, recvbufsize
);
519 printf("%s - soreserve(%llx, %u, %u) error %d\n",
520 __func__
, (uint64_t)VM_KERNEL_ADDRPERM(so
),
521 sendbufsize
, recvbufsize
, error
);
528 soisdisconnected(so
);
529 lck_mtx_lock(ctl_mtx
);
530 TAILQ_REMOVE(&kctl
->kcb_head
, kcb
, next
);
532 kcb
->sac
.sc_unit
= 0;
533 kctlstat
.kcs_pcbcount
--;
534 kctlstat
.kcs_gencnt
++;
535 kctlstat
.kcs_conn_fail
++;
536 lck_mtx_unlock(ctl_mtx
);
542 ctl_bind(struct socket
*so
, struct sockaddr
*nam
, struct proc
*p
)
545 struct ctl_cb
*kcb
= (struct ctl_cb
*)so
->so_pcb
;
548 panic("ctl_bind so_pcb null\n");
551 lck_mtx_t
*mtx_held
= socket_getlock(so
, PR_F_WILLUNLOCK
);
552 ctl_kcb_increment_use_count(kcb
, mtx_held
);
554 error
= ctl_setup_kctl(so
, nam
, p
);
559 if (kcb
->kctl
== NULL
) {
560 panic("ctl_bind kctl null\n");
563 if (kcb
->kctl
->bind
== NULL
) {
568 socket_unlock(so
, 0);
569 error
= (*kcb
->kctl
->bind
)(kcb
->kctl
->kctlref
, &kcb
->sac
, &kcb
->userdata
);
573 clt_kcb_decrement_use_count(kcb
);
578 ctl_connect(struct socket
*so
, struct sockaddr
*nam
, struct proc
*p
)
581 struct ctl_cb
*kcb
= (struct ctl_cb
*)so
->so_pcb
;
584 panic("ctl_connect so_pcb null\n");
587 lck_mtx_t
*mtx_held
= socket_getlock(so
, PR_F_WILLUNLOCK
);
588 ctl_kcb_increment_use_count(kcb
, mtx_held
);
590 error
= ctl_setup_kctl(so
, nam
, p
);
595 if (kcb
->kctl
== NULL
) {
596 panic("ctl_connect kctl null\n");
600 socket_unlock(so
, 0);
601 error
= (*kcb
->kctl
->connect
)(kcb
->kctl
->kctlref
, &kcb
->sac
, &kcb
->userdata
);
609 if (error
&& kcb
->kctl
->disconnect
) {
611 * XXX Make sure we Don't check the return value
612 * of disconnect here.
613 * ipsec/utun_ctl_disconnect will return error when
614 * disconnect gets called after connect failure.
615 * However if we decide to check for disconnect return
616 * value here. Please make sure to revisit
617 * ipsec/utun_ctl_disconnect.
619 socket_unlock(so
, 0);
620 (*kcb
->kctl
->disconnect
)(kcb
->kctl
->kctlref
, kcb
->sac
.sc_unit
, kcb
->userdata
);
624 soisdisconnected(so
);
625 lck_mtx_lock(ctl_mtx
);
626 TAILQ_REMOVE(&kcb
->kctl
->kcb_head
, kcb
, next
);
628 kcb
->sac
.sc_unit
= 0;
629 kctlstat
.kcs_pcbcount
--;
630 kctlstat
.kcs_gencnt
++;
631 kctlstat
.kcs_conn_fail
++;
632 lck_mtx_unlock(ctl_mtx
);
635 clt_kcb_decrement_use_count(kcb
);
640 ctl_disconnect(struct socket
*so
)
642 struct ctl_cb
*kcb
= (struct ctl_cb
*)so
->so_pcb
;
644 if ((kcb
= (struct ctl_cb
*)so
->so_pcb
)) {
645 lck_mtx_t
*mtx_held
= socket_getlock(so
, PR_F_WILLUNLOCK
);
646 ctl_kcb_increment_use_count(kcb
, mtx_held
);
647 struct kctl
*kctl
= kcb
->kctl
;
649 if (kctl
&& kctl
->disconnect
) {
650 socket_unlock(so
, 0);
651 (*kctl
->disconnect
)(kctl
->kctlref
, kcb
->sac
.sc_unit
,
656 soisdisconnected(so
);
658 socket_unlock(so
, 0);
659 lck_mtx_lock(ctl_mtx
);
661 kcb
->sac
.sc_unit
= 0;
662 while (kcb
->usecount
!= 0) {
663 msleep(&kcb
->usecount
, ctl_mtx
, 0, "kcb->usecount", 0);
665 TAILQ_REMOVE(&kctl
->kcb_head
, kcb
, next
);
666 kctlstat
.kcs_pcbcount
--;
667 kctlstat
.kcs_gencnt
++;
668 lck_mtx_unlock(ctl_mtx
);
670 clt_kcb_decrement_use_count(kcb
);
676 ctl_peeraddr(struct socket
*so
, struct sockaddr
**nam
)
678 struct ctl_cb
*kcb
= (struct ctl_cb
*)so
->so_pcb
;
680 struct sockaddr_ctl sc
;
682 if (kcb
== NULL
) { /* sanity check */
686 if ((kctl
= kcb
->kctl
) == NULL
) {
690 bzero(&sc
, sizeof(struct sockaddr_ctl
));
691 sc
.sc_len
= sizeof(struct sockaddr_ctl
);
692 sc
.sc_family
= AF_SYSTEM
;
693 sc
.ss_sysaddr
= AF_SYS_CONTROL
;
695 sc
.sc_unit
= kcb
->sac
.sc_unit
;
697 *nam
= dup_sockaddr((struct sockaddr
*)&sc
, 1);
703 ctl_sbrcv_trim(struct socket
*so
)
705 struct sockbuf
*sb
= &so
->so_rcv
;
707 if (sb
->sb_hiwat
> sb
->sb_idealsize
) {
712 * The difference between the ideal size and the
713 * current size is the upper bound of the trimage
715 diff
= sb
->sb_hiwat
- sb
->sb_idealsize
;
717 * We cannot trim below the outstanding data
719 trim
= sb
->sb_hiwat
- sb
->sb_cc
;
721 trim
= imin(trim
, (int32_t)diff
);
724 sbreserve(sb
, (sb
->sb_hiwat
- trim
));
727 printf("%s - shrunk to %d\n",
728 __func__
, sb
->sb_hiwat
);
735 ctl_usr_rcvd(struct socket
*so
, int flags
)
738 struct ctl_cb
*kcb
= (struct ctl_cb
*)so
->so_pcb
;
745 lck_mtx_t
*mtx_held
= socket_getlock(so
, PR_F_WILLUNLOCK
