1 /*
2 * Copyright (c) 1999-2020 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 /*
30 * Kernel Control domain - allows control connections to
31 * and to read/write data.
32 *
33 * Vincent Lubet, 040506
34 * Christophe Allie, 010928
35 * Justin C. Walker, 990319
36 */
37
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>
47 #include <sys/mbuf.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>
55
56 #include <mach/vm_types.h>
57
58 #include <kern/thread.h>
59
60 struct kctl {
61 TAILQ_ENTRY(kctl) next; /* controller chain */
62 kern_ctl_ref kctlref;
63
64 /* controller information provided when registering */
65 char name[MAX_KCTL_NAME]; /* unique identifier */
66 u_int32_t id;
67 u_int32_t reg_unit;
68
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 */
73
74 /* Dispatch functions */
75 ctl_setup_func setup; /* Setup contact */
76 ctl_bind_func bind; /* Prepare contact */
77 ctl_connect_func connect; /* Make contact */
78 ctl_disconnect_func disconnect; /* Break contact */
79 ctl_send_func send; /* Send data to nke */
80 ctl_send_list_func send_list; /* Send list of packets */
81 ctl_setopt_func setopt; /* set kctl configuration */
82 ctl_getopt_func getopt; /* get kctl configuration */
83 ctl_rcvd_func rcvd; /* Notify nke when client reads data */
84
85 TAILQ_HEAD(, ctl_cb) kcb_head;
86 u_int32_t lastunit;
87 };
88
89 #if DEVELOPMENT || DEBUG
90 enum ctl_status {
91 KCTL_DISCONNECTED = 0,
92 KCTL_CONNECTING = 1,
93 KCTL_CONNECTED = 2
94 };
95 #endif /* DEVELOPMENT || DEBUG */
96
97 struct ctl_cb {
98 TAILQ_ENTRY(ctl_cb) next; /* controller chain */
99 lck_mtx_t mtx;
100 struct socket *so; /* controlling socket */
101 struct kctl *kctl; /* back pointer to controller */
102 void *userdata;
103 struct sockaddr_ctl sac;
104 u_int32_t usecount;
105 u_int32_t kcb_usecount;
106 u_int32_t require_clearing_count;
107 #if DEVELOPMENT || DEBUG
108 enum ctl_status status;
109 #endif /* DEVELOPMENT || DEBUG */
110 };
111
112 #ifndef ROUNDUP64
113 #define ROUNDUP64(x) P2ROUNDUP((x), sizeof (u_int64_t))
114 #endif
115
116 #ifndef ADVANCE64
117 #define ADVANCE64(p, n) (void*)((char *)(p) + ROUNDUP64(n))
118 #endif
119
120 /*
121 * Definitions and vars for we support
122 */
123
124 #define CTL_SENDSIZE (2 * 1024) /* default buffer size */
125 #define CTL_RECVSIZE (8 * 1024) /* default buffer size */
126
127 /*
128 * Definitions and vars for we support
129 */
130
131 const u_int32_t ctl_maxunit = 65536;
132 static LCK_ATTR_DECLARE(ctl_lck_attr, 0, 0);
133 static LCK_GRP_DECLARE(ctl_lck_grp, "Kernel Control Protocol");
134 static LCK_MTX_DECLARE_ATTR(ctl_mtx, &ctl_lck_grp, &ctl_lck_attr);
135
136 /* all the controllers are chained */
137 TAILQ_HEAD(kctl_list, kctl) ctl_head = TAILQ_HEAD_INITIALIZER(ctl_head);
138
139 static int ctl_attach(struct socket *, int, struct proc *);
140 static int ctl_detach(struct socket *);
141 static int ctl_sofreelastref(struct socket *so);
142 static int ctl_bind(struct socket *, struct sockaddr *, struct proc *);
143 static int ctl_connect(struct socket *, struct sockaddr *, struct proc *);
144 static int ctl_disconnect(struct socket *);
145 static int ctl_ioctl(struct socket *so, u_long cmd, caddr_t data,
146 struct ifnet *ifp, struct proc *p);
147 static int ctl_send(struct socket *, int, struct mbuf *,
148 struct sockaddr *, struct mbuf *, struct proc *);
149 static int ctl_send_list(struct socket *, int, struct mbuf *,
150 struct sockaddr *, struct mbuf *, struct proc *);
151 static int ctl_ctloutput(struct socket *, struct sockopt *);
152 static int ctl_peeraddr(struct socket *so, struct sockaddr **nam);
153 static int ctl_usr_rcvd(struct socket *so, int flags);
154
155 static struct kctl *ctl_find_by_name(const char *);
156 static struct kctl *ctl_find_by_id_unit(u_int32_t id, u_int32_t unit);
157
158 static struct socket *kcb_find_socket(kern_ctl_ref kctlref, u_int32_t unit,
159 u_int32_t *);
160 static struct ctl_cb *kcb_find(struct kctl *, u_int32_t unit);
161 static void ctl_post_msg(u_int32_t event_code, u_int32_t id);
162
163 static int ctl_lock(struct socket *, int, void *);
164 static int ctl_unlock(struct socket *, int, void *);
165 static lck_mtx_t * ctl_getlock(struct socket *, int);
166
167 static struct pr_usrreqs ctl_usrreqs = {
168 .pru_attach = ctl_attach,
169 .pru_bind = ctl_bind,
170 .pru_connect = ctl_connect,
171 .pru_control = ctl_ioctl,
172 .pru_detach = ctl_detach,
173 .pru_disconnect = ctl_disconnect,
174 .pru_peeraddr = ctl_peeraddr,
175 .pru_rcvd = ctl_usr_rcvd,
176 .pru_send = ctl_send,
177 .pru_send_list = ctl_send_list,
178 .pru_sosend = sosend,
179 .pru_sosend_list = sosend_list,
180 .pru_soreceive = soreceive,
181 .pru_soreceive_list = soreceive_list,
182 };
183
184 static struct protosw kctlsw[] = {
185 {
186 .pr_type = SOCK_DGRAM,
187 .pr_protocol = SYSPROTO_CONTROL,
188 .pr_flags = PR_ATOMIC | PR_CONNREQUIRED | PR_PCBLOCK | PR_WANTRCVD,
189 .pr_ctloutput = ctl_ctloutput,
190 .pr_usrreqs = &ctl_usrreqs,
191 .pr_lock = ctl_lock,
192 .pr_unlock = ctl_unlock,
193 .pr_getlock = ctl_getlock,
194 },
195 {
196 .pr_type = SOCK_STREAM,
197 .pr_protocol = SYSPROTO_CONTROL,
198 .pr_flags = PR_CONNREQUIRED | PR_PCBLOCK | PR_WANTRCVD,
199 .pr_ctloutput = ctl_ctloutput,
200 .pr_usrreqs = &ctl_usrreqs,
201 .pr_lock = ctl_lock,
202 .pr_unlock = ctl_unlock,
203 .pr_getlock = ctl_getlock,
204 }
205 };
206
207 __private_extern__ int kctl_reg_list SYSCTL_HANDLER_ARGS;
208 __private_extern__ int kctl_pcblist SYSCTL_HANDLER_ARGS;
209 __private_extern__ int kctl_getstat SYSCTL_HANDLER_ARGS;
210
211
212 SYSCTL_NODE(_net_systm, OID_AUTO, kctl,
213 CTLFLAG_RW | CTLFLAG_LOCKED, 0, "Kernel control family");
214
215 struct kctlstat kctlstat;
216 SYSCTL_PROC(_net_systm_kctl, OID_AUTO, stats,
217 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0,
218 kctl_getstat, "S,kctlstat", "");
219
220 SYSCTL_PROC(_net_systm_kctl, OID_AUTO, reg_list,
221 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0,
222 kctl_reg_list, "S,xkctl_reg", "");
223
224 SYSCTL_PROC(_net_systm_kctl, OID_AUTO, pcblist,
225 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0,
226 kctl_pcblist, "S,xkctlpcb", "");
227
228 u_int32_t ctl_autorcvbuf_max = 256 * 1024;
229 SYSCTL_INT(_net_systm_kctl, OID_AUTO, autorcvbufmax,
230 CTLFLAG_RW | CTLFLAG_LOCKED, &ctl_autorcvbuf_max, 0, "");
231
232 u_int32_t ctl_autorcvbuf_high = 0;
233 SYSCTL_INT(_net_systm_kctl, OID_AUTO, autorcvbufhigh,
234 CTLFLAG_RD | CTLFLAG_LOCKED, &ctl_autorcvbuf_high, 0, "");
235
236 u_int32_t ctl_debug = 0;
237 SYSCTL_INT(_net_systm_kctl, OID_AUTO, debug,
238 CTLFLAG_RW | CTLFLAG_LOCKED, &ctl_debug, 0, "");
239
240 #if DEVELOPMENT || DEBUG
241 u_int32_t ctl_panic_debug = 0;
242 SYSCTL_INT(_net_systm_kctl, OID_AUTO, panicdebug,
243 CTLFLAG_RW | CTLFLAG_LOCKED, &ctl_panic_debug, 0, "");
244 #endif /* DEVELOPMENT || DEBUG */
245
246 #define KCTL_TBL_INC 16
247
248 static uintptr_t kctl_tbl_size = 0;
249 static u_int32_t kctl_tbl_growing = 0;
250 static u_int32_t kctl_tbl_growing_waiting = 0;
251 static uintptr_t kctl_tbl_count = 0;
252 static struct kctl **kctl_table = NULL;
253 static uintptr_t kctl_ref_gencnt = 0;
254
255 static void kctl_tbl_grow(void);
256 static kern_ctl_ref kctl_make_ref(struct kctl *kctl);
257 static void kctl_delete_ref(kern_ctl_ref);
258 static struct kctl *kctl_from_ref(kern_ctl_ref);
259
260 /*
261 * Install the protosw's for the Kernel Control manager.
