xref: /netbsd-src/sys/kern/sysv_msg.c (revision c29d51755812ace2e87aeefdb06cb2b4dac7087a)
1 /*	$NetBSD: sysv_msg.c,v 1.69 2015/05/13 01:16:15 pgoyette Exp $	*/
2 
3 /*-
4  * Copyright (c) 1999, 2006, 2007 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Jason R. Thorpe of the Numerical Aerospace Simulation Facility,
9  * NASA Ames Research Center, and by Andrew Doran.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
21  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
22  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
23  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
24  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
25  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
26  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
27  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
28  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
29  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
30  * POSSIBILITY OF SUCH DAMAGE.
31  */
32 
33 /*
34  * Implementation of SVID messages
35  *
36  * Author: Daniel Boulet
37  *
38  * Copyright 1993 Daniel Boulet and RTMX Inc.
39  *
40  * This system call was implemented by Daniel Boulet under contract from RTMX.
41  *
42  * Redistribution and use in source forms, with and without modification,
43  * are permitted provided that this entire comment appears intact.
44  *
45  * Redistribution in binary form may occur without any restrictions.
46  * Obviously, it would be nice if you gave credit where credit is due
47  * but requiring it would be too onerous.
48  *
49  * This software is provided ``AS IS'' without any warranties of any kind.
50  */
51 
52 #include <sys/cdefs.h>
53 __KERNEL_RCSID(0, "$NetBSD: sysv_msg.c,v 1.69 2015/05/13 01:16:15 pgoyette Exp $");
54 
55 #ifdef _KERNEL_OPT
56 #include "opt_sysv.h"
57 #endif
58 
59 #include <sys/param.h>
60 #include <sys/kernel.h>
61 #include <sys/msg.h>
62 #include <sys/sysctl.h>
63 #include <sys/mount.h>		/* XXX for <sys/syscallargs.h> */
64 #include <sys/syscallargs.h>
65 #include <sys/kauth.h>
66 
67 #define MSG_DEBUG
68 #undef MSG_DEBUG_OK
69 
70 #ifdef MSG_DEBUG_OK
71 #define MSG_PRINTF(a)	printf a
72 #else
73 #define MSG_PRINTF(a)
74 #endif
75 
76 static int	nfree_msgmaps;		/* # of free map entries */
77 static short	free_msgmaps;	/* head of linked list of free map entries */
78 static struct	__msg *free_msghdrs;	/* list of free msg headers */
79 static char	*msgpool;		/* MSGMAX byte long msg buffer pool */
80 static struct	msgmap *msgmaps;	/* MSGSEG msgmap structures */
81 static struct __msg *msghdrs;		/* MSGTQL msg headers */
82 
83 kmsq_t	*msqs;				/* MSGMNI msqid_ds struct's */
84 kmutex_t msgmutex;			/* subsystem lock */
85 
86 static u_int	msg_waiters = 0;	/* total number of msgrcv waiters */
87 static bool	msg_realloc_state;
88 static kcondvar_t msg_realloc_cv;
89 
90 static void msg_freehdr(struct __msg *);
91 
92 extern int kern_has_sysvmsg;
93 
94 void
95 msginit(void)
96 {
97 	int i, sz;
98 	vaddr_t v;
99 
100 	/*
101 	 * msginfo.msgssz should be a power of two for efficiency reasons.
102 	 * It is also pretty silly if msginfo.msgssz is less than 8
103 	 * or greater than about 256 so ...