);
746 ctl_kcb_increment_use_count(kcb
, mtx_held
);
748 if ((kctl
= kcb
->kctl
) == NULL
) {
754 socket_unlock(so
, 0);
755 (*kctl
->rcvd
)(kctl
->kctlref
, kcb
->sac
.sc_unit
, kcb
->userdata
, flags
);
762 clt_kcb_decrement_use_count(kcb
);
767 ctl_send(struct socket
*so
, int flags
, struct mbuf
*m
,
768 struct sockaddr
*addr
, struct mbuf
*control
,
771 #pragma unused(addr, p)
773 struct ctl_cb
*kcb
= (struct ctl_cb
*)so
->so_pcb
;
780 if (kcb
== NULL
) { /* sanity check */
784 lck_mtx_t
*mtx_held
= socket_getlock(so
, PR_F_WILLUNLOCK
);
785 ctl_kcb_increment_use_count(kcb
, mtx_held
);
787 if (error
== 0 && (kctl
= kcb
->kctl
) == NULL
) {
791 if (error
== 0 && kctl
->send
) {
792 so_tc_update_stats(m
, so
, m_get_service_class(m
));
793 socket_unlock(so
, 0);
794 error
= (*kctl
->send
)(kctl
->kctlref
, kcb
->sac
.sc_unit
, kcb
->userdata
,
804 OSIncrementAtomic64((SInt64
*)&kctlstat
.kcs_send_fail
);
806 clt_kcb_decrement_use_count(kcb
);
812 ctl_send_list(struct socket
*so
, int flags
, struct mbuf
*m
,
813 __unused
struct sockaddr
*addr
, struct mbuf
*control
,
814 __unused
struct proc
*p
)
817 struct ctl_cb
*kcb
= (struct ctl_cb
*)so
->so_pcb
;
821 m_freem_list(control
);
824 if (kcb
== NULL
) { /* sanity check */
828 lck_mtx_t
*mtx_held
= socket_getlock(so
, PR_F_WILLUNLOCK
);
829 ctl_kcb_increment_use_count(kcb
, mtx_held
);
831 if (error
== 0 && (kctl
= kcb
->kctl
) == NULL
) {
835 if (error
== 0 && kctl
->send_list
) {
838 for (nxt
= m
; nxt
!= NULL
; nxt
= nxt
->m_nextpkt
) {
839 so_tc_update_stats(nxt
, so
, m_get_service_class(nxt
));
842 socket_unlock(so
, 0);
843 error
= (*kctl
->send_list
)(kctl
->kctlref
, kcb
->sac
.sc_unit
,
844 kcb
->userdata
, m
, flags
);
846 } else if (error
== 0 && kctl
->send
) {
847 while (m
!= NULL
&& error
== 0) {
848 struct mbuf
*nextpkt
= m
->m_nextpkt
;
851 so_tc_update_stats(m
, so
, m_get_service_class(m
));
852 socket_unlock(so
, 0);
853 error
= (*kctl
->send
)(kctl
->kctlref
, kcb
->sac
.sc_unit
,
854 kcb
->userdata
, m
, flags
);
868 OSIncrementAtomic64((SInt64
*)&kctlstat
.kcs_send_list_fail
);
870 clt_kcb_decrement_use_count(kcb
);
876 ctl_rcvbspace(struct socket
*so
, u_int32_t datasize
,
877 u_int32_t kctlflags
, u_int32_t flags
)
879 struct sockbuf
*sb
= &so
->so_rcv
;
880 u_int32_t space
= sbspace(sb
);
883 if ((kctlflags
& CTL_FLAG_REG_CRIT
) == 0) {
884 if ((u_int32_t
) space
>= datasize
) {
889 } else if ((flags
& CTL_DATA_CRIT
) == 0) {
891 * Reserve 25% for critical messages
893 if (space
< (sb
->sb_hiwat
>> 2) ||
900 u_int32_t autorcvbuf_max
;
903 * Allow overcommit of 25%
905 autorcvbuf_max
= min(sb
->sb_idealsize
+ (sb
->sb_idealsize
>> 2),
908 if ((u_int32_t
) space
>= datasize
) {
910 } else if (tcp_cansbgrow(sb
) &&
911 sb
->sb_hiwat
< autorcvbuf_max
) {
913 * Grow with a little bit of leeway
915 u_int32_t grow
= datasize
- space
+ MSIZE
;
918 min((sb
->sb_hiwat
+ grow
), autorcvbuf_max
)) == 1) {
919 if (sb
->sb_hiwat
> ctl_autorcvbuf_high
) {
920 ctl_autorcvbuf_high
= sb
->sb_hiwat
;
926 if ((u_int32_t
) sbspace(sb
) >= datasize
) {
933 printf("%s - grown to %d error %d\n",
934 __func__
, sb
->sb_hiwat
, error
);
947 ctl_enqueuembuf(kern_ctl_ref kctlref
, u_int32_t unit
, struct mbuf
*m
,
952 int len
= m
->m_pkthdr
.len
;
955 so
= kcb_find_socket(kctlref
, unit
, &kctlflags
);
960 if (ctl_rcvbspace(so
, len
, kctlflags
, flags
) != 0) {
962 OSIncrementAtomic64((SInt64
*)&kctlstat
.kcs_enqueue_fullsock
);
965 if ((flags
& CTL_DATA_EOR
)) {
969 so_recv_data_stat(so
, m
, 0);
970 if (sbappend(&so
->so_rcv
, m
) != 0) {
971 if ((flags
& CTL_DATA_NOWAKEUP
) == 0) {
976 OSIncrementAtomic64((SInt64
*)&kctlstat
.kcs_enqueue_fullsock
);
979 if (ctl_debug
&& error
!= 0 && (flags
& CTL_DATA_CRIT
)) {
980 printf("%s - crit data err %d len %d hiwat %d cc: %d\n",
981 __func__
, error
, len
,
982 so
->so_rcv
.sb_hiwat
, so
->so_rcv
.sb_cc
);
985 socket_unlock(so
, 1);
987 OSIncrementAtomic64((SInt64
*)&kctlstat
.kcs_enqueue_fail
);
994 * Compute space occupied by mbuf like sbappendrecord
997 m_space(struct mbuf
*m
)
1002 for (nxt
= m
; nxt
!= NULL
; nxt
= nxt
->m_next
) {
1003 space
+= nxt
->m_len
;
1010 ctl_enqueuembuf_list(void *kctlref
, u_int32_t unit
, struct mbuf
*m_list
,
1011 u_int32_t flags
, struct mbuf
**m_remain
)