262 */
263 __private_extern__ void
kern_control_init(struct domain * dp)264 kern_control_init(struct domain *dp)
265 {
266 struct protosw *pr;
267 int i;
268 int kctl_proto_count = (sizeof(kctlsw) / sizeof(struct protosw));
269
270 VERIFY(!(dp->dom_flags & DOM_INITIALIZED));
271 VERIFY(dp == systemdomain);
272
273 for (i = 0, pr = &kctlsw[0]; i < kctl_proto_count; i++, pr++) {
274 net_add_proto(pr, dp, 1);
275 }
276 }
277
278 static void
kcb_delete(struct ctl_cb * kcb)279 kcb_delete(struct ctl_cb *kcb)
280 {
281 if (kcb != 0) {
282 lck_mtx_destroy(&kcb->mtx, &ctl_lck_grp);
283 kfree_type(struct ctl_cb, kcb);
284 }
285 }
286
287 /*
288 * Kernel Controller user-request functions
289 * attach function must exist and succeed
290 * detach not necessary
291 * we need a pcb for the per socket mutex
292 */
293 static int
ctl_attach(struct socket * so,int proto,struct proc * p)294 ctl_attach(struct socket *so, int proto, struct proc *p)
295 {
296 #pragma unused(proto, p)
297 struct ctl_cb *kcb = 0;
298
299 kcb = kalloc_type(struct ctl_cb, Z_WAITOK | Z_ZERO | Z_NOFAIL);
300
301 lck_mtx_init(&kcb->mtx, &ctl_lck_grp, &ctl_lck_attr);
302 kcb->so = so;
303 so->so_pcb = (caddr_t)kcb;
304
305 return 0;
306 }
307
308 static int
ctl_sofreelastref(struct socket * so)309 ctl_sofreelastref(struct socket *so)
310 {
311 struct ctl_cb *kcb = (struct ctl_cb *)so->so_pcb;
312
313 so->so_pcb = 0;
314
315 if (kcb != 0) {
316 struct kctl *kctl;
317 if ((kctl = kcb->kctl) != 0) {
318 lck_mtx_lock(&ctl_mtx);
319 TAILQ_REMOVE(&kctl->kcb_head, kcb, next);
320 kctlstat.kcs_pcbcount--;
321 kctlstat.kcs_gencnt++;
322 lck_mtx_unlock(&ctl_mtx);
323 }
324 kcb_delete(kcb);
325 }
326 sofreelastref(so, 1);
327 return 0;
328 }
329
330 /*
331 * Use this function and ctl_kcb_require_clearing to serialize
332 * critical calls into the kctl subsystem
333 */
334 static void
ctl_kcb_increment_use_count(struct ctl_cb * kcb,lck_mtx_t * mutex_held)335 ctl_kcb_increment_use_count(struct ctl_cb *kcb, lck_mtx_t *mutex_held)
336 {
337 LCK_MTX_ASSERT(mutex_held, LCK_MTX_ASSERT_OWNED);
338 while (kcb->require_clearing_count > 0) {
339 msleep(&kcb->require_clearing_count, mutex_held, PSOCK | PCATCH, "kcb_require_clearing", NULL);
340 }
341 kcb->kcb_usecount++;
342 }
343
344 static void
ctl_kcb_require_clearing(struct ctl_cb * kcb,lck_mtx_t * mutex_held)345 ctl_kcb_require_clearing(struct ctl_cb *kcb, lck_mtx_t *mutex_held)
346 {
347 assert(kcb->kcb_usecount != 0);
348 kcb->require_clearing_count++;
349 kcb->kcb_usecount--;
350 while (kcb->kcb_usecount > 0) { // we need to wait until no one else is running
351 msleep(&kcb->kcb_usecount, mutex_held, PSOCK | PCATCH, "kcb_usecount", NULL);
352 }
353 kcb->kcb_usecount++;
354 }
355
356 static void
ctl_kcb_done_clearing(struct ctl_cb * kcb)357 ctl_kcb_done_clearing(struct ctl_cb *kcb)
358 {
359 assert(kcb->require_clearing_count != 0);
360 kcb->require_clearing_count--;
361 wakeup((caddr_t)&kcb->require_clearing_count);
362 }
363
364 static void
ctl_kcb_decrement_use_count(struct ctl_cb * kcb)365 ctl_kcb_decrement_use_count(struct ctl_cb *kcb)
366 {
367 assert(kcb->kcb_usecount != 0);
368 kcb->kcb_usecount--;
369 wakeup((caddr_t)&kcb->kcb_usecount);
370 }
371
372 static int
ctl_detach(struct socket * so)373 ctl_detach(struct socket *so)
374 {
375 struct ctl_cb *kcb = (struct ctl_cb *)so->so_pcb;
376
377 if (kcb == 0) {
378 return 0;
379 }
380
381 lck_mtx_t *mtx_held = socket_getlock(so, PR_F_WILLUNLOCK);
382 ctl_kcb_increment_use_count(kcb, mtx_held);
383 ctl_kcb_require_clearing(kcb, mtx_held);
384
385 if (kcb->kctl != NULL && kcb->kctl->bind != NULL &&
386 kcb->userdata != NULL && !(so->so_state & SS_ISCONNECTED)) {
387 // The unit was bound, but not connected
388 // Invoke the disconnected call to cleanup
389 if (kcb->kctl->disconnect != NULL) {
390 socket_unlock(so, 0);
391 (*kcb->kctl->disconnect)(kcb->kctl->kctlref,
392 kcb->sac.sc_unit, kcb->userdata);
393 socket_lock(so, 0);
394 }
395 }
396
397 soisdisconnected(so);
398 #if DEVELOPMENT || DEBUG
399 kcb->status = KCTL_DISCONNECTED;
400 #endif /* DEVELOPMENT || DEBUG */
401 so->so_flags |= SOF_PCBCLEARING;
402 ctl_kcb_done_clearing(kcb);
403 ctl_kcb_decrement_use_count(kcb);
404 return 0;
405 }
406
407 static int
ctl_setup_kctl(struct socket * so,struct sockaddr * nam,struct proc * p)408 ctl_setup_kctl(struct socket *so, struct sockaddr *nam, struct proc *p)
409 {
410 struct kctl *kctl = NULL;
411 int error = 0;
412 struct sockaddr_ctl sa;
413 struct ctl_cb *kcb = (struct ctl_cb *)so->so_pcb;
414 struct ctl_cb *kcb_next = NULL;
415 u_quad_t sbmaxsize;
416 u_int32_t recvbufsize, sendbufsize;
417
418 if (kcb == 0) {
419 panic("ctl_setup_kctl so_pcb null");
420 }
421
422 if (kcb->kctl != NULL) {
423 // Already set up, skip
424 return 0;
425 }
426
427 if (nam->sa_len != sizeof(struct sockaddr_ctl)) {
428 return EINVAL;
429 }
430
431 bcopy(nam, &sa, sizeof(struct sockaddr_ctl));
432
433 lck_mtx_lock(&ctl_mtx);
434 kctl = ctl_find_by_id_unit(sa.sc_id, sa.sc_unit);
435 if (kctl == NULL) {
436 lck_mtx_unlock(&ctl_mtx);
437 return ENOENT;
438 }
439
440 if (((kctl->flags & CTL_FLAG_REG_SOCK_STREAM) &&
441 (so->so_type != SOCK_STREAM)) ||
442 (!(kctl->flags & CTL_FLAG_REG_SOCK_STREAM) &&
443 (so->so_type != SOCK_DGRAM))) {
444 lck_mtx_unlock(&ctl_mtx);
445 return EPROTOTYPE;
446 }
447
448 if (kctl->flags & CTL_FLAG_PRIVILEGED) {
449 if (p == 0) {
450 lck_mtx_unlock(&ctl_mtx);
451 return EINVAL;
452 }
453 if (kauth_cred_issuser(kauth_cred_get()) == 0) {
454 lck_mtx_unlock(&ctl_mtx);
455 return EPERM;
456 }
457 }
458
459 if ((kctl->flags & CTL_FLAG_REG_ID_UNIT) || sa.sc_unit != 0) {
460 if (kcb_find(kctl, sa.sc_unit) != NULL) {
461 lck_mtx_unlock(&ctl_mtx);
462 return EBUSY;
463 }
464 } else if (kctl->setup != NULL) {
465 error = (*kctl->setup)(&sa.sc_unit, &kcb->userdata);
466 if (error != 0) {
467 lck_mtx_unlock(&ctl_mtx);
468 return error;
469 }
470 } else {
471 /* Find an unused ID, assumes control IDs are in order */
472 u_int32_t unit = 1;
473
474 TAILQ_FOREACH(kcb_next, &kctl->kcb_head, next) {
475 if (kcb_next->sac.sc_unit > unit) {
476 /* Found a gap, lets fill it in */
477 break;
478 }
479 unit = kcb_next->sac.sc_unit + 1;
480 if (unit == ctl_maxunit) {
481 break;
482 }
483 }
484
485 if (unit == ctl_maxunit) {
486 lck_mtx_unlock(&ctl_mtx);
487 return EBUSY;
488 }
489
490 sa.sc_unit = unit;
491 }
492
493 bcopy(&sa, &kcb->sac, sizeof(struct sockaddr_ctl));
494 kcb->kctl = kctl;
495 if (kcb_next != NULL) {
496 TAILQ_INSERT_BEFORE(kcb_next, kcb, next);
497 } else {
498 TAILQ_INSERT_TAIL(&kctl->kcb_head, kcb, next);
499 }
500 kctlstat.kcs_pcbcount++;
501 kctlstat.kcs_gencnt++;
502 kctlstat.kcs_connections++;
503 lck_mtx_unlock(&ctl_mtx);
504
505 /*
506 * rdar://15526688: Limit the send and receive sizes to sb_max
507 * by using the same scaling as sbreserve()
508 */
509 sbmaxsize = (u_quad_t)sb_max * MCLBYTES / (MSIZE + MCLBYTES);
510
511 if (kctl->sendbufsize > sbmaxsize) {
512 sendbufsize = (u_int32_t)sbmaxsize;
513 } else {
514 sendbufsize = kctl->sendbufsize;
515 }
516
517 if (kctl->recvbufsize > sbmaxsize) {
518 recvbufsize = (u_int32_t)sbmaxsize;
519 } else {
520 recvbufsize = kctl->recvbufsize;
521 }
522
523 error = soreserve(so, sendbufsize, recvbufsize);
524 if (error) {
525 if (ctl_debug) {
526 printf("%s - soreserve(%llx, %u, %u) error %d\n",
527 __func__, (uint64_t)VM_KERNEL_ADDRPERM(so),
528 sendbufsize, recvbufsize, error);
529 }
530 goto done;
531 }
532
533 done:
534 if (error) {
535 soisdisconnected(so);
536 #if DEVELOPMENT || DEBUG
537 kcb->status = KCTL_DISCONNECTED;
538 #endif /* DEVELOPMENT || DEBUG */
539 lck_mtx_lock(&ctl_mtx);
540 TAILQ_REMOVE(&kctl->kcb_head, kcb, next);
541 kcb->kctl = NULL;
542 kcb->sac.sc_unit = 0;
543 kctlstat.kcs_pcbcount--;
544 kctlstat.kcs_gencnt++;
545 kctlstat.kcs_conn_fail++;
546 lck_mtx_unlock(&ctl_mtx);
547 }
548 return error;
549 }
550
551 static int
ctl_bind(struct socket * so,struct sockaddr * nam,struct proc * p)552 ctl_bind(struct socket *so, struct sockaddr *nam, struct proc *p)
553 {
554 int error = 0;
555 struct ctl_cb *kcb = (struct ctl_cb *)so->so_pcb;
556
557 if (kcb == NULL) {
558 panic("ctl_bind so_pcb null");
559 }
560
561 lck_mtx_t *mtx_held = socket_getlock(so, PR_F_WILLUNLOCK);
562 ctl_kcb_increment_use_count(kcb, mtx_held);
563 ctl_kcb_require_clearing(kcb, mtx_held);
564
565 error = ctl_setup_kctl(so, nam, p);
566 if (error) {
567 goto out;
568 }
569
570 if (kcb->kctl == NULL) {
571 panic("ctl_bind kctl null");
572 }
573
574 if (kcb->kctl->bind == NULL) {
575 error = EINVAL;
576 goto out;
577 }
578
579 socket_unlock(so, 0);
580 error = (*kcb->kctl->bind)(kcb->kctl->kctlref, &kcb->sac, &kcb->userdata);
581 socket_lock(so, 0);
582
583 out:
584 ctl_kcb_done_clearing(kcb);
585 ctl_kcb_decrement_use_count(kcb);
586 return error;
587 }
588
589 static int
ctl_connect(struct socket * so,struct sockaddr * nam,struct proc * p)590 ctl_connect(struct socket *so, struct sockaddr *nam, struct proc *p)
591 {
592 int error = 0;
593 struct ctl_cb *kcb = (struct ctl_cb *)so->so_pcb;
594
595 if (kcb == NULL) {
596 panic("ctl_connect so_pcb null");
597 }
598
599 lck_mtx_t *mtx_held = socket_getlock(so, PR_F_WILLUNLOCK);
600 ctl_kcb_increment_use_count(kcb, mtx_held);
601 ctl_kcb_require_clearing(kcb, mtx_held);
602
603 #if DEVELOPMENT || DEBUG
604 if (kcb->status != KCTL_DISCONNECTED && ctl_panic_debug) {
605 panic("kctl already connecting/connected");
606 }
607 kcb->status = KCTL_CONNECTING;
608 #endif /* DEVELOPMENT || DEBUG */
609
610 error = ctl_setup_kctl(so, nam, p);
611 if (error) {
612 goto out;
613 }
614
615 if (kcb->kctl == NULL) {
616 panic("ctl_connect kctl null");
617 }
618
619 soisconnecting(so);
620 socket_unlock(so, 0);
621 error = (*kcb->kctl->connect)(kcb->kctl->kctlref, &kcb->sac, &kcb->userdata);
622 socket_lock(so, 0);
623 if (error) {
624 goto end;
625 }
626 soisconnected(so);
627 #if DEVELOPMENT || DEBUG
628 kcb->status = KCTL_CONNECTED;
629 #endif /* DEVELOPMENT || DEBUG */
630
631 end:
632 if (error && kcb->kctl->disconnect) {
633 /*
634 * XXX Make sure we Don't check the return value
635 * of disconnect here.