104 	 */
105 
106 	i = 8;
107 	while (i < 1024 && i != msginfo.msgssz)
108 		i <<= 1;
109 	if (i != msginfo.msgssz) {
110 		panic("msginfo.msgssz = %d, not a small power of 2",
111 		    msginfo.msgssz);
112 	}
113 
114 	if (msginfo.msgseg > 32767) {
115 		panic("msginfo.msgseg = %d > 32767", msginfo.msgseg);
116 	}
117 
118 	/* Allocate the wired memory for our structures */
119 	sz = ALIGN(msginfo.msgmax) +
120 	    ALIGN(msginfo.msgseg * sizeof(struct msgmap)) +
121 	    ALIGN(msginfo.msgtql * sizeof(struct __msg)) +
122 	    ALIGN(msginfo.msgmni * sizeof(kmsq_t));
123 	sz = round_page(sz);
124 	v = uvm_km_alloc(kernel_map, sz, 0, UVM_KMF_WIRED|UVM_KMF_ZERO);
125 	if (v == 0)
126 		panic("sysv_msg: cannot allocate memory");
127 	msgpool = (void *)v;
128 	msgmaps = (void *)((uintptr_t)msgpool + ALIGN(msginfo.msgmax));
129 	msghdrs = (void *)((uintptr_t)msgmaps +
130 	    ALIGN(msginfo.msgseg * sizeof(struct msgmap)));
131 	msqs = (void *)((uintptr_t)msghdrs +
132 	    ALIGN(msginfo.msgtql * sizeof(struct __msg)));
133 
134 	for (i = 0; i < (msginfo.msgseg - 1); i++)
135 		msgmaps[i].next = i + 1;
136 	msgmaps[msginfo.msgseg - 1].next = -1;
137 
138 	free_msgmaps = 0;
139 	nfree_msgmaps = msginfo.msgseg;
140 
141 	for (i = 0; i < (msginfo.msgtql - 1); i++) {
142 		msghdrs[i].msg_type = 0;
143 		msghdrs[i].msg_next = &msghdrs[i + 1];
144 	}
145 	i = msginfo.msgtql - 1;
146 	msghdrs[i].msg_type = 0;
147 	msghdrs[i].msg_next = NULL;
148 	free_msghdrs = &msghdrs[0];
149 
150 	for (i = 0; i < msginfo.msgmni; i++) {
151 		cv_init(&msqs[i].msq_cv, "msgwait");
152 		/* Implies entry is available */
153 		msqs[i].msq_u.msg_qbytes = 0;
154 		/* Reset to a known value */
155 		msqs[i].msq_u.msg_perm._seq = 0;
156 	}
157 
158 	mutex_init(&msgmutex, MUTEX_DEFAULT, IPL_NONE);
159 	cv_init(&msg_realloc_cv, "msgrealc");
160 	msg_realloc_state = false;
161 
162 	kern_has_sysvmsg = 1;
163 
164 	sysvipcinit();
165 }
166 
167 int
168 msgfini(void)
169 {
170 	int i, sz;
171 	vaddr_t v = (vaddr_t)msgpool;
172 
173 	mutex_enter(&msgmutex);
174 	for (i = 0; i < msginfo.msgmni; i++) {
175 		if (msqs[i].msq_u.msg_qbytes != 0) {
176 			mutex_exit(&msgmutex);
177 			return 1; /* queue not available, prevent unload! */
178 		}
179 	}
180 /*
181  * Destroy all condvars and free the memory we're using
182  */
183 	for (i = 0; i < msginfo.msgmni; i++) {
184 		cv_destroy(&msqs[i].msq_cv);
185 	}
186 	sz = ALIGN(msginfo.msgmax) +
187 	    ALIGN(msginfo.msgseg * sizeof(struct msgmap)) +
188 	    ALIGN(msginfo.msgtql * sizeof(struct __msg)) +
189 	    ALIGN(msginfo.msgmni * sizeof(kmsq_t));
190 	sz = round_page(sz);
191 	uvm_km_free(kernel_map, v, sz, UVM_KMF_WIRED);
192 
193 	mutex_exit(&msgmutex);
194 	mutex_destroy(&msgmutex);
195 
196 	kern_has_sysvmsg = 0;
197 
198 	return 0;
199 }
200 
201 static int
202 msgrealloc(int newmsgmni, int newmsgseg)
203 {
204 	struct msgmap *new_msgmaps;
205 	struct __msg *new_msghdrs, *new_free_msghdrs;
206 	char *old_msgpool, *new_msgpool;
207 	kmsq_t *new_msqs;
208 	vaddr_t v;
209 	int i, sz, msqid, newmsgmax, new_nfree_msgmaps;
210 	short new_free_msgmaps;
211 
212 	if (newmsgmni < 1 || newmsgseg < 1)
213 		return EINVAL;
214 
215 	/* Allocate the wired memory for our structures */
216 	newmsgmax = msginfo.msgssz * newmsgseg;
217 	sz = ALIGN(newmsgmax) +
218 	    ALIGN(newmsgseg * sizeof(struct msgmap)) +
219 	    ALIGN(msginfo.msgtql * sizeof(struct __msg)) +
220 	    ALIGN(newmsgmni * sizeof(kmsq_t));
221 	sz = round_page(sz);
222 	v = uvm_km_alloc(kernel_map, sz, 0, UVM_KMF_WIRED|UVM_KMF_ZERO);
223 	if (v == 0)
224 		return ENOMEM;
225 
226 	mutex_enter(&msgmutex);
227 	if (msg_realloc_state) {
228 		mutex_exit(&msgmutex);
229 		uvm_km_free(kernel_map, v, sz, UVM_KMF_WIRED);
230 		return EBUSY;
231 	}
232 	msg_realloc_state = true;
233 	if (msg_waiters) {
234 		/*
235 		 * Mark reallocation state, wake-up all waiters,
236 		 * and wait while they will all exit.