1013 struct socket
*so
= NULL
;
1015 struct mbuf
*m
, *nextpkt
;
1018 u_int32_t kctlflags
;
1021 * Need to point the beginning of the list in case of early exit
1026 * kcb_find_socket takes the socket lock with a reference
1028 so
= kcb_find_socket(kctlref
, unit
, &kctlflags
);
1034 if (kctlflags
& CTL_FLAG_REG_SOCK_STREAM
) {
1038 if (flags
& CTL_DATA_EOR
) {
1043 for (m
= m_list
; m
!= NULL
; m
= nextpkt
) {
1044 nextpkt
= m
->m_nextpkt
;
1046 if (m
->m_pkthdr
.len
== 0 && ctl_debug
) {
1047 printf("%s: %llx m_pkthdr.len is 0",
1048 __func__
, (uint64_t)VM_KERNEL_ADDRPERM(m
));
1052 * The mbuf is either appended or freed by sbappendrecord()
1053 * so it's not reliable from a data standpoint
1056 if (ctl_rcvbspace(so
, len
, kctlflags
, flags
) != 0) {
1058 OSIncrementAtomic64(
1059 (SInt64
*)&kctlstat
.kcs_enqueue_fullsock
);
1063 * Unlink from the list, m is on its own
1065 m
->m_nextpkt
= NULL
;
1066 so_recv_data_stat(so
, m
, 0);
1067 if (sbappendrecord(&so
->so_rcv
, m
) != 0) {
1071 * We free or return the remaining
1076 OSIncrementAtomic64(
1077 (SInt64
*)&kctlstat
.kcs_enqueue_fullsock
);
1082 if (needwakeup
&& (flags
& CTL_DATA_NOWAKEUP
) == 0) {
1088 if (ctl_debug
&& error
!= 0 && (flags
& CTL_DATA_CRIT
)) {
1089 printf("%s - crit data err %d len %d hiwat %d cc: %d\n",
1090 __func__
, error
, len
,
1091 so
->so_rcv
.sb_hiwat
, so
->so_rcv
.sb_cc
);
1094 socket_unlock(so
, 1);
1099 if (m
!= NULL
&& socket_debug
&& so
!= NULL
&&
1100 (so
->so_options
& SO_DEBUG
)) {
1103 printf("%s m_list %llx\n", __func__
,
1104 (uint64_t) VM_KERNEL_ADDRPERM(m_list
));
1105 for (n
= m
; n
!= NULL
; n
= n
->m_nextpkt
) {
1106 printf(" remain %llx m_next %llx\n",
1107 (uint64_t) VM_KERNEL_ADDRPERM(n
),
1108 (uint64_t) VM_KERNEL_ADDRPERM(n
->m_next
));
1117 OSIncrementAtomic64((SInt64
*)&kctlstat
.kcs_enqueue_fail
);
1123 ctl_enqueuedata(void *kctlref
, u_int32_t unit
, void *data
, size_t len
,
1129 unsigned int num_needed
;
1132 u_int32_t kctlflags
;
1134 so
= kcb_find_socket(kctlref
, unit
, &kctlflags
);
1139 if (ctl_rcvbspace(so
, len
, kctlflags
, flags
) != 0) {
1141 OSIncrementAtomic64((SInt64
*)&kctlstat
.kcs_enqueue_fullsock
);
1146 m
= m_allocpacket_internal(&num_needed
, len
, NULL
, M_NOWAIT
, 1, 0);
1148 kctlstat
.kcs_enqdata_mb_alloc_fail
++;
1150 printf("%s: m_allocpacket_internal(%lu) failed\n",
1157 for (n
= m
; n
!= NULL
; n
= n
->m_next
) {
1158 size_t mlen
= mbuf_maxlen(n
);
1160 if (mlen
+ curlen
> len
) {
1161 mlen
= len
- curlen
;
1164 bcopy((char *)data
+ curlen
, n
->m_data
, mlen
);
1167 mbuf_pkthdr_setlen(m
, curlen
);
1169 if ((flags
& CTL_DATA_EOR
)) {
1170 m
->m_flags
|= M_EOR
;
1172 so_recv_data_stat(so
, m
, 0);
1173 if (sbappend(&so
->so_rcv
, m
) != 0) {
1174 if ((flags
& CTL_DATA_NOWAKEUP
) == 0) {
1178 kctlstat
.kcs_enqdata_sbappend_fail
++;
1180 OSIncrementAtomic64((SInt64
*)&kctlstat
.kcs_enqueue_fullsock
);
1184 if (ctl_debug
&& error
!= 0 && (flags
& CTL_DATA_CRIT
)) {
1185 printf("%s - crit data err %d len %d hiwat %d cc: %d\n",
1186 __func__
, error
, (int)len
,
1187 so
->so_rcv
.sb_hiwat
, so
->so_rcv
.sb_cc
);
1190 socket_unlock(so
, 1);
1192 OSIncrementAtomic64((SInt64
*)&kctlstat
.kcs_enqueue_fail
);
1198 ctl_getenqueuepacketcount(kern_ctl_ref kctlref
, u_int32_t unit
, u_int32_t
*pcnt
)
1208 so
= kcb_find_socket(kctlref
, unit
, NULL
);
1214 m1
= so
->so_rcv
.sb_mb
;
1215 while (m1
!= NULL
) {
1216 if (m1
->m_type
== MT_DATA
||
1217 m1
->m_type
== MT_HEADER
||
1218 m1
->m_type
== MT_OOBDATA
) {
1225 socket_unlock(so
, 1);
1231 ctl_getenqueuespace(kern_ctl_ref kctlref
, u_int32_t unit
, size_t *space
)
1236 if (space
== NULL
) {
1240 so
= kcb_find_socket(kctlref
, unit
, NULL
);
1245 avail
= sbspace(&so
->so_rcv
);
1246 *space
= (avail
< 0) ? 0 : avail
;
1247 socket_unlock(so
, 1);
1253 ctl_getenqueuereadable(kern_ctl_ref kctlref
, u_int32_t unit
,
1254 u_int32_t
*difference
)
1258 if (difference
== NULL
) {
1262 so
= kcb_find_socket(kctlref
, unit
, NULL
);
1267 if (so
->so_rcv
.sb_cc
>= so
->so_rcv
.sb_lowat
) {
1270 *difference
= (so
->so_rcv
.sb_lowat
- so
->so_rcv
.sb_cc
);
1272 socket_unlock(so
, 1);
1278 ctl_ctloutput(struct socket
*so
, struct sockopt
*sopt
)
1280 struct ctl_cb
*kcb
= (struct ctl_cb
*)so
->so_pcb
;
1286 if (sopt
->sopt_level
!= SYSPROTO_CONTROL