636 * ipsec/utun_ctl_disconnect will return error when
637 * disconnect gets called after connect failure.
638 * However if we decide to check for disconnect return
639 * value here. Please make sure to revisit
640 * ipsec/utun_ctl_disconnect.
641 */
642 socket_unlock(so, 0);
643 (*kcb->kctl->disconnect)(kcb->kctl->kctlref, kcb->sac.sc_unit, kcb->userdata);
644 socket_lock(so, 0);
645 }
646 if (error) {
647 soisdisconnected(so);
648 #if DEVELOPMENT || DEBUG
649 kcb->status = KCTL_DISCONNECTED;
650 #endif /* DEVELOPMENT || DEBUG */
651 lck_mtx_lock(&ctl_mtx);
652 TAILQ_REMOVE(&kcb->kctl->kcb_head, kcb, next);
653 kcb->kctl = NULL;
654 kcb->sac.sc_unit = 0;
655 kctlstat.kcs_pcbcount--;
656 kctlstat.kcs_gencnt++;
657 kctlstat.kcs_conn_fail++;
658 lck_mtx_unlock(&ctl_mtx);
659 }
660 out:
661 ctl_kcb_done_clearing(kcb);
662 ctl_kcb_decrement_use_count(kcb);
663 return error;
664 }
665
666 static int
ctl_disconnect(struct socket * so)667 ctl_disconnect(struct socket *so)
668 {
669 struct ctl_cb *kcb = (struct ctl_cb *)so->so_pcb;
670
671 if ((kcb = (struct ctl_cb *)so->so_pcb)) {
672 lck_mtx_t *mtx_held = socket_getlock(so, PR_F_WILLUNLOCK);
673 ctl_kcb_increment_use_count(kcb, mtx_held);
674 ctl_kcb_require_clearing(kcb, mtx_held);
675 struct kctl *kctl = kcb->kctl;
676
677 if (kctl && kctl->disconnect) {
678 socket_unlock(so, 0);
679 (*kctl->disconnect)(kctl->kctlref, kcb->sac.sc_unit,
680 kcb->userdata);
681 socket_lock(so, 0);
682 }
683
684 soisdisconnected(so);
685 #if DEVELOPMENT || DEBUG
686 kcb->status = KCTL_DISCONNECTED;
687 #endif /* DEVELOPMENT || DEBUG */
688
689 socket_unlock(so, 0);
690 lck_mtx_lock(&ctl_mtx);
691 kcb->kctl = 0;
692 kcb->sac.sc_unit = 0;
693 while (kcb->usecount != 0) {
694 msleep(&kcb->usecount, &ctl_mtx, 0, "kcb->usecount", 0);
695 }
696 TAILQ_REMOVE(&kctl->kcb_head, kcb, next);
697 kctlstat.kcs_pcbcount--;
698 kctlstat.kcs_gencnt++;
699 lck_mtx_unlock(&ctl_mtx);
700 socket_lock(so, 0);
701 ctl_kcb_done_clearing(kcb);
702 ctl_kcb_decrement_use_count(kcb);
703 }
704 return 0;
705 }
706
707 static int
ctl_peeraddr(struct socket * so,struct sockaddr ** nam)708 ctl_peeraddr(struct socket *so, struct sockaddr **nam)
709 {
710 struct ctl_cb *kcb = (struct ctl_cb *)so->so_pcb;
711 struct kctl *kctl;
712 struct sockaddr_ctl sc;
713
714 if (kcb == NULL) { /* sanity check */
715 return ENOTCONN;
716 }
717
718 if ((kctl = kcb->kctl) == NULL) {
719 return EINVAL;
720 }
721
722 bzero(&sc, sizeof(struct sockaddr_ctl));
723 sc.sc_len = sizeof(struct sockaddr_ctl);
724 sc.sc_family = AF_SYSTEM;
725 sc.ss_sysaddr = AF_SYS_CONTROL;
726 sc.sc_id = kctl->id;
727 sc.sc_unit = kcb->sac.sc_unit;
728
729 *nam = dup_sockaddr((struct sockaddr *)&sc, 1);
730
731 return 0;
732 }
733
734 static void
ctl_sbrcv_trim(struct socket * so)735 ctl_sbrcv_trim(struct socket *so)
736 {
737 struct sockbuf *sb = &so->so_rcv;
738
739 if (sb->sb_hiwat > sb->sb_idealsize) {
740 u_int32_t diff;
741 int32_t trim;
742
743 /*
744 * The difference between the ideal size and the
745 * current size is the upper bound of the trimage
746 */
747 diff = sb->sb_hiwat - sb->sb_idealsize;
748 /*
749 * We cannot trim below the outstanding data
750 */
751 trim = sb->sb_hiwat - sb->sb_cc;
752
753 trim = imin(trim, (int32_t)diff);
754
755 if (trim > 0) {
756 sbreserve(sb, (sb->sb_hiwat - trim));
757
758 if (ctl_debug) {
759 printf("%s - shrunk to %d\n",
760 __func__, sb->sb_hiwat);
761 }
762 }
763 }
764 }
765
766 static int
ctl_usr_rcvd(struct socket * so,int flags)767 ctl_usr_rcvd(struct socket *so, int flags)
768 {
769 int error = 0;
770 struct ctl_cb *kcb = (struct ctl_cb *)so->so_pcb;
771 struct kctl *kctl;
772
773 if (kcb == NULL) {
774 return ENOTCONN;
775 }
776
777 lck_mtx_t *mtx_held = socket_getlock(so, PR_F_WILLUNLOCK);
778 ctl_kcb_increment_use_count(kcb, mtx_held);
779
780 if ((kctl = kcb->kctl) == NULL) {
781 error = EINVAL;
782 goto out;
783 }
784
785 if (kctl->rcvd) {
786 socket_unlock(so, 0);
787 (*kctl->rcvd)(kctl->kctlref, kcb->sac.sc_unit, kcb->userdata, flags);
788 socket_lock(so, 0);
789 }
790
791 ctl_sbrcv_trim(so);
792
793 out:
794 ctl_kcb_decrement_use_count(kcb);
795 return error;
796 }
797
798 static int
ctl_send(struct socket * so,int flags,struct mbuf * m,struct sockaddr * addr,struct mbuf * control,struct proc * p)799 ctl_send(struct socket *so, int flags, struct mbuf *m,
800 struct sockaddr *addr, struct mbuf *control,
801 struct proc *p)
802 {
803 #pragma unused(addr, p)
804 int error = 0;
805 struct ctl_cb *kcb = (struct ctl_cb *)so->so_pcb;
806 struct kctl *kctl;
807
808 if (control) {
809 m_freem(control);
810 }
811
812 if (kcb == NULL) { /* sanity check */
813 error = ENOTCONN;
814 }
815
816 lck_mtx_t *mtx_held = socket_getlock(so, PR_F_WILLUNLOCK);
817 ctl_kcb_increment_use_count(kcb, mtx_held);
818
819 if (error == 0 && (kctl = kcb->kctl) == NULL) {
820 error = EINVAL;
821 }
822
823 if (error == 0 && kctl->send) {
824 so_tc_update_stats(m, so, m_get_service_class(m));
825 socket_unlock(so, 0);
826 error = (*kctl->send)(kctl->kctlref, kcb->sac.sc_unit, kcb->userdata,
827 m, flags);
828 socket_lock(so, 0);
829 } else {
830 m_freem(m);
831 if (error == 0) {
832 error = ENOTSUP;
833 }
834 }
835 if (error != 0) {
836 OSIncrementAtomic64((SInt64 *)&kctlstat.kcs_send_fail);
837 }
838 ctl_kcb_decrement_use_count(kcb);
839
840 return error;
841 }
842
843 static int
ctl_send_list(struct socket * so,int flags,struct mbuf * m,__unused struct sockaddr * addr,struct mbuf * control,__unused struct proc * p)844 ctl_send_list(struct socket *so, int flags, struct mbuf *m,
845 __unused struct sockaddr *addr, struct mbuf *control,
846 __unused struct proc *p)
847 {
848 int error = 0;
849 struct ctl_cb *kcb = (struct ctl_cb *)so->so_pcb;
850 struct kctl *kctl;
851
852 if (control) {
853 m_freem_list(control);
854 }
855
856 if (kcb == NULL) { /* sanity check */
857 error = ENOTCONN;
858 }
859
860 lck_mtx_t *mtx_held = socket_getlock(so, PR_F_WILLUNLOCK);
861 ctl_kcb_increment_use_count(kcb, mtx_held);
862
863 if (error == 0 && (kctl = kcb->kctl) == NULL) {
864 error = EINVAL;
865 }
866
867 if (error == 0 && kctl->send_list) {
868 struct mbuf *nxt;
869
870 for (nxt = m; nxt != NULL; nxt = nxt->m_nextpkt) {
871 so_tc_update_stats(nxt, so, m_get_service_class(nxt));
872 }
873
874 socket_unlock(so, 0);
875 error = (*kctl->send_list)(kctl->kctlref, kcb->sac.sc_unit,
876 kcb->userdata, m, flags);
877 socket_lock(so, 0);
878 } else if (error == 0 && kctl->send) {
879 while (m != NULL && error == 0) {
880 struct mbuf *nextpkt = m->m_nextpkt;
881
882 m->m_nextpkt = NULL;
883 so_tc_update_stats(m, so, m_get_service_class(m));
884 socket_unlock(so, 0);
885 error = (*kctl->send)(kctl->kctlref, kcb->sac.sc_unit,
886 kcb->userdata, m, flags);
887 socket_lock(so, 0);
888 m = nextpkt;
889 }
890 if (m != NULL) {
891 m_freem_list(m);
892 }
893 } else {
894 m_freem_list(m);
895 if (error == 0) {
896 error = ENOTSUP;
897 }
898 }
899 if (error != 0) {
900 OSIncrementAtomic64((SInt64 *)&kctlstat.kcs_send_list_fail);
901 }
902 ctl_kcb_decrement_use_count(kcb);
903
904 return error;
905 }
906
907 static errno_t
ctl_rcvbspace(struct socket * so,size_t datasize,u_int32_t kctlflags,u_int32_t flags)908 ctl_rcvbspace(struct socket *so, size_t datasize,
909 u_int32_t kctlflags, u_int32_t flags)
910 {
911 struct sockbuf *sb = &so->so_rcv;
912 u_int32_t space = sbspace(sb);
913 errno_t error;
914
915 if ((kctlflags & CTL_FLAG_REG_CRIT) == 0) {
916 if ((u_int32_t) space >= datasize) {
917 error = 0;
918 } else {
919 error = ENOBUFS;