237 		 */
238 		for (i = 0; i < msginfo.msgmni; i++)
239 			cv_broadcast(&msqs[i].msq_cv);
240 		while (msg_waiters)
241 			cv_wait(&msg_realloc_cv, &msgmutex);
242 	}
243 	old_msgpool = msgpool;
244 
245 	/* We cannot reallocate less memory than we use */
246 	i = 0;
247 	for (msqid = 0; msqid < msginfo.msgmni; msqid++) {
248 		struct msqid_ds *mptr;
249 		kmsq_t *msq;
250 
251 		msq = &msqs[msqid];
252 		mptr = &msq->msq_u;
253 		if (mptr->msg_qbytes || (mptr->msg_perm.mode & MSG_LOCKED))
254 			i = msqid;
255 	}
256 	if (i >= newmsgmni || (msginfo.msgseg - nfree_msgmaps) > newmsgseg) {
257 		mutex_exit(&msgmutex);
258 		uvm_km_free(kernel_map, v, sz, UVM_KMF_WIRED);
259 		return EBUSY;
260 	}
261 
262 	new_msgpool = (void *)v;
263 	new_msgmaps = (void *)((uintptr_t)new_msgpool + ALIGN(newmsgmax));
264 	new_msghdrs = (void *)((uintptr_t)new_msgmaps +
265 	    ALIGN(newmsgseg * sizeof(struct msgmap)));
266 	new_msqs = (void *)((uintptr_t)new_msghdrs +
267 	    ALIGN(msginfo.msgtql * sizeof(struct __msg)));
268 
269 	/* Initialize the structures */
270 	for (i = 0; i < (newmsgseg - 1); i++)
271 		new_msgmaps[i].next = i + 1;
272 	new_msgmaps[newmsgseg - 1].next = -1;
273 	new_free_msgmaps = 0;
274 	new_nfree_msgmaps = newmsgseg;
275 
276 	for (i = 0; i < (msginfo.msgtql - 1); i++) {
277 		new_msghdrs[i].msg_type = 0;
278 		new_msghdrs[i].msg_next = &new_msghdrs[i + 1];
279 	}
280 	i = msginfo.msgtql - 1;
281 	new_msghdrs[i].msg_type = 0;
282 	new_msghdrs[i].msg_next = NULL;
283 	new_free_msghdrs = &new_msghdrs[0];
284 
285 	for (i = 0; i < newmsgmni; i++) {
286 		new_msqs[i].msq_u.msg_qbytes = 0;
287 		new_msqs[i].msq_u.msg_perm._seq = 0;
288 		cv_init(&new_msqs[i].msq_cv, "msgwait");
289 	}
290 
291 	/*
292 	 * Copy all message queue identifiers, message headers and buffer
293 	 * pools to the new memory location.
294 	 */
295 	for (msqid = 0; msqid < msginfo.msgmni; msqid++) {
296 		struct __msg *nmsghdr, *msghdr, *pmsghdr;
297 		struct msqid_ds *nmptr, *mptr;
298 		kmsq_t *nmsq, *msq;
299 
300 		msq = &msqs[msqid];
301 		mptr = &msq->msq_u;
302 
303 		if (mptr->msg_qbytes == 0 &&
304 		    (mptr->msg_perm.mode & MSG_LOCKED) == 0)
305 			continue;
306 
307 		nmsq = &new_msqs[msqid];
308 		nmptr = &nmsq->msq_u;
309 		memcpy(nmptr, mptr, sizeof(struct msqid_ds));
310 
311 		/*
312 		 * Go through the message headers, and and copy each
313 		 * one by taking the new ones, and thus defragmenting.