) {
1290 if (kcb
== NULL
) { /* sanity check */
1294 if ((kctl
= kcb
->kctl
) == NULL
) {
1298 lck_mtx_t
*mtx_held
= socket_getlock(so
, PR_F_WILLUNLOCK
);
1299 ctl_kcb_increment_use_count(kcb
, mtx_held
);
1301 switch (sopt
->sopt_dir
) {
1303 if (kctl
->setopt
== NULL
) {
1307 if (sopt
->sopt_valsize
!= 0) {
1308 MALLOC(data
, void *, sopt
->sopt_valsize
, M_TEMP
,
1314 error
= sooptcopyin(sopt
, data
,
1315 sopt
->sopt_valsize
, sopt
->sopt_valsize
);
1318 socket_unlock(so
, 0);
1319 error
= (*kctl
->setopt
)(kctl
->kctlref
,
1320 kcb
->sac
.sc_unit
, kcb
->userdata
, sopt
->sopt_name
,
1321 data
, sopt
->sopt_valsize
);
1331 if (kctl
->getopt
== NULL
) {
1336 if (sopt
->sopt_valsize
&& sopt
->sopt_val
) {
1337 MALLOC(data
, void *, sopt
->sopt_valsize
, M_TEMP
,
1344 * 4108337 - copy user data in case the
1345 * kernel control needs it
1347 error
= sooptcopyin(sopt
, data
,
1348 sopt
->sopt_valsize
, sopt
->sopt_valsize
);
1352 len
= sopt
->sopt_valsize
;
1353 socket_unlock(so
, 0);
1354 error
= (*kctl
->getopt
)(kctl
->kctlref
, kcb
->sac
.sc_unit
,
1355 kcb
->userdata
, sopt
->sopt_name
,
1357 if (data
!= NULL
&& len
> sopt
->sopt_valsize
) {
1358 panic_plain("ctl_ctloutput: ctl %s returned "
1359 "len (%lu) > sopt_valsize (%lu)\n",
1360 kcb
->kctl
->name
, len
,
1361 sopt
->sopt_valsize
);
1366 error
= sooptcopyout(sopt
, data
, len
);
1368 sopt
->sopt_valsize
= len
;
1379 clt_kcb_decrement_use_count(kcb
);
1384 ctl_ioctl(struct socket
*so
, u_long cmd
, caddr_t data
,
1385 struct ifnet
*ifp
, struct proc
*p
)
1387 #pragma unused(so, ifp, p)
1388 int error
= ENOTSUP
;
1391 /* get the number of controllers */
1392 case CTLIOCGCOUNT
: {
1396 lck_mtx_lock(ctl_mtx
);
1397 TAILQ_FOREACH(kctl
, &ctl_head
, next
)
1399 lck_mtx_unlock(ctl_mtx
);
1401 bcopy(&n
, data
, sizeof(n
));
1406 struct ctl_info ctl_info
;
1407 struct kctl
*kctl
= 0;
1410 bcopy(data
, &ctl_info
, sizeof(ctl_info
));
1411 name_len
= strnlen(ctl_info
.ctl_name
, MAX_KCTL_NAME
);
1413 if (name_len
== 0 || name_len
+ 1 > MAX_KCTL_NAME
) {
1417 lck_mtx_lock(ctl_mtx
);
1418 kctl
= ctl_find_by_name(ctl_info
.ctl_name
);
1419 lck_mtx_unlock(ctl_mtx
);
1424 ctl_info
.ctl_id
= kctl
->id
;
1425 bcopy(&ctl_info
, data
, sizeof(ctl_info
));
1430 /* add controls to get list of NKEs */
1439 struct kctl
**new_table
;
1442 lck_mtx_assert(ctl_mtx
, LCK_MTX_ASSERT_OWNED
);
1444 if (kctl_tbl_growing
) {
1445 /* Another thread is allocating */
1446 kctl_tbl_growing_waiting
++;
1449 (void) msleep((caddr_t
) &kctl_tbl_growing
, ctl_mtx
,
1450 PSOCK
| PCATCH
, "kctl_tbl_growing", 0);
1451 } while (kctl_tbl_growing
);
1452 kctl_tbl_growing_waiting
--;
1454 /* Another thread grew the table */
1455 if (kctl_table
!= NULL
&& kctl_tbl_count
< kctl_tbl_size
) {
1459 /* Verify we have a sane size */
1460 if (kctl_tbl_size
+ KCTL_TBL_INC
>= UINT16_MAX
) {
1461 kctlstat
.kcs_tbl_size_too_big
++;
1463 printf("%s kctl_tbl_size %lu too big\n",
1464 __func__
, kctl_tbl_size
);
1468 kctl_tbl_growing
= 1;
1470 new_size
= kctl_tbl_size
+ KCTL_TBL_INC
;
1472 lck_mtx_unlock(ctl_mtx
);
1473 new_table
= _MALLOC(sizeof(struct kctl
*) * new_size
,
1474 M_TEMP
, M_WAIT
| M_ZERO
);
1475 lck_mtx_lock(ctl_mtx
);
1477 if (new_table
!= NULL
) {
1478 if (kctl_table
!= NULL
) {
1479 bcopy(kctl_table
, new_table
,
1480 kctl_tbl_size
* sizeof(struct kctl
*));
1482 _FREE(kctl_table
, M_TEMP
);
1484 kctl_table
= new_table
;
1485 kctl_tbl_size
= new_size
;
1488 kctl_tbl_growing
= 0;
1490 if (kctl_tbl_growing_waiting
) {
1491 wakeup(&kctl_tbl_growing
);
1495 #define KCTLREF_INDEX_MASK 0x0000FFFF
1496 #define KCTLREF_GENCNT_MASK 0xFFFF0000
1497 #define KCTLREF_GENCNT_SHIFT 16
1500 kctl_make_ref(struct kctl
*kctl
)
1504 lck_mtx_assert(ctl_mtx
, LCK_MTX_ASSERT_OWNED
);
1506 if (kctl_tbl_count
>= kctl_tbl_size
) {
1510 kctl
->kctlref
= NULL
;
1511 for (i
= 0; i
< kctl_tbl_size
; i
++) {
1512 if (kctl_table
[i
] == NULL
) {
1516 * Reference is index plus one
1518 kctl_ref_gencnt
+= 1;
1521 * Add generation count as salt to reference to prevent
1522 * use after deregister
1524 ref
= ((kctl_ref_gencnt
<< KCTLREF_GENCNT_SHIFT
) &
1525 KCTLREF_GENCNT_MASK
) +
1526 ((i
+ 1) & KCTLREF_INDEX_MASK
);
1528 kctl
->kctlref
= (void *)(ref
);
1529 kctl_table
[i
] = kctl
;
1535 if (kctl
->kctlref