920 }
921 } else if ((flags & CTL_DATA_CRIT) == 0) {
922 /*
923 * Reserve 25% for critical messages
924 */
925 if (space < (sb->sb_hiwat >> 2) ||
926 space < datasize) {
927 error = ENOBUFS;
928 } else {
929 error = 0;
930 }
931 } else {
932 size_t autorcvbuf_max;
933
934 /*
935 * Allow overcommit of 25%
936 */
937 autorcvbuf_max = min(sb->sb_idealsize + (sb->sb_idealsize >> 2),
938 ctl_autorcvbuf_max);
939
940 if ((u_int32_t) space >= datasize) {
941 error = 0;
942 } else if (tcp_cansbgrow(sb) &&
943 sb->sb_hiwat < autorcvbuf_max) {
944 /*
945 * Grow with a little bit of leeway
946 */
947 size_t grow = datasize - space + MSIZE;
948 u_int32_t cc = (u_int32_t)MIN(MIN((sb->sb_hiwat + grow), autorcvbuf_max), UINT32_MAX);
949
950 if (sbreserve(sb, cc) == 1) {
951 if (sb->sb_hiwat > ctl_autorcvbuf_high) {
952 ctl_autorcvbuf_high = sb->sb_hiwat;
953 }
954
955 /*
956 * A final check
957 */
958 if ((u_int32_t) sbspace(sb) >= datasize) {
959 error = 0;
960 } else {
961 error = ENOBUFS;
962 }
963
964 if (ctl_debug) {
965 printf("%s - grown to %d error %d\n",
966 __func__, sb->sb_hiwat, error);
967 }
968 } else {
969 error = ENOBUFS;
970 }
971 } else {
972 error = ENOBUFS;
973 }
974 }
975 return error;
976 }
977
978 errno_t
ctl_enqueuembuf(kern_ctl_ref kctlref,u_int32_t unit,struct mbuf * m,u_int32_t flags)979 ctl_enqueuembuf(kern_ctl_ref kctlref, u_int32_t unit, struct mbuf *m,
980 u_int32_t flags)
981 {
982 struct socket *so;
983 errno_t error = 0;
984 int len = m->m_pkthdr.len;
985 u_int32_t kctlflags;
986
987 so = kcb_find_socket(kctlref, unit, &kctlflags);
988 if (so == NULL) {
989 return EINVAL;
990 }
991
992 if (ctl_rcvbspace(so, len, kctlflags, flags) != 0) {
993 error = ENOBUFS;
994 OSIncrementAtomic64((SInt64 *)&kctlstat.kcs_enqueue_fullsock);
995 goto bye;
996 }
997 if ((flags & CTL_DATA_EOR)) {
998 m->m_flags |= M_EOR;
999 }
1000
1001 so_recv_data_stat(so, m, 0);
1002 if (sbappend_nodrop(&so->so_rcv, m) != 0) {
1003 if ((flags & CTL_DATA_NOWAKEUP) == 0) {
1004 sorwakeup(so);
1005 }
1006 } else {
1007 error = ENOBUFS;
1008 OSIncrementAtomic64((SInt64 *)&kctlstat.kcs_enqueue_fullsock);
1009 }
1010 bye:
1011 if (ctl_debug && error != 0 && (flags & CTL_DATA_CRIT)) {
1012 printf("%s - crit data err %d len %d hiwat %d cc: %d\n",
1013 __func__, error, len,
1014 so->so_rcv.sb_hiwat, so->so_rcv.sb_cc);
1015 }
1016
1017 socket_unlock(so, 1);
1018 if (error != 0) {
1019 OSIncrementAtomic64((SInt64 *)&kctlstat.kcs_enqueue_fail);
1020 }
1021
1022 return error;
1023 }
1024
1025 /*
1026 * Compute space occupied by mbuf like sbappendrecord
1027 */
1028 static int
m_space(struct mbuf * m)1029 m_space(struct mbuf *m)
1030 {
1031 int space = 0;
1032 struct mbuf *nxt;
1033
1034 for (nxt = m; nxt != NULL; nxt = nxt->m_next) {
1035 space += nxt->m_len;
1036 }
1037
1038 return space;
1039 }
1040
1041 errno_t
ctl_enqueuembuf_list(void * kctlref,u_int32_t unit,struct mbuf * m_list,u_int32_t flags,struct mbuf ** m_remain)1042 ctl_enqueuembuf_list(void *kctlref, u_int32_t unit, struct mbuf *m_list,
1043 u_int32_t flags, struct mbuf **m_remain)
1044 {
1045 struct socket *so = NULL;
1046 errno_t error = 0;
1047 struct mbuf *m, *nextpkt;
1048 int needwakeup = 0;
1049 int len = 0;
1050 u_int32_t kctlflags;
1051
1052 /*
1053 * Need to point the beginning of the list in case of early exit
1054 */
1055 m = m_list;
1056
1057 /*
1058 * kcb_find_socket takes the socket lock with a reference
1059 */
1060 so = kcb_find_socket(kctlref, unit, &kctlflags);
1061 if (so == NULL) {
1062 error = EINVAL;
1063 goto done;
1064 }
1065
1066 if (kctlflags & CTL_FLAG_REG_SOCK_STREAM) {
1067 error = EOPNOTSUPP;
1068 goto done;
1069 }
1070 if (flags & CTL_DATA_EOR) {
1071 error = EINVAL;
1072 goto done;
1073 }
1074
1075 for (m = m_list; m != NULL; m = nextpkt) {
1076 nextpkt = m->m_nextpkt;
1077
1078 if (m->m_pkthdr.len == 0 && ctl_debug) {
1079 printf("%s: %llx m_pkthdr.len is 0",
1080 __func__, (uint64_t)VM_KERNEL_ADDRPERM(m));
1081 }
1082
1083 /*
1084 * The mbuf is either appended or freed by sbappendrecord()
1085 * so it's not reliable from a data standpoint
1086 */
1087 len = m_space(m);
1088 if (ctl_rcvbspace(so, len, kctlflags, flags) != 0) {
1089 error = ENOBUFS;
1090 OSIncrementAtomic64(
1091 (SInt64 *)&kctlstat.kcs_enqueue_fullsock);
1092 break;
1093 } else {
1094 /*
1095 * Unlink from the list, m is on its own
1096 */
1097 m->m_nextpkt = NULL;
1098 so_recv_data_stat(so, m, 0);
1099 if (sbappendrecord_nodrop(&so->so_rcv, m) != 0) {
1100 needwakeup = 1;
1101 } else {
1102 /*
1103 * We free or return the remaining
1104 * mbufs in the list
1105 */
1106 m = nextpkt;
1107 error = ENOBUFS;
1108 OSIncrementAtomic64(
1109 (SInt64 *)&kctlstat.kcs_enqueue_fullsock);
1110 break;
1111 }
1112 }
1113 }
1114 if (needwakeup && (flags & CTL_DATA_NOWAKEUP) == 0) {
1115 sorwakeup(so);
1116 }
1117
1118 done:
1119 if (so != NULL) {
1120 if (ctl_debug && error != 0 && (flags & CTL_DATA_CRIT)) {
1121 printf("%s - crit data err %d len %d hiwat %d cc: %d\n",
1122 __func__, error, len,
1123 so->so_rcv.sb_hiwat, so->so_rcv.sb_cc);
1124 }
1125
1126 socket_unlock(so, 1);
1127 }
1128 if (m_remain) {
1129 *m_remain = m;
1130
1131 if (m != NULL && socket_debug && so != NULL &&
1132 (so->so_options & SO_DEBUG)) {
1133 struct mbuf *n;
1134
1135 printf("%s m_list %llx\n", __func__,
1136 (uint64_t) VM_KERNEL_ADDRPERM(m_list));
1137 for (n = m; n != NULL; n = n->m_nextpkt) {
1138 printf(" remain %llx m_next %llx\n",
1139 (uint64_t) VM_KERNEL_ADDRPERM(n),
1140 (uint64_t) VM_KERNEL_ADDRPERM(n->m_next));
1141 }
1142 }
1143 } else {
1144 if (m != NULL) {
1145 m_freem_list(m);
1146 }
1147 }
1148 if (error != 0) {
1149 OSIncrementAtomic64((SInt64 *)&kctlstat.kcs_enqueue_fail);
1150 }
1151 return error;
1152 }
1153
1154 errno_t
ctl_enqueuedata(void * kctlref,u_int32_t unit,void * data,size_t len,u_int32_t flags)1155 ctl_enqueuedata(void *kctlref, u_int32_t unit, void *data, size_t len,
1156 u_int32_t flags)
1157 {
1158 struct socket *so;
1159 struct mbuf *m;
1160 errno_t error = 0;
1161 unsigned int num_needed;
1162 struct mbuf *n;
1163 size_t curlen = 0;
1164 u_int32_t kctlflags;
1165
1166 so = kcb_find_socket(kctlref, unit, &kctlflags);
1167 if (so == NULL) {
1168 return EINVAL;
1169 }
1170
1171 if (ctl_rcvbspace(so, len, kctlflags, flags) != 0) {
1172 error = ENOBUFS;
1173 OSIncrementAtomic64((SInt64 *)&kctlstat.kcs_enqueue_fullsock);
1174 goto bye;
1175 }
1176
1177 num_needed = 1;
1178 m = m_allocpacket_internal(&num_needed, len, NULL, M_NOWAIT, 1, 0);
1179 if (m == NULL) {
1180 kctlstat.kcs_enqdata_mb_alloc_fail++;
1181 if (ctl_debug) {
1182 printf("%s: m_allocpacket_internal(%lu) failed\n",
1183 __func__, len);
1184 }
1185 error = ENOMEM;
1186 goto bye;
1187 }
1188
1189 for (n = m; n != NULL; n = n->m_next) {
1190 size_t mlen = mbuf_maxlen(n);
1191
1192 if (mlen + curlen > len) {
1193 mlen = len - curlen;
1194 }
1195 n->m_len = (int32_t)mlen;
1196 bcopy((char *)data + curlen, n->m_data, mlen);
1197 curlen += mlen;
1198 }
1199 mbuf_pkthdr_setlen(m, curlen);
1200
1201 if ((flags & CTL_DATA_EOR)) {
1202 m->m_flags |= M_EOR;
1203 }
1204 so_recv_data_stat(so, m, 0);
1205 /*
1206 * No need to call the "nodrop" variant of sbappend
1207 * because the mbuf is local to the scope of the function
1208 */
1209 if (sbappend(&so->so_rcv, m) != 0) {
1210 if ((flags & CTL_DATA_NOWAKEUP) == 0) {
1211 sorwakeup(so);
1212 }
1213 } else {
1214 kctlstat.kcs_enqdata_sbappend_fail++;
1215 error = ENOBUFS;
1216 OSIncrementAtomic64((SInt64 *)&kctlstat.kcs_enqueue_fullsock);
1217 }
1218
1219 bye:
1220 if (ctl_debug && error != 0 && (flags & CTL_DATA_CRIT)) {
1221 printf("%s - crit data err %d len %d hiwat %d cc: %d\n",
1222 __func__, error, (int)len,
1223 so->so_rcv.sb_hiwat, so->so_rcv.sb_cc);
1224 }
1225
1226 socket_unlock(so, 1);
1227 if (error != 0) {
1228 OSIncrementAtomic64((SInt64 *)&kctlstat.kcs_enqueue_fail);
1229 }
1230 return error;
1231 }
1232
1233 errno_t
ctl_getenqueuepacketcount(kern_ctl_ref kctlref,u_int32_t unit,u_int32_t * pcnt)1234 ctl_getenqueuepacketcount(kern_ctl_ref kctlref, u_int32_t unit, u_int32_t *pcnt)
1235 {
1236 struct socket *so;
1237 u_int32_t cnt;
1238 struct mbuf *m1;
1239
1240 if (pcnt == NULL) {
1241 return EINVAL;
1242 }
1243
1244 so = kcb_find_socket(kctlref, unit, NULL);
1245 if (so == NULL) {
1246 return EINVAL;
1247 }
1248
1249 cnt = 0;
1250 m1 = so->so_rcv.sb_mb;
1251 while (m1 != NULL) {
1252 if (m1->m_type == MT_DATA ||
1253 m1->m_type == MT_HEADER ||
1254 m1->m_type == MT_OOBDATA) {
1255 cnt += 1;
1256 }
1257 m1 = m1->m_nextpkt;
1258 }
1259 *pcnt = cnt;
1260
1261 socket_unlock(so, 1);
1262
1263 return 0;
1264 }
1265
1266 errno_t
ctl_getenqueuespace(kern_ctl_ref kctlref,u_int32_t unit,size_t * space)1267 ctl_getenqueuespace(kern_ctl_ref kctlref, u_int32_t unit, size_t *space)
1268 {
1269 struct socket *so;
1270 long avail;
1271
1272 if (space == NULL) {
1273 return EINVAL;
1274 }
1275
1276 so = kcb_find_socket(kctlref, unit, NULL);
1277 if (so == NULL) {
1278 return EINVAL;
1279 }
1280
1281 avail = sbspace(&so->so_rcv);
1282 *space = (avail < 0) ? 0 : avail;
1283 socket_unlock(so, 1);
1284
1285 return 0;
1286 }
1287
1288 errno_t
ctl_getenqueuereadable(kern_ctl_ref kctlref,u_int32_t unit,u_int32_t * difference)1289 ctl_getenqueuereadable(kern_ctl_ref kctlref, u_int32_t unit,
1290 u_int32_t *difference)
1291 {
1292 struct socket *so;
1293
1294 if (difference == NULL) {
1295 return EINVAL;
1296 }
1297
1298 so = kcb_find_socket(kctlref, unit, NULL);
1299 if (so == NULL) {
1300 return EINVAL;
1301 }
1302
1303 if (so->so_rcv.sb_cc >= so->so_rcv.sb_lowat) {
1304 *difference = 0;
1305 } else {
1306 *difference = (so->so_rcv.sb_lowat - so->so_rcv.sb_cc);
1307 }
1308 socket_unlock(so, 1);
1309
1310 return 0;
1311 }
1312
1313 static int
ctl_ctloutput(struct socket * so,struct sockopt * sopt)1314 ctl_ctloutput(struct socket *so, struct sockopt *sopt)
1315 {
1316 struct ctl_cb *kcb = (struct ctl_cb *)so->so_pcb;
1317 struct kctl *kctl;
1318 int error = 0;
1319 void *data = NULL;
1320 size_t data_len = 0;
1321 size_t len;
1322
1323 if (sopt->sopt_level != SYSPROTO_CONTROL) {
1324 return EINVAL;
1325 }
1326
1327 if (kcb == NULL) { /* sanity check */
1328 return ENOTCONN;
1329 }
1330
1331 if ((kctl = kcb->kctl) == NULL) {
1332 return EINVAL;
1333 }
1334
1335 lck_mtx_t *mtx_held = socket_getlock(so, PR_F_WILLUNLOCK);
1336 ctl_kcb_increment_use_count(kcb, mtx_held);
1337
1338 switch (sopt->sopt_dir) {
1339 case SOPT_SET:
1340 if (kctl->setopt == NULL) {
1341 error = ENOTSUP;
1342 goto out;
1343 }
1344 if (sopt->sopt_valsize != 0) {
1345 data_len = sopt->sopt_valsize;
1346 data = kalloc_data(data_len, Z_WAITOK | Z_ZERO);
1347 if (data == NULL) {
1348 data_len = 0;
1349 error = ENOMEM;
1350 goto out;
1351 }
1352 error = sooptcopyin(sopt, data,
1353 sopt->sopt_valsize, sopt->sopt_valsize);
1354 }
1355 if (error == 0) {
1356 socket_unlock(so, 0);
1357 error = (*kctl->setopt)(kctl->kctlref,
1358 kcb->sac.sc_unit, kcb->userdata, sopt->sopt_name,
1359 data, sopt->sopt_valsize);
1360 socket_lock(so, 0);
1361 }
1362
1363 kfree_data(data, data_len);
1364 break;
1365
1366 case SOPT_GET:
1367 if (kctl->getopt == NULL) {
1368 error = ENOTSUP;
1369 goto out;
1370 }
1371
1372 if (sopt->sopt_valsize && sopt->sopt_val) {
1373 data_len = sopt->sopt_valsize;
1374 data = kalloc_data(data_len, Z_WAITOK | Z_ZERO);
1375 if (data == NULL) {
1376 data_len = 0;
1377 error = ENOMEM;
1378 goto out;
1379 }
1380 /*
1381 * 4108337 - copy user data in case the
1382 * kernel control needs it
1383 */
1384 error = sooptcopyin(sopt, data,
1385 sopt->sopt_valsize, sopt->sopt_valsize);
1386 }
1387
1388 if (error == 0) {
1389 len = sopt->sopt_valsize;
1390 socket_unlock(so, 0);
1391 error = (*kctl->getopt)(kctl->kctlref, kcb->sac.sc_unit,
1392 kcb->userdata, sopt->sopt_name,
1393 data, &len);
1394 if (data != NULL && len > sopt->sopt_valsize) {
1395 panic_plain("ctl_ctloutput: ctl %s returned "
1396 "len (%lu) > sopt_valsize (%lu)\n",
1397 kcb->kctl->name, len,
1398 sopt->sopt_valsize);
1399 }
1400 socket_lock(so, 0);
1401 if (error == 0) {
1402 if (data != NULL) {
1403 error = sooptcopyout(sopt, data, len);
1404 } else {
1405 sopt->sopt_valsize = len;
1406 }
1407 }
1408 }
1409
1410 kfree_data(data, data_len);
1411 break;
1412 }
1413
1414 out:
1415 ctl_kcb_decrement_use_count(kcb);
1416 return error;
1417 }
1418
1419 static int
ctl_ioctl(struct socket * so,u_long cmd,caddr_t data,struct ifnet * ifp,struct proc * p)1420 ctl_ioctl(struct socket *so, u_long cmd, caddr_t data,
1421 struct ifnet *ifp, struct proc *p)
1422 {
1423 #pragma unused(so, ifp, p)
1424 int error = ENOTSUP;
1425
1426 switch (cmd) {
1427 /* get the number of controllers */
1428 case CTLIOCGCOUNT: {
1429 struct kctl *kctl;
1430 u_int32_t n = 0;
1431
1432 lck_mtx_lock(&ctl_mtx);
1433 TAILQ_FOREACH(kctl, &ctl_head, next)
1434 n++;
1435 lck_mtx_unlock(&ctl_mtx);
1436
1437 bcopy(&n, data, sizeof(n));
1438 error = 0;
1439 break;
1440 }
1441 case CTLIOCGINFO: {
1442 struct ctl_info ctl_info;
1443 struct kctl *kctl = 0;
1444 size_t name_len;
1445
1446 bcopy(data, &ctl_info, sizeof(ctl_info));
1447 name_len = strnlen(ctl_info.ctl_name, MAX_KCTL_NAME);
1448
1449 if (name_len == 0 || name_len + 1 > MAX_KCTL_NAME) {
1450 error = EINVAL;
1451 break;
1452 }
1453 lck_mtx_lock(&ctl_mtx);
1454 kctl = ctl_find_by_name(ctl_info.ctl_name);
1455 lck_mtx_unlock(&ctl_mtx);
1456 if (kctl == 0) {
1457 error = ENOENT;
1458 break;
1459 }
1460 ctl_info.ctl_id = kctl->id;
1461 bcopy(&ctl_info, data, sizeof(ctl_info));
1462 error = 0;
1463 break;
1464 }
1465
1466 /* add controls to get list of NKEs */
1467 }
1468
1469 return error;
1470 }
1471
1472 static void
kctl_tbl_grow(void)1473 kctl_tbl_grow(void)
1474 {
1475 struct kctl **new_table;
1476 uintptr_t new_size;
1477
1478 lck_mtx_assert(&ctl_mtx, LCK_MTX_ASSERT_OWNED);
1479
1480 if (kctl_tbl_growing) {
1481 /* Another thread is allocating */
1482 kctl_tbl_growing_waiting++;
1483
1484 do {
1485 (void) msleep((caddr_t) &kctl_tbl_growing, &ctl_mtx,
1486 PSOCK | PCATCH, "kctl_tbl_growing", 0);
1487 } while (kctl_tbl_growing);
1488 kctl_tbl_growing_waiting--;
1489 }
1490 /* Another thread grew the table */
1491 if (kctl_table != NULL && kctl_tbl_count < kctl_tbl_size) {
1492 return;
1493 }
1494
1495 /* Verify we have a sane size */
1496 if (kctl_tbl_size + KCTL_TBL_INC >= UINT16_MAX) {
1497 kctlstat.kcs_tbl_size_too_big++;
1498 if (ctl_debug) {
1499 printf("%s kctl_tbl_size %lu too big\n",
1500 __func__, kctl_tbl_size);
1501 }
1502 return;
1503 }
1504 kctl_tbl_growing = 1;
1505
1506 new_size = kctl_tbl_size + KCTL_TBL_INC;
1507
1508 lck_mtx_unlock(&ctl_mtx);
1509 new_table = kalloc_type(struct kctl *, new_size, Z_WAITOK | Z_ZERO);
1510 lck_mtx_lock(&ctl_mtx);
1511
1512 if (new_table != NULL) {
1513 if (kctl_table != NULL) {
1514 bcopy(kctl_table, new_table,
1515 kctl_tbl_size * sizeof(struct kctl *));
1516
1517 kfree_type(struct kctl *, kctl_tbl_size, kctl_table);
1518 }
1519 kctl_table = new_table;
1520 kctl_tbl_size = new_size;
1521 }
1522
1523 kctl_tbl_growing = 0;
1524
1525 if (kctl_tbl_growing_waiting) {