== NULL
) {
1536 panic("%s no space in table", __func__
);
1539 if (ctl_debug
> 0) {
1540 printf("%s %p for %p\n",
1541 __func__
, kctl
->kctlref
, kctl
);
1544 return kctl
->kctlref
;
1548 kctl_delete_ref(kern_ctl_ref kctlref
)
1551 * Reference is index plus one
1553 uintptr_t i
= (((uintptr_t)kctlref
) & KCTLREF_INDEX_MASK
) - 1;
1555 lck_mtx_assert(ctl_mtx
, LCK_MTX_ASSERT_OWNED
);
1557 if (i
< kctl_tbl_size
) {
1558 struct kctl
*kctl
= kctl_table
[i
];
1560 if (kctl
->kctlref
== kctlref
) {
1561 kctl_table
[i
] = NULL
;
1564 kctlstat
.kcs_bad_kctlref
++;
1567 kctlstat
.kcs_bad_kctlref
++;
1571 static struct kctl
*
1572 kctl_from_ref(kern_ctl_ref kctlref
)
1575 * Reference is index plus one
1577 uintptr_t i
= (((uintptr_t)kctlref
) & KCTLREF_INDEX_MASK
) - 1;
1578 struct kctl
*kctl
= NULL
;
1580 lck_mtx_assert(ctl_mtx
, LCK_MTX_ASSERT_OWNED
);
1582 if (i
>= kctl_tbl_size
) {
1583 kctlstat
.kcs_bad_kctlref
++;
1586 kctl
= kctl_table
[i
];
1587 if (kctl
->kctlref
!= kctlref
) {
1588 kctlstat
.kcs_bad_kctlref
++;
1595 * Register/unregister a NKE
1598 ctl_register(struct kern_ctl_reg
*userkctl
, kern_ctl_ref
*kctlref
)
1600 struct kctl
*kctl
= NULL
;
1601 struct kctl
*kctl_next
= NULL
;
1604 int is_extended
= 0;
1606 if (userkctl
== NULL
) { /* sanity check */
1609 if (userkctl
->ctl_connect
== NULL
) {
1612 name_len
= strlen(userkctl
->ctl_name
);
1613 if (name_len
== 0 || name_len
+ 1 > MAX_KCTL_NAME
) {
1617 MALLOC(kctl
, struct kctl
*, sizeof(*kctl
), M_TEMP
, M_WAITOK
);
1621 bzero((char *)kctl
, sizeof(*kctl
));
1623 lck_mtx_lock(ctl_mtx
);
1625 if (kctl_make_ref(kctl
) == NULL
) {
1626 lck_mtx_unlock(ctl_mtx
);
1632 * Kernel Control IDs
1634 * CTL_FLAG_REG_ID_UNIT indicates the control ID and unit number are
1635 * static. If they do not exist, add them to the list in order. If the
1636 * flag is not set, we must find a new unique value. We assume the
1637 * list is in order. We find the last item in the list and add one. If
1638 * this leads to wrapping the id around, we start at the front of the
1639 * list and look for a gap.
1642 if ((userkctl
->ctl_flags
& CTL_FLAG_REG_ID_UNIT
) == 0) {
1643 /* Must dynamically assign an unused ID */
1645 /* Verify the same name isn't already registered */
1646 if (ctl_find_by_name(userkctl
->ctl_name
) != NULL
) {
1647 kctl_delete_ref(kctl
->kctlref
);
1648 lck_mtx_unlock(ctl_mtx
);
1653 /* Start with 1 in case the list is empty */
1655 kctl_next
= TAILQ_LAST(&ctl_head
, kctl_list
);
1657 if (kctl_next
!= NULL
) {
1658 /* List was not empty, add one to the last item */
1659 id
= kctl_next
->id
+ 1;
1663 * If this wrapped the id number, start looking at
1664 * the front of the list for an unused id.
1667 /* Find the next unused ID */
1670 TAILQ_FOREACH(kctl_next
, &ctl_head
, next
) {
1671 if (kctl_next
->id
> id
) {
1672 /* We found a gap */
1676 id
= kctl_next
->id
+ 1;
1681 userkctl
->ctl_id
= id
;
1683 kctl
->reg_unit
= -1;
1685 TAILQ_FOREACH(kctl_next
, &ctl_head
, next
) {
1686 if (kctl_next
->id
> userkctl
->ctl_id
) {
1691 if (ctl_find_by_id_unit(userkctl
->ctl_id
, userkctl
->ctl_unit
)) {
1692 kctl_delete_ref(kctl
->kctlref
);
1693 lck_mtx_unlock(ctl_mtx
);
1697 kctl
->id
= userkctl
->ctl_id
;
1698 kctl
->reg_unit
= userkctl
->ctl_unit
;
1701 is_extended
= (userkctl
->ctl_flags
& CTL_FLAG_REG_EXTENDED
);
1703 strlcpy(kctl
->name
, userkctl
->ctl_name
, MAX_KCTL_NAME
);
1704 kctl
->flags
= userkctl
->ctl_flags
;
1707 * Let the caller know the default send and receive sizes
1709 if (userkctl
->ctl_sendsize
== 0) {
1710 kctl
->sendbufsize
= CTL_SENDSIZE
;
1711 userkctl
->ctl_sendsize
= kctl
->sendbufsize
;
1713 kctl
->sendbufsize
= userkctl
->ctl_sendsize
;
1715 if (userkctl
->ctl_recvsize
== 0) {
1716 kctl
->recvbufsize
= CTL_RECVSIZE
;
1717 userkctl
->ctl_recvsize
= kctl
->recvbufsize
;
1719 kctl
->recvbufsize
= userkctl
->ctl_recvsize
;
1722 kctl
->bind
= userkctl
->ctl_bind
;
1723 kctl
->connect
= userkctl
->ctl_connect
;
1724 kctl
->disconnect
= userkctl
->ctl_disconnect
;
1725 kctl
->send
= userkctl
->ctl_send
;
1726 kctl
->setopt
= userkctl
->ctl_setopt
;
1727 kctl
->getopt
= userkctl
->ctl_getopt
;
1729 kctl
->rcvd
= userkctl
->ctl_rcvd
;
1730 kctl
->send_list
= userkctl
->ctl_send_list
;
1733 TAILQ_INIT(&kctl
->kcb_head
);
1736 TAILQ_INSERT_BEFORE(kctl_next