1526 wakeup(&kctl_tbl_growing);
1527 }
1528 }
1529
1530 #define KCTLREF_INDEX_MASK 0x0000FFFF
1531 #define KCTLREF_GENCNT_MASK 0xFFFF0000
1532 #define KCTLREF_GENCNT_SHIFT 16
1533
1534 static kern_ctl_ref
kctl_make_ref(struct kctl * kctl)1535 kctl_make_ref(struct kctl *kctl)
1536 {
1537 uintptr_t i;
1538
1539 lck_mtx_assert(&ctl_mtx, LCK_MTX_ASSERT_OWNED);
1540
1541 if (kctl_tbl_count >= kctl_tbl_size) {
1542 kctl_tbl_grow();
1543 }
1544
1545 kctl->kctlref = NULL;
1546 for (i = 0; i < kctl_tbl_size; i++) {
1547 if (kctl_table[i] == NULL) {
1548 uintptr_t ref;
1549
1550 /*
1551 * Reference is index plus one
1552 */
1553 kctl_ref_gencnt += 1;
1554
1555 /*
1556 * Add generation count as salt to reference to prevent
1557 * use after deregister
1558 */
1559 ref = ((kctl_ref_gencnt << KCTLREF_GENCNT_SHIFT) &
1560 KCTLREF_GENCNT_MASK) +
1561 ((i + 1) & KCTLREF_INDEX_MASK);
1562
1563 kctl->kctlref = (void *)(ref);
1564 kctl_table[i] = kctl;
1565 kctl_tbl_count++;
1566 break;
1567 }
1568 }
1569
1570 if (kctl->kctlref == NULL) {
1571 panic("%s no space in table", __func__);
1572 }
1573
1574 if (ctl_debug > 0) {
1575 printf("%s %p for %p\n",
1576 __func__, kctl->kctlref, kctl);
1577 }
1578
1579 return kctl->kctlref;
1580 }
1581
1582 static void
kctl_delete_ref(kern_ctl_ref kctlref)1583 kctl_delete_ref(kern_ctl_ref kctlref)
1584 {
1585 /*
1586 * Reference is index plus one
1587 */
1588 uintptr_t i = (((uintptr_t)kctlref) & KCTLREF_INDEX_MASK) - 1;
1589
1590 lck_mtx_assert(&ctl_mtx, LCK_MTX_ASSERT_OWNED);
1591
1592 if (i < kctl_tbl_size) {
1593 struct kctl *kctl = kctl_table[i];
1594
1595 if (kctl->kctlref == kctlref) {
1596 kctl_table[i] = NULL;
1597 kctl_tbl_count--;
1598 } else {
1599 kctlstat.kcs_bad_kctlref++;
1600 }
1601 } else {
1602 kctlstat.kcs_bad_kctlref++;
1603 }
1604 }
1605
1606 static struct kctl *
kctl_from_ref(kern_ctl_ref kctlref)1607 kctl_from_ref(kern_ctl_ref kctlref)
1608 {
1609 /*
1610 * Reference is index plus one
1611 */
1612 uintptr_t i = (((uintptr_t)kctlref) & KCTLREF_INDEX_MASK) - 1;
1613 struct kctl *kctl = NULL;
1614
1615 lck_mtx_assert(&ctl_mtx, LCK_MTX_ASSERT_OWNED);
1616
1617 if (i >= kctl_tbl_size) {
1618 kctlstat.kcs_bad_kctlref++;
1619 return NULL;
1620 }
1621 kctl = kctl_table[i];
1622 if (kctl->kctlref != kctlref) {
1623 kctlstat.kcs_bad_kctlref++;
1624 return NULL;
1625 }
1626 return kctl;
1627 }
1628
1629 /*
1630 * Register/unregister a NKE
1631 */
1632 errno_t
ctl_register(struct kern_ctl_reg * userkctl,kern_ctl_ref * kctlref)1633 ctl_register(struct kern_ctl_reg *userkctl, kern_ctl_ref *kctlref)
1634 {
1635 struct kctl *kctl = NULL;
1636 struct kctl *kctl_next = NULL;
1637 u_int32_t id = 1;
1638 size_t name_len;
1639 int is_extended = 0;
1640 int is_setup = 0;
1641
1642 if (userkctl == NULL) { /* sanity check */
1643 return EINVAL;
1644 }
1645 if (userkctl->ctl_connect == NULL) {
1646 return EINVAL;
1647 }
1648 name_len = strlen(userkctl->ctl_name);
1649 if (name_len == 0 || name_len + 1 > MAX_KCTL_NAME) {
1650 return EINVAL;
1651 }
1652
1653 kctl = kalloc_type(struct kctl, Z_WAITOK | Z_ZERO | Z_NOFAIL);
1654
1655 lck_mtx_lock(&ctl_mtx);
1656
1657 if (kctl_make_ref(kctl) == NULL) {
1658 lck_mtx_unlock(&ctl_mtx);
1659 kfree_type(struct kctl, kctl);
1660 return ENOMEM;
1661 }
1662
1663 /*
1664 * Kernel Control IDs
1665 *
1666 * CTL_FLAG_REG_ID_UNIT indicates the control ID and unit number are
1667 * static. If they do not exist, add them to the list in order. If the
1668 * flag is not set, we must find a new unique value. We assume the
1669 * list is in order. We find the last item in the list and add one. If
1670 * this leads to wrapping the id around, we start at the front of the
1671 * list and look for a gap.
1672 */
1673
1674 if ((userkctl->ctl_flags & CTL_FLAG_REG_ID_UNIT) == 0) {
1675 /* Must dynamically assign an unused ID */
1676
1677 /* Verify the same name isn't already registered */
1678 if (ctl_find_by_name(userkctl->ctl_name) != NULL) {
1679 kctl_delete_ref(kctl->kctlref);
1680 lck_mtx_unlock(&ctl_mtx);
1681 kfree_type(struct kctl, kctl);
1682 return EEXIST;
1683 }
1684
1685 /* Start with 1 in case the list is empty */
1686 id = 1;
1687 kctl_next = TAILQ_LAST(&ctl_head, kctl_list);
1688
1689 if (kctl_next != NULL) {
1690 /* List was not empty, add one to the last item */
1691 id = kctl_next->id + 1;
1692 kctl_next = NULL;
1693
1694 /*
1695 * If this wrapped the id number, start looking at
1696 * the front of the list for an unused id.
1697 */
1698 if (id == 0) {
1699 /* Find the next unused ID */
1700 id = 1;
1701
1702 TAILQ_FOREACH(kctl_next, &ctl_head, next) {
1703 if (kctl_next->id > id) {
1704 /* We found a gap */
1705 break;
1706 }
1707
1708 id = kctl_next->id + 1;
1709 }
1710 }
1711 }
1712
1713 userkctl->ctl_id = id;
1714 kctl->id = id;
1715 kctl->reg_unit = -1;
1716 } else {
1717 TAILQ_FOREACH(kctl_next, &ctl_head, next) {
1718 if (kctl_next->id > userkctl->ctl_id) {
1719 break;
1720 }
1721 }
1722
1723 if (ctl_find_by_id_unit(userkctl->ctl_id, userkctl->ctl_unit)) {
1724 kctl_delete_ref(kctl->kctlref);
1725 lck_mtx_unlock(&ctl_mtx);
1726 kfree_type(struct kctl, kctl);
1727 return EEXIST;
1728 }
1729 kctl->id = userkctl->ctl_id;
1730 kctl->reg_unit = userkctl->ctl_unit;
1731 }
1732
1733 is_extended = (userkctl->ctl_flags & CTL_FLAG_REG_EXTENDED);
1734 is_setup = (userkctl->ctl_flags & CTL_FLAG_REG_SETUP);
1735
1736 strlcpy(kctl->name, userkctl->ctl_name, MAX_KCTL_NAME);
1737 kctl->flags = userkctl->ctl_flags;
1738
1739 /*
1740 * Let the caller know the default send and receive sizes
1741 */
1742 if (userkctl->ctl_sendsize == 0) {
1743 kctl->sendbufsize = CTL_SENDSIZE;
1744 userkctl->ctl_sendsize = kctl->sendbufsize;
1745 } else {
1746 kctl->sendbufsize = userkctl->ctl_sendsize;
1747 }
1748 if (userkctl->ctl_recvsize == 0) {
1749 kctl->recvbufsize = CTL_RECVSIZE;
1750 userkctl->ctl_recvsize = kctl->recvbufsize;
1751 } else {
1752 kctl->recvbufsize = userkctl->ctl_recvsize;
1753 }
1754
1755 if (is_setup) {
1756 kctl->setup = userkctl->ctl_setup;
1757 }
1758 kctl->bind = userkctl->ctl_bind;
1759 kctl->connect = userkctl->ctl_connect;
1760 kctl->disconnect = userkctl->ctl_disconnect;
1761 kctl->send = userkctl->ctl_send;
1762 kctl->setopt = userkctl->ctl_setopt;
1763 kctl->getopt = userkctl->ctl_getopt;
1764 if (is_extended) {
1765 kctl->rcvd = userkctl->ctl_rcvd;
1766 kctl->send_list = userkctl->ctl_send_list;
1767 }
1768
1769 TAILQ_INIT(&kctl->kcb_head);
1770
1771 if (kctl_next) {
1772 TAILQ_INSERT_BEFORE(kctl_next, kctl, next);
1773 } else {
1774 TAILQ_INSERT_TAIL(&ctl_head, kctl, next);
1775 }
1776
1777 kctlstat.kcs_reg_count++;
1778 kctlstat.kcs_gencnt++;
1779
1780 lck_mtx_unlock(&ctl_mtx);
1781
1782 *kctlref = kctl->kctlref;
1783
1784 ctl_post_msg(KEV_CTL_REGISTERED, kctl->id);
1785 return 0;
1786 }
1787
1788 errno_t
ctl_deregister(void * kctlref)1789 ctl_deregister(void *kctlref)
1790 {
1791 struct kctl *kctl;
1792
1793 lck_mtx_lock(&ctl_mtx);
1794 if ((kctl = kctl_from_ref(kctlref)) == NULL) {
1795 kctlstat.kcs_bad_kctlref++;
1796 lck_mtx_unlock(&ctl_mtx);
1797 if (ctl_debug != 0) {
1798 printf("%s invalid kctlref %p\n",
1799 __func__, kctlref);
1800 }
1801 return EINVAL;
1802 }
1803
1804 if (!TAILQ_EMPTY(&kctl->kcb_head)) {
1805 lck_mtx_unlock(&ctl_mtx);
1806 return EBUSY;
1807 }
1808
1809 TAILQ_REMOVE(&ctl_head, kctl, next);
1810
1811 kctlstat.kcs_reg_count--;
1812 kctlstat.kcs_gencnt++;
1813
1814 kctl_delete_ref(kctl->kctlref);
1815 lck_mtx_unlock(&ctl_mtx);
1816
1817 ctl_post_msg(KEV_CTL_DEREGISTERED, kctl->id);