, kctl
, next
);
1738 TAILQ_INSERT_TAIL(&ctl_head
, kctl
, next
);
1741 kctlstat
.kcs_reg_count
++;
1742 kctlstat
.kcs_gencnt
++;
1744 lck_mtx_unlock(ctl_mtx
);
1746 *kctlref
= kctl
->kctlref
;
1748 ctl_post_msg(KEV_CTL_REGISTERED
, kctl
->id
);
1753 ctl_deregister(void *kctlref
)
1757 lck_mtx_lock(ctl_mtx
);
1758 if ((kctl
= kctl_from_ref(kctlref
)) == NULL
) {
1759 kctlstat
.kcs_bad_kctlref
++;
1760 lck_mtx_unlock(ctl_mtx
);
1761 if (ctl_debug
!= 0) {
1762 printf("%s invalid kctlref %p\n",
1768 if (!TAILQ_EMPTY(&kctl
->kcb_head
)) {
1769 lck_mtx_unlock(ctl_mtx
);
1773 TAILQ_REMOVE(&ctl_head
, kctl
, next
);
1775 kctlstat
.kcs_reg_count
--;
1776 kctlstat
.kcs_gencnt
++;
1778 kctl_delete_ref(kctl
->kctlref
);
1779 lck_mtx_unlock(ctl_mtx
);
1781 ctl_post_msg(KEV_CTL_DEREGISTERED
, kctl
->id
);
1787 * Must be called with global ctl_mtx lock taked
1789 static struct kctl
*
1790 ctl_find_by_name(const char *name
)
1794 lck_mtx_assert(ctl_mtx
, LCK_MTX_ASSERT_OWNED
);
1796 TAILQ_FOREACH(kctl
, &ctl_head
, next
)
1797 if (strncmp(kctl
->name
, name
, sizeof(kctl
->name
)) == 0) {
1805 ctl_id_by_name(const char *name
)
1807 u_int32_t ctl_id
= 0;
1810 lck_mtx_lock(ctl_mtx
);
1811 kctl
= ctl_find_by_name(name
);
1815 lck_mtx_unlock(ctl_mtx
);
1821 ctl_name_by_id(u_int32_t id
, char *out_name
, size_t maxsize
)
1826 lck_mtx_lock(ctl_mtx
);
1827 TAILQ_FOREACH(kctl
, &ctl_head
, next
) {
1828 if (kctl
->id
== id
) {
1834 if (maxsize
> MAX_KCTL_NAME
) {
1835 maxsize
= MAX_KCTL_NAME
;
1837 strlcpy(out_name
, kctl
->name
, maxsize
);
1840 lck_mtx_unlock(ctl_mtx
);
1842 return found
? 0 : ENOENT
;
1846 * Must be called with global ctl_mtx lock taked
1849 static struct kctl
*
1850 ctl_find_by_id_unit(u_int32_t id
, u_int32_t unit
)
1854 lck_mtx_assert(ctl_mtx
, LCK_MTX_ASSERT_OWNED
);
1856 TAILQ_FOREACH(kctl
, &ctl_head
, next
) {
1857 if (kctl
->id
== id
&& (kctl
->flags
& CTL_FLAG_REG_ID_UNIT
) == 0) {
1859 } else if (kctl
->id
== id
&& kctl
->reg_unit
== unit
) {
1867 * Must be called with kernel controller lock taken
1869 static struct ctl_cb
*
1870 kcb_find(struct kctl
*kctl
, u_int32_t unit
)
1874 lck_mtx_assert(ctl_mtx
, LCK_MTX_ASSERT_OWNED
);
1876 TAILQ_FOREACH(kcb
, &kctl
->kcb_head
, next
)
1877 if (kcb
->sac
.sc_unit
== unit
) {
1884 static struct socket
*
1885 kcb_find_socket(kern_ctl_ref kctlref
, u_int32_t unit
, u_int32_t
*kctlflags
)
1887 struct socket
*so
= NULL
;
1893 lr_saved
= __builtin_return_address(0);
1895 lck_mtx_lock(ctl_mtx
);
1897 * First validate the kctlref
1899 if ((kctl
= kctl_from_ref(kctlref
)) == NULL
) {
1900 kctlstat
.kcs_bad_kctlref
++;
1901 lck_mtx_unlock(ctl_mtx
);
1902 if (ctl_debug
!= 0) {
1903 printf("%s invalid kctlref %p\n",
1909 kcb
= kcb_find(kctl
, unit
);
1910 if (kcb
== NULL
|| kcb
->kctl
!= kctl
|| (so
= kcb
->so
) == NULL
) {
1911 lck_mtx_unlock(ctl_mtx
);
1915 * This prevents the socket from being closed
1919 * Respect lock ordering: socket before ctl_mtx
1921 lck_mtx_unlock(ctl_mtx
);
1925 * The socket lock history is more useful if we store
1926 * the address of the caller.
1928 i
= (so
->next_lock_lr
+ SO_LCKDBG_MAX
- 1) % SO_LCKDBG_MAX
;
1929 so
->lock_lr
[i
] = lr_saved
;
1931 lck_mtx_lock(ctl_mtx
);
1933 if ((kctl
= kctl_from_ref(kctlref
)) == NULL
|| kcb
->kctl
== NULL
) {
1934 lck_mtx_unlock(ctl_mtx
);
1935 socket_unlock(so
, 1);
1937 lck_mtx_lock(ctl_mtx
);
1938 } else if (kctlflags
!= NULL
) {
1939 *kctlflags
= kctl
->flags
;
1943 if (kcb
->usecount
== 0) {
1944 wakeup((event_t
)&kcb
->usecount
);
1947 lck_mtx_unlock(ctl_mtx
);
1953 ctl_post_msg(u_int32_t event_code
, u_int32_t id
)
1955 struct ctl_event_data ctl_ev_data
;
1956 struct kev_msg ev_msg
;
1958 lck_mtx_assert(ctl_mtx
, LCK_MTX_ASSERT_NOTOWNED
);
1960 bzero(&ev_msg
, sizeof(struct kev_msg
));
1961 ev_msg
.vendor_code
= KEV_VENDOR_APPLE
;
1963 ev_msg
.kev_class
= KEV_SYSTEM_CLASS
;
1964 ev_msg
.kev_subclass
= KEV_CTL_SUBCLASS
;
1965 ev_msg
.event_code
= event_code
;
1967 /* common nke subclass data */
1968 bzero(&ctl_ev_data
, sizeof(ctl_ev_data
));
1969 ctl_ev_data
.ctl_id
= id
;
1970 ev_msg
.dv
[0].data_ptr
= &ctl_ev_data
;
1971 ev_msg
.dv
[0].data_length
= sizeof(ctl_ev_data
);
1973 ev_msg
.dv
[1].data_length
= 0;
1975 kev_post_msg(&ev_msg
);
1979 ctl_lock(struct socket
*so