1818 kfree_type(struct kctl, kctl);
1819 return 0;
1820 }
1821
1822 /*
1823 * Must be called with global ctl_mtx lock taked
1824 */
1825 static struct kctl *
ctl_find_by_name(const char * name)1826 ctl_find_by_name(const char *name)
1827 {
1828 struct kctl *kctl;
1829
1830 lck_mtx_assert(&ctl_mtx, LCK_MTX_ASSERT_OWNED);
1831
1832 TAILQ_FOREACH(kctl, &ctl_head, next)
1833 if (strncmp(kctl->name, name, sizeof(kctl->name)) == 0) {
1834 return kctl;
1835 }
1836
1837 return NULL;
1838 }
1839
1840 u_int32_t
ctl_id_by_name(const char * name)1841 ctl_id_by_name(const char *name)
1842 {
1843 u_int32_t ctl_id = 0;
1844 struct kctl *kctl;
1845
1846 lck_mtx_lock(&ctl_mtx);
1847 kctl = ctl_find_by_name(name);
1848 if (kctl) {
1849 ctl_id = kctl->id;
1850 }
1851 lck_mtx_unlock(&ctl_mtx);
1852
1853 return ctl_id;
1854 }
1855
1856 errno_t
ctl_name_by_id(u_int32_t id,char * out_name,size_t maxsize)1857 ctl_name_by_id(u_int32_t id, char *out_name, size_t maxsize)
1858 {
1859 int found = 0;
1860 struct kctl *kctl;
1861
1862 lck_mtx_lock(&ctl_mtx);
1863 TAILQ_FOREACH(kctl, &ctl_head, next) {
1864 if (kctl->id == id) {
1865 break;
1866 }
1867 }
1868
1869 if (kctl) {
1870 if (maxsize > MAX_KCTL_NAME) {
1871 maxsize = MAX_KCTL_NAME;
1872 }
1873 strlcpy(out_name, kctl->name, maxsize);
1874 found = 1;
1875 }
1876 lck_mtx_unlock(&ctl_mtx);
1877
1878 return found ? 0 : ENOENT;
1879 }
1880
1881 /*
1882 * Must be called with global ctl_mtx lock taked
1883 *
1884 */
1885 static struct kctl *
ctl_find_by_id_unit(u_int32_t id,u_int32_t unit)1886 ctl_find_by_id_unit(u_int32_t id, u_int32_t unit)
1887 {
1888 struct kctl *kctl;
1889
1890 lck_mtx_assert(&ctl_mtx, LCK_MTX_ASSERT_OWNED);
1891
1892 TAILQ_FOREACH(kctl, &ctl_head, next) {
1893 if (kctl->id == id && (kctl->flags & CTL_FLAG_REG_ID_UNIT) == 0) {
1894 return kctl;
1895 } else if (kctl->id == id && kctl->reg_unit == unit) {
1896 return kctl;
1897 }
1898 }
1899 return NULL;
1900 }
1901
1902 /*
1903 * Must be called with kernel controller lock taken
1904 */
1905 static struct ctl_cb *
kcb_find(struct kctl * kctl,u_int32_t unit)1906 kcb_find(struct kctl *kctl, u_int32_t unit)
1907 {
1908 struct ctl_cb *kcb;
1909
1910 lck_mtx_assert(&ctl_mtx, LCK_MTX_ASSERT_OWNED);
1911
1912 TAILQ_FOREACH(kcb, &kctl->kcb_head, next)
1913 if (kcb->sac.sc_unit == unit) {
1914 return kcb;
1915 }
1916
1917 return NULL;
1918 }
1919
1920 static struct socket *
kcb_find_socket(kern_ctl_ref kctlref,u_int32_t unit,u_int32_t * kctlflags)1921 kcb_find_socket(kern_ctl_ref kctlref, u_int32_t unit, u_int32_t *kctlflags)
1922 {
1923 struct socket *so = NULL;
1924 struct ctl_cb *kcb;
1925 void *lr_saved;
1926 struct kctl *kctl;
1927 int i;
1928
1929 lr_saved = __builtin_return_address(0);
1930
1931 lck_mtx_lock(&ctl_mtx);
1932 /*
1933 * First validate the kctlref
1934 */
1935 if ((kctl = kctl_from_ref(kctlref)) == NULL) {
1936 kctlstat.kcs_bad_kctlref++;
1937 lck_mtx_unlock(&ctl_mtx);
1938 if (ctl_debug != 0) {
1939 printf("%s invalid kctlref %p\n",
1940 __func__, kctlref);
1941 }
1942 return NULL;
1943 }
1944
1945 kcb = kcb_find(kctl, unit);
1946 if (kcb == NULL || kcb->kctl != kctl || (so = kcb->so) == NULL) {
1947 lck_mtx_unlock(&ctl_mtx);
1948 return NULL;
1949 }
1950 /*
1951 * This prevents the socket from being closed
1952 */
1953 kcb->usecount++;
1954 /*
1955 * Respect lock ordering: socket before ctl_mtx
1956 */
1957 lck_mtx_unlock(&ctl_mtx);
1958
1959 socket_lock(so, 1);
1960 /*
1961 * The socket lock history is more useful if we store
1962 * the address of the caller.
1963 */
1964 i = (so->next_lock_lr + SO_LCKDBG_MAX - 1) % SO_LCKDBG_MAX;
1965 so->lock_lr[i] = lr_saved;
1966
1967 lck_mtx_lock(&ctl_mtx);
1968
1969 if ((kctl = kctl_from_ref(kctlref)) == NULL || kcb->kctl == NULL) {
1970 lck_mtx_unlock(&ctl_mtx);
1971 socket_unlock(so, 1);
1972 so = NULL;
1973 lck_mtx_lock(&ctl_mtx);
1974 } else if (kctlflags != NULL) {
1975 *kctlflags = kctl->flags;
1976 }
1977
1978 kcb->usecount--;
1979 if (kcb->usecount == 0) {
1980 wakeup((event_t)&kcb->usecount);
1981 }
1982
1983 lck_mtx_unlock(&ctl_mtx);
1984
1985 return so;
1986 }
1987
1988 static void
ctl_post_msg(u_int32_t event_code,u_int32_t id)1989 ctl_post_msg(u_int32_t event_code, u_int32_t id)
1990 {
1991 struct ctl_event_data ctl_ev_data;
1992 struct kev_msg ev_msg;
1993
1994 lck_mtx_assert(&ctl_mtx, LCK_MTX_ASSERT_NOTOWNED);
1995
1996 bzero(&ev_msg, sizeof(struct kev_msg));
1997 ev_msg.vendor_code = KEV_VENDOR_APPLE;
1998
1999 ev_msg.kev_class = KEV_SYSTEM_CLASS;
2000 ev_msg.kev_subclass = KEV_CTL_SUBCLASS;
2001 ev_msg.event_code = event_code;
2002
2003 /* common nke subclass data */
2004 bzero(&ctl_ev_data, sizeof(ctl_ev_data));
2005 ctl_ev_data.ctl_id = id;
2006 ev_msg.dv[0].data_ptr = &ctl_ev_data;
2007 ev_msg.dv[0].data_length = sizeof(ctl_ev_data);
2008
2009 ev_msg.dv[1].data_length = 0;
2010
2011 kev_post_msg(&ev_msg);
2012 }
2013
2014 static int
ctl_lock(struct socket * so,int refcount,void * lr)2015 ctl_lock(struct socket *so, int refcount, void *lr)
2016 {
2017 void *lr_saved;
2018
2019 if (lr == NULL) {
2020 lr_saved = __builtin_return_address(0);
2021 } else {
2022 lr_saved = lr;
2023 }
2024
2025 if (so->so_pcb != NULL) {
2026 lck_mtx_lock(&((struct ctl_cb *)so->so_pcb)->mtx);
2027 } else {
2028 panic("ctl_lock: so=%p NO PCB! lr=%p lrh= %s",
2029 so, lr_saved, solockhistory_nr(so));
2030 /* NOTREACHED */
2031 }
2032
2033 if (so->so_usecount < 0) {
2034 panic("ctl_lock: so=%p so_pcb=%p lr=%p ref=%x lrh= %s",
2035 so, so->so_pcb, lr_saved, so->so_usecount,
2036 solockhistory_nr(so));
2037 /* NOTREACHED */
2038 }
2039
2040 if (refcount) {
2041 so->so_usecount++;
2042 }
2043
2044 so->lock_lr[so->next_lock_lr] = lr_saved;
2045 so->next_lock_lr = (so->next_lock_lr + 1) % SO_LCKDBG_MAX;
2046 return 0;
2047 }
2048
2049 static int
ctl_unlock(struct socket * so,int refcount,void * lr)2050 ctl_unlock(struct socket *so, int refcount, void *lr)
2051 {
2052 void *lr_saved;
2053 lck_mtx_t *mutex_held;
2054
2055 if (lr == NULL) {
2056 lr_saved = __builtin_return_address(0);
2057 } else {
2058 lr_saved = lr;
2059 }
2060
2061 #if (MORE_KCTLLOCK_DEBUG && (DEVELOPMENT || DEBUG))
2062 printf("ctl_unlock: so=%llx sopcb=%x lock=%llx ref=%u lr=%llx\n",
2063 (uint64_t)VM_KERNEL_ADDRPERM(so),
2064 (uint64_t)VM_KERNEL_ADDRPERM(so->so_pcb,
2065 (uint64_t)VM_KERNEL_ADDRPERM(&((struct ctl_cb *)so->so_pcb)->mtx),
2066 so->so_usecount, (uint64_t)VM_KERNEL_ADDRPERM(lr_saved));
2067 #endif /* (MORE_KCTLLOCK_DEBUG && (DEVELOPMENT || DEBUG)) */
2068 if (refcount) {
2069 so->so_usecount--;
2070 }
2071
2072 if (so->so_usecount < 0) {
2073 panic("ctl_unlock: so=%p usecount=%x lrh= %s",
2074 so, so->so_usecount, solockhistory_nr(so));
2075 /* NOTREACHED */
2076 }
2077 if (so->so_pcb == NULL) {
2078 panic("ctl_unlock: so=%p NO PCB usecount=%x lr=%p lrh= %s",
2079 so, so->so_usecount, (void *)lr_saved,
2080 solockhistory_nr(so));
2081 /* NOTREACHED */
2082 }
2083 mutex_held = &((struct ctl_cb *)so->so_pcb)->mtx;
2084
2085 lck_mtx_assert(mutex_held, LCK_MTX_ASSERT_OWNED);
2086 so->unlock_lr[so->next_unlock_lr] = lr_saved;
2087 so->next_unlock_lr = (so->next_unlock_lr + 1) % SO_LCKDBG_MAX;
2088 lck_mtx_unlock(mutex_held);
2089
2090 if (so->so_usecount == 0) {
2091 ctl_sofreelastref(so);
2092 }
2093
2094 return 0;
2095 }
2096
2097 static lck_mtx_t *
2098 ctl_getlock(struct socket *so, int flags)
2099 {
2100 #pragma unused(flags)
2101 struct ctl_cb *kcb = (struct ctl_cb *)so->so_pcb;
2102
2103 if (so->so_pcb) {
2104 if (so->so_usecount < 0) {
2105 panic("ctl_getlock: so=%p usecount=%x lrh= %s",