, int refcount
, void *lr
)
1984 lr_saved
= __builtin_return_address(0);
1989 if (so
->so_pcb
!= NULL
) {
1990 lck_mtx_lock(((struct ctl_cb
*)so
->so_pcb
)->mtx
);
1992 panic("ctl_lock: so=%p NO PCB! lr=%p lrh= %s\n",
1993 so
, lr_saved
, solockhistory_nr(so
));
1997 if (so
->so_usecount
< 0) {
1998 panic("ctl_lock: so=%p so_pcb=%p lr=%p ref=%x lrh= %s\n",
1999 so
, so
->so_pcb
, lr_saved
, so
->so_usecount
,
2000 solockhistory_nr(so
));
2008 so
->lock_lr
[so
->next_lock_lr
] = lr_saved
;
2009 so
->next_lock_lr
= (so
->next_lock_lr
+ 1) % SO_LCKDBG_MAX
;
2014 ctl_unlock(struct socket
*so
, int refcount
, void *lr
)
2017 lck_mtx_t
*mutex_held
;
2020 lr_saved
= __builtin_return_address(0);
2025 #if (MORE_KCTLLOCK_DEBUG && (DEVELOPMENT || DEBUG))
2026 printf("ctl_unlock: so=%llx sopcb=%x lock=%llx ref=%u lr=%llx\n",
2027 (uint64_t)VM_KERNEL_ADDRPERM(so
),
2028 (uint64_t)VM_KERNEL_ADDRPERM(so
->so_pcb
,
2029 (uint64_t)VM_KERNEL_ADDRPERM(((struct ctl_cb
*)so
->so_pcb
)->mtx
),
2030 so
->so_usecount
, (uint64_t)VM_KERNEL_ADDRPERM(lr_saved
));
2031 #endif /* (MORE_KCTLLOCK_DEBUG && (DEVELOPMENT || DEBUG)) */
2036 if (so
->so_usecount
< 0) {
2037 panic("ctl_unlock: so=%p usecount=%x lrh= %s\n",
2038 so
, so
->so_usecount
, solockhistory_nr(so
));
2041 if (so
->so_pcb
== NULL
) {
2042 panic("ctl_unlock: so=%p NO PCB usecount=%x lr=%p lrh= %s\n",
2043 so
, so
->so_usecount
, (void *)lr_saved
,
2044 solockhistory_nr(so
));
2047 mutex_held
= ((struct ctl_cb
*)so
->so_pcb
)->mtx
;
2049 lck_mtx_assert(mutex_held
, LCK_MTX_ASSERT_OWNED
);
2050 so
->unlock_lr
[so
->next_unlock_lr
] = lr_saved
;
2051 so
->next_unlock_lr
= (so
->next_unlock_lr
+ 1) % SO_LCKDBG_MAX
;
2052 lck_mtx_unlock(mutex_held
);
2054 if (so
->so_usecount
== 0) {
2055 ctl_sofreelastref(so
);
2062 ctl_getlock(struct socket
*so
, int flags
)
2064 #pragma unused(flags)
2065 struct ctl_cb
*kcb
= (struct ctl_cb
*)so
->so_pcb
;
2068 if (so
->so_usecount
< 0) {
2069 panic("ctl_getlock: so=%p usecount=%x lrh= %s\n",
2070 so
, so
->so_usecount
, solockhistory_nr(so
));
2074 panic("ctl_getlock: so=%p NULL NO so_pcb %s\n",
2075 so
, solockhistory_nr(so
));
2076 return so
->so_proto
->pr_domain
->dom_mtx
;
2080 __private_extern__
int
2081 kctl_reg_list SYSCTL_HANDLER_ARGS
2083 #pragma unused(oidp, arg1, arg2)
2086 struct xsystmgen xsg
;
2089 size_t item_size
= ROUNDUP64(sizeof(struct xkctl_reg
));
2091 buf
= _MALLOC(item_size
, M_TEMP
, M_WAITOK
| M_ZERO
);
2096 lck_mtx_lock(ctl_mtx
);
2098 n
= kctlstat
.kcs_reg_count
;
2100 if (req
->oldptr
== USER_ADDR_NULL
) {
2101 req
->oldidx
= (n
+ n
/ 8) * sizeof(struct xkctl_reg
);
2104 if (req
->newptr
!= USER_ADDR_NULL
) {
2108 bzero(&xsg
, sizeof(xsg
));
2109 xsg
.xg_len
= sizeof(xsg
);
2111 xsg
.xg_gen
= kctlstat
.kcs_gencnt
;
2112 xsg
.xg_sogen
= so_gencnt
;
2113 error
= SYSCTL_OUT(req
, &xsg
, sizeof(xsg
));
2118 * We are done if there is no pcb
2125 for (i
= 0, kctl
= TAILQ_FIRST(&ctl_head
);
2126 i
< n
&& kctl
!= NULL
;
2127 i
++, kctl
= TAILQ_NEXT(kctl
, next
)) {
2128 struct xkctl_reg
*xkr
= (struct xkctl_reg
*)buf
;
2130 u_int32_t pcbcount
= 0;
2132 TAILQ_FOREACH(kcb
, &kctl
->kcb_head
, next
)
2135 bzero(buf
, item_size
);
2137 xkr
->xkr_len
= sizeof(struct xkctl_reg
);
2138 xkr
->xkr_kind
= XSO_KCREG
;
2139 xkr
->xkr_id
= kctl
->id
;
2140 xkr
->xkr_reg_unit
= kctl
->reg_unit
;
2141 xkr
->xkr_flags
= kctl
->flags
;
2142 xkr
->xkr_kctlref
= (uint64_t)(kctl
->kctlref
);
2143 xkr
->xkr_recvbufsize
= kctl
->recvbufsize
;
2144 xkr
->xkr_sendbufsize
= kctl
->sendbufsize
;
2145 xkr
->xkr_lastunit
= kctl
->lastunit
;
2146 xkr
->xkr_pcbcount
= pcbcount
;
2147 xkr
->xkr_connect
= (uint64_t)VM_KERNEL_UNSLIDE(kctl
->connect
);
2148 xkr
->xkr_disconnect
=
2149 (uint64_t)VM_KERNEL_UNSLIDE(kctl
->disconnect
);
2150 xkr
->xkr_send
= (uint64_t)VM_KERNEL_UNSLIDE(kctl
->send
);
2151 xkr
->xkr_send_list
=
2152 (uint64_t)VM_KERNEL_UNSLIDE(kctl
->send_list
);
2153 xkr
->xkr_setopt
= (uint64_t)VM_KERNEL_UNSLIDE(kctl
->setopt
);
2154 xkr
->xkr_getopt
= (uint64_t)VM_KERNEL_UNSLIDE(kctl
->getopt
);
2155 xkr
->xkr_rcvd
= (uint64_t)VM_KERNEL_UNSLIDE(kctl
->rcvd
);
2156 strlcpy(xkr
->xkr_name
, kctl
->name
, sizeof(xkr
->xkr_name
));
2158 error
= SYSCTL_OUT(req
, buf
, item_size
);
2163 * Give the user an updated idea of our state.