2106 so, so->so_usecount, solockhistory_nr(so));
2107 }
2108 return &kcb->mtx;
2109 } else {
2110 panic("ctl_getlock: so=%p NULL NO so_pcb %s",
2111 so, solockhistory_nr(so));
2112 return so->so_proto->pr_domain->dom_mtx;
2113 }
2114 }
2115
2116 __private_extern__ int
2117 kctl_reg_list SYSCTL_HANDLER_ARGS
2118 {
2119 #pragma unused(oidp, arg1, arg2)
2120 int error = 0;
2121 u_int64_t i, n;
2122 struct xsystmgen xsg;
2123 void *buf = NULL;
2124 struct kctl *kctl;
2125 size_t item_size = ROUNDUP64(sizeof(struct xkctl_reg));
2126
2127 buf = kalloc_data(item_size, Z_WAITOK | Z_ZERO | Z_NOFAIL);
2128
2129 lck_mtx_lock(&ctl_mtx);
2130
2131 n = kctlstat.kcs_reg_count;
2132
2133 if (req->oldptr == USER_ADDR_NULL) {
2134 req->oldidx = (size_t)(n + n / 8) * sizeof(struct xkctl_reg);
2135 goto done;
2136 }
2137 if (req->newptr != USER_ADDR_NULL) {
2138 error = EPERM;
2139 goto done;
2140 }
2141 bzero(&xsg, sizeof(xsg));
2142 xsg.xg_len = sizeof(xsg);
2143 xsg.xg_count = n;
2144 xsg.xg_gen = kctlstat.kcs_gencnt;
2145 xsg.xg_sogen = so_gencnt;
2146 error = SYSCTL_OUT(req, &xsg, sizeof(xsg));
2147 if (error) {
2148 goto done;
2149 }
2150 /*
2151 * We are done if there is no pcb
2152 */
2153 if (n == 0) {
2154 goto done;
2155 }
2156
2157 for (i = 0, kctl = TAILQ_FIRST(&ctl_head);
2158 i < n && kctl != NULL;
2159 i++, kctl = TAILQ_NEXT(kctl, next)) {
2160 struct xkctl_reg *xkr = (struct xkctl_reg *)buf;
2161 struct ctl_cb *kcb;
2162 u_int32_t pcbcount = 0;
2163
2164 TAILQ_FOREACH(kcb, &kctl->kcb_head, next)
2165 pcbcount++;
2166
2167 bzero(buf, item_size);
2168
2169 xkr->xkr_len = sizeof(struct xkctl_reg);
2170 xkr->xkr_kind = XSO_KCREG;
2171 xkr->xkr_id = kctl->id;
2172 xkr->xkr_reg_unit = kctl->reg_unit;
2173 xkr->xkr_flags = kctl->flags;
2174 xkr->xkr_kctlref = (uint64_t)(kctl->kctlref);
2175 xkr->xkr_recvbufsize = kctl->recvbufsize;
2176 xkr->xkr_sendbufsize = kctl->sendbufsize;
2177 xkr->xkr_lastunit = kctl->lastunit;
2178 xkr->xkr_pcbcount = pcbcount;
2179 xkr->xkr_connect = (uint64_t)VM_KERNEL_UNSLIDE(kctl->connect);
2180 xkr->xkr_disconnect =
2181 (uint64_t)VM_KERNEL_UNSLIDE(kctl->disconnect);
2182 xkr->xkr_send = (uint64_t)VM_KERNEL_UNSLIDE(kctl->send);
2183 xkr->xkr_send_list =
2184 (uint64_t)VM_KERNEL_UNSLIDE(kctl->send_list);
2185 xkr->xkr_setopt = (uint64_t)VM_KERNEL_UNSLIDE(kctl->setopt);
2186 xkr->xkr_getopt = (uint64_t)VM_KERNEL_UNSLIDE(kctl->getopt);
2187 xkr->xkr_rcvd = (uint64_t)VM_KERNEL_UNSLIDE(kctl->rcvd);
2188 strlcpy(xkr->xkr_name, kctl->name, sizeof(xkr->xkr_name));
2189
2190 error = SYSCTL_OUT(req, buf, item_size);
2191 }
2192
2193 if (error == 0) {
2194 /*
2195 * Give the user an updated idea of our state.
2196 * If the generation differs from what we told
2197 * her before, she knows that something happened
2198 * while we were processing this request, and it
2199 * might be necessary to retry.
2200 */
2201 bzero(&xsg, sizeof(xsg));
2202 xsg.xg_len = sizeof(xsg);
2203 xsg.xg_count = n;
2204 xsg.xg_gen = kctlstat.kcs_gencnt;
2205 xsg.xg_sogen = so_gencnt;
2206 error = SYSCTL_OUT(req, &xsg, sizeof(xsg));
2207 if (error) {
2208 goto done;
2209 }
2210 }
2211
2212 done:
2213 lck_mtx_unlock(&ctl_mtx);
2214
2215 kfree_data(buf, item_size);
2216
2217 return error;
2218 }
2219
2220 __private_extern__ int
2221 kctl_pcblist SYSCTL_HANDLER_ARGS
2222 {
2223 #pragma unused(oidp, arg1, arg2)
2224 int error = 0;
2225 u_int64_t n, i;
2226 struct xsystmgen xsg;
2227 void *buf = NULL;
2228 struct kctl *kctl;
2229 size_t item_size = ROUNDUP64(sizeof(struct xkctlpcb)) +
2230 ROUNDUP64(sizeof(struct xsocket_n)) +
2231 2 * ROUNDUP64(sizeof(struct xsockbuf_n)) +
2232 ROUNDUP64(sizeof(struct xsockstat_n));
2233
2234 buf = kalloc_data(item_size, Z_WAITOK | Z_ZERO | Z_NOFAIL);
2235
2236 lck_mtx_lock(&ctl_mtx);
2237
2238 n = kctlstat.kcs_pcbcount;
2239
2240 if (req->oldptr == USER_ADDR_NULL) {
2241 req->oldidx = (size_t)(n + n / 8) * item_size;
2242 goto done;
2243 }
2244 if (req->newptr != USER_ADDR_NULL) {
2245 error = EPERM;
2246 goto done;
2247 }
2248 bzero(&xsg, sizeof(xsg));
2249 xsg.xg_len = sizeof(xsg);
2250 xsg.xg_count = n;
2251 xsg.xg_gen = kctlstat.kcs_gencnt;
2252 xsg.xg_sogen = so_gencnt;
2253 error = SYSCTL_OUT(req, &xsg, sizeof(xsg));
2254 if (error) {
2255 goto done;
2256 }
2257 /*
2258 * We are done if there is no pcb
2259 */
2260 if (n == 0) {
2261 goto done;
2262 }
2263
2264 for (i = 0, kctl = TAILQ_FIRST(&ctl_head);
2265 i < n && kctl != NULL;
2266 kctl = TAILQ_NEXT(kctl, next)) {
2267 struct ctl_cb *kcb;
2268
2269 for (kcb = TAILQ_FIRST(&kctl->kcb_head);
2270 i < n && kcb != NULL;
2271 i++, kcb = TAILQ_NEXT(kcb, next)) {
2272 struct xkctlpcb *xk = (struct xkctlpcb *)buf;
2273 struct xsocket_n *xso = (struct xsocket_n *)
2274 ADVANCE64(xk, sizeof(*xk));
2275 struct xsockbuf_n *xsbrcv = (struct xsockbuf_n *)
2276 ADVANCE64(xso, sizeof(*xso));
2277 struct xsockbuf_n *xsbsnd = (struct xsockbuf_n *)
2278 ADVANCE64(xsbrcv, sizeof(*xsbrcv));
2279 struct xsockstat_n *xsostats = (struct xsockstat_n *)
2280 ADVANCE64(xsbsnd, sizeof(*xsbsnd));
2281
2282 bzero(buf, item_size);
2283
2284 xk->xkp_len = sizeof(struct xkctlpcb);
2285 xk->xkp_kind = XSO_KCB;
2286 xk->xkp_unit = kcb->sac.sc_unit;
2287 xk->xkp_kctpcb = (uint64_t)VM_KERNEL_ADDRPERM(kcb);
2288 xk->xkp_kctlref = (uint64_t)VM_KERNEL_ADDRPERM(kctl);
2289 xk->xkp_kctlid = kctl->id;
2290 strlcpy(xk->xkp_kctlname, kctl->name,
2291 sizeof(xk->xkp_kctlname));
2292
2293 sotoxsocket_n(kcb->so, xso);
2294 sbtoxsockbuf_n(kcb->so ?
2295 &kcb->so->so_rcv : NULL, xsbrcv);
2296 sbtoxsockbuf_n(kcb->so ?
2297 &kcb->so->so_snd : NULL, xsbsnd);
2298 sbtoxsockstat_n(kcb->so, xsostats);
2299
2300 error = SYSCTL_OUT(req, buf, item_size);
2301 }
2302 }
2303
2304 if (error == 0) {
2305 /*
2306 * Give the user an updated idea of our state.
2307 * If the generation differs from what we told
2308 * her before, she knows that something happened
2309 * while we were processing this request, and it
2310 * might be necessary to retry.
2311 */
2312 bzero(&xsg, sizeof(xsg));
2313 xsg.xg_len = sizeof(xsg);
2314 xsg.xg_count = n;
2315 xsg.xg_gen = kctlstat.kcs_gencnt;
2316 xsg.xg_sogen = so_gencnt;
2317 error = SYSCTL_OUT(req, &xsg, sizeof(xsg));
2318 if (error) {
2319 goto done;
2320 }
2321 }
2322
2323 done:
2324 lck_mtx_unlock(&ctl_mtx);
2325
2326 kfree_data(buf, item_size);
2327 return error;
2328 }
2329
2330 int
2331 kctl_getstat SYSCTL_HANDLER_ARGS
2332 {
2333 #pragma unused(oidp, arg1, arg2)
2334 int error = 0;
2335
2336 lck_mtx_lock(&ctl_mtx);
2337
2338 if (req->newptr != USER_ADDR_NULL) {
2339 error = EPERM;
2340 goto done;
2341 }
2342 if (req->oldptr == USER_ADDR_NULL) {
2343 req->oldidx = sizeof(struct kctlstat);
2344 goto done;
2345 }
2346
2347 error = SYSCTL_OUT(req, &kctlstat,
2348 MIN(sizeof(struct kctlstat), req->oldlen));
2349 done:
2350 lck_mtx_unlock(&ctl_mtx);
2351 return error;
2352 }
2353
2354 void
2355 kctl_fill_socketinfo(struct socket *so, struct socket_info *si)
2356 {
2357 struct ctl_cb *kcb = (struct ctl_cb *)so->so_pcb;
2358 struct kern_ctl_info *kcsi =
2359 &si->soi_proto.pri_kern_ctl;
2360 struct kctl *kctl = kcb->kctl;
2361
2362 si->soi_kind = SOCKINFO_KERN_CTL;
2363
2364 if (kctl == 0) {
2365 return;
2366 }
2367
2368 kcsi->kcsi_id = kctl->id;
2369 kcsi->kcsi_reg_unit = kctl->reg_unit;
2370 kcsi->kcsi_flags = kctl->flags;
2371 kcsi->kcsi_recvbufsize = kctl->recvbufsize;
2372 kcsi->kcsi_sendbufsize = kctl->sendbufsize;
2373 kcsi->kcsi_unit = kcb->sac.sc_unit;
2374 strlcpy(kcsi->kcsi_name, kctl->name, MAX_KCTL_NAME);
2375 }
2376