2164 * If the generation differs from what we told
2165 * her before, she knows that something happened
2166 * while we were processing this request, and it
2167 * might be necessary to retry.
2169 bzero(&xsg
, sizeof(xsg
));
2170 xsg
.xg_len
= sizeof(xsg
);
2172 xsg
.xg_gen
= kctlstat
.kcs_gencnt
;
2173 xsg
.xg_sogen
= so_gencnt
;
2174 error
= SYSCTL_OUT(req
, &xsg
, sizeof(xsg
));
2181 lck_mtx_unlock(ctl_mtx
);
2190 __private_extern__
int
2191 kctl_pcblist SYSCTL_HANDLER_ARGS
2193 #pragma unused(oidp, arg1, arg2)
2196 struct xsystmgen xsg
;
2199 size_t item_size
= ROUNDUP64(sizeof(struct xkctlpcb
)) +
2200 ROUNDUP64(sizeof(struct xsocket_n
)) +
2201 2 * ROUNDUP64(sizeof(struct xsockbuf_n
)) +
2202 ROUNDUP64(sizeof(struct xsockstat_n
));
2204 buf
= _MALLOC(item_size
, M_TEMP
, M_WAITOK
| M_ZERO
);
2209 lck_mtx_lock(ctl_mtx
);
2211 n
= kctlstat
.kcs_pcbcount
;
2213 if (req
->oldptr
== USER_ADDR_NULL
) {
2214 req
->oldidx
= (n
+ n
/ 8) * item_size
;
2217 if (req
->newptr
!= USER_ADDR_NULL
) {
2221 bzero(&xsg
, sizeof(xsg
));
2222 xsg
.xg_len
= sizeof(xsg
);
2224 xsg
.xg_gen
= kctlstat
.kcs_gencnt
;
2225 xsg
.xg_sogen
= so_gencnt
;
2226 error
= SYSCTL_OUT(req
, &xsg
, sizeof(xsg
));
2231 * We are done if there is no pcb
2238 for (i
= 0, kctl
= TAILQ_FIRST(&ctl_head
);
2239 i
< n
&& kctl
!= NULL
;
2240 kctl
= TAILQ_NEXT(kctl
, next
)) {
2243 for (kcb
= TAILQ_FIRST(&kctl
->kcb_head
);
2244 i
< n
&& kcb
!= NULL
;
2245 i
++, kcb
= TAILQ_NEXT(kcb
, next
)) {
2246 struct xkctlpcb
*xk
= (struct xkctlpcb
*)buf
;
2247 struct xsocket_n
*xso
= (struct xsocket_n
*)
2248 ADVANCE64(xk
, sizeof(*xk
));
2249 struct xsockbuf_n
*xsbrcv
= (struct xsockbuf_n
*)
2250 ADVANCE64(xso
, sizeof(*xso
));
2251 struct xsockbuf_n
*xsbsnd
= (struct xsockbuf_n
*)
2252 ADVANCE64(xsbrcv
, sizeof(*xsbrcv
));
2253 struct xsockstat_n
*xsostats
= (struct xsockstat_n
*)
2254 ADVANCE64(xsbsnd
, sizeof(*xsbsnd
));
2256 bzero(buf
, item_size
);
2258 xk
->xkp_len
= sizeof(struct xkctlpcb
);
2259 xk
->xkp_kind
= XSO_KCB
;
2260 xk
->xkp_unit
= kcb
->sac
.sc_unit
;
2261 xk
->xkp_kctpcb
= (uint64_t)VM_KERNEL_ADDRPERM(kcb
);
2262 xk
->xkp_kctlref
= (uint64_t)VM_KERNEL_ADDRPERM(kctl
);
2263 xk
->xkp_kctlid
= kctl
->id
;
2264 strlcpy(xk
->xkp_kctlname
, kctl
->name
,
2265 sizeof(xk
->xkp_kctlname
));
2267 sotoxsocket_n(kcb
->so
, xso
);
2268 sbtoxsockbuf_n(kcb
->so
?
2269 &kcb
->so
->so_rcv
: NULL
, xsbrcv
);
2270 sbtoxsockbuf_n(kcb
->so
?
2271 &kcb
->so
->so_snd
: NULL
, xsbsnd
);
2272 sbtoxsockstat_n(kcb
->so
, xsostats
);
2274 error
= SYSCTL_OUT(req
, buf
, item_size
);
2280 * Give the user an updated idea of our state.
2281 * If the generation differs from what we told
2282 * her before, she knows that something happened
2283 * while we were processing this request, and it
2284 * might be necessary to retry.
2286 bzero(&xsg
, sizeof(xsg
));
2287 xsg
.xg_len
= sizeof(xsg
);
2289 xsg
.xg_gen
= kctlstat
.kcs_gencnt
;
2290 xsg
.xg_sogen
= so_gencnt
;
2291 error
= SYSCTL_OUT(req
, &xsg
, sizeof(xsg
));
2298 lck_mtx_unlock(ctl_mtx
);
2304 kctl_getstat SYSCTL_HANDLER_ARGS
2306 #pragma unused(oidp, arg1, arg2)
2309 lck_mtx_lock(ctl_mtx
);
2311 if (req
->newptr
!= USER_ADDR_NULL
) {
2315 if (req
->oldptr
== USER_ADDR_NULL
) {
2316 req
->oldidx
= sizeof(struct kctlstat
);
2320 error
= SYSCTL_OUT(req
, &kctlstat
,
2321 MIN(sizeof(struct kctlstat
), req
->oldlen
));
2323 lck_mtx_unlock(ctl_mtx
);
2328 kctl_fill_socketinfo(struct socket
*so
, struct socket_info
*si
)
2330 struct ctl_cb
*kcb
= (struct ctl_cb
*)so
->so_pcb
;
2331 struct kern_ctl_info
*kcsi
=
2332 &si
->soi_proto
.pri_kern_ctl
;
2333 struct kctl
*kctl
= kcb
->kctl
;
2335 si
->soi_kind
= SOCKINFO_KERN_CTL
;
2341 kcsi
->kcsi_id
= kctl
->id
;
2342 kcsi
->kcsi_reg_unit
= kctl
->reg_unit
;
2343 kcsi
->kcsi_flags
= kctl
->flags
;
2344 kcsi
->kcsi_recvbufsize
= kctl
->recvbufsize
;
2345 kcsi
->kcsi_sendbufsize
= kctl
->sendbufsize
;
2346 kcsi
->kcsi_unit
= kcb
->sac
.sc_unit
;
2347 strlcpy(kcsi
->kcsi_name
, kctl
->name
, MAX_KCTL_NAME
);