314 		 */
315 		nmsghdr = pmsghdr = NULL;
316 		msghdr = mptr->_msg_first;
317 		while (msghdr) {
318 			short nnext = 0, next;
319 			u_short msgsz, segcnt;
320 
321 			/* Take an entry from the new list of free msghdrs */
322 			nmsghdr = new_free_msghdrs;
323 			KASSERT(nmsghdr != NULL);
324 			new_free_msghdrs = nmsghdr->msg_next;
325 
326 			nmsghdr->msg_next = NULL;
327 			if (pmsghdr) {
328 				pmsghdr->msg_next = nmsghdr;
329 			} else {
330 				nmptr->_msg_first = nmsghdr;
331 				pmsghdr = nmsghdr;
332 			}
333 			nmsghdr->msg_ts = msghdr->msg_ts;
334 			nmsghdr->msg_spot = -1;
335 
336 			/* Compute the amount of segments and reserve them */
337 			msgsz = msghdr->msg_ts;
338 			segcnt = (msgsz + msginfo.msgssz - 1) / msginfo.msgssz;
339 			if (segcnt == 0)
340 				continue;
341 			while (segcnt--) {
342 				nnext = new_free_msgmaps;
343 				new_free_msgmaps = new_msgmaps[nnext].next;
344 				new_nfree_msgmaps--;
345 				new_msgmaps[nnext].next = nmsghdr->msg_spot;
346 				nmsghdr->msg_spot = nnext;
347 			}
348 
349 			/* Copy all segments */
350 			KASSERT(nnext == nmsghdr->msg_spot);
351 			next = msghdr->msg_spot;
352 			while (msgsz > 0) {
353 				size_t tlen;
354 
355 				if (msgsz >= msginfo.msgssz) {
356 					tlen = msginfo.msgssz;
357 					msgsz -= msginfo.msgssz;
358 				} else {
359 					tlen = msgsz;
360 					msgsz = 0;
361 				}
362 
363 				/* Copy the message buffer */
364 				memcpy(&new_msgpool[nnext * msginfo.msgssz],
365 				    &msgpool[next * msginfo.msgssz], tlen);
366 
367 				/* Next entry of the map */
368 				nnext = msgmaps[nnext].next;
369 				next = msgmaps[next].next;
370 			}
371 
372 			/* Next message header */
373 			msghdr = msghdr->msg_next;
374 		}
375 		nmptr->_msg_last = nmsghdr;
376 	}
377 	KASSERT((msginfo.msgseg - nfree_msgmaps) ==
378 	    (newmsgseg - new_nfree_msgmaps));
379 
380 	sz = ALIGN(msginfo.msgmax) +
381 	    ALIGN(msginfo.msgseg * sizeof(struct msgmap)) +
382 	    ALIGN(msginfo.msgtql * sizeof(struct __msg)) +
383 	    ALIGN(msginfo.msgmni * sizeof(kmsq_t));
384 	sz = round_page(sz);
385 
386 	for (i = 0; i < msginfo.msgmni; i++)
387 		cv_destroy(&msqs[i].msq_cv);
388 
389 	/* Set the pointers and update the new values */
390 	msgpool = new_msgpool;
391 	msgmaps = new_msgmaps;
392 	msghdrs = new_msghdrs;
393 	msqs = new_msqs;
394 
395 	free_msghdrs = new_free_msghdrs;
396 	free_msgmaps = new_free_msgmaps;
397 	nfree_msgmaps = new_nfree_msgmaps;
398 	msginfo.msgmni = newmsgmni;
399 	msginfo.msgseg = newmsgseg;
400 	msginfo.msgmax = newmsgmax;
401 
402 	/* Reallocation completed - notify all waiters, if any */
403 	msg_realloc_state = false;
404 	cv_broadcast(&msg_realloc_cv);
405 	mutex_exit(&msgmutex);
406 
407 	uvm_km_free(kernel_map, (vaddr_t)old_msgpool, sz, UVM_KMF_WIRED);
408 	return 0;
409 }
410 
411 static void
412 msg_freehdr(struct __msg *msghdr)
413 {
414 
415 	KASSERT(mutex_owned(&msgmutex));
416 
417 	while (msghdr->msg_ts > 0) {
418 		short next;
419 		KASSERT(msghdr->msg_spot >= 0);
420 		KASSERT(msghdr->msg_spot < msginfo.msgseg);
421 
422 		next = msgmaps[msghdr->msg_spot].next;
423 		msgmaps[msghdr->msg_spot].next = free_msgmaps;
424 		free_msgmaps = msghdr->msg_spot;
425 		nfree_msgmaps++;
426 		msghdr->msg_spot = next;
427 		if (msghdr->msg_ts >= msginfo.msgssz)
428 			msghdr->msg_ts -= msginfo.msgssz;
429 		else
430 			msghdr->msg_ts = 0;
431 	}
432 	KASSERT(msghdr->msg_spot == -1);
433 	msghdr->msg_next = free_msghdrs;
434 	free_msghdrs = msghdr;
435 }
436 
437 int
438 sys___msgctl50(struct lwp *l, const struct sys___msgctl50_args *uap,
439     register_t *retval)
440 {
441 	/* {
442 		syscallarg(int) msqid;
443 		syscallarg(int) cmd;
444 		syscallarg(struct msqid_ds *) buf;
445 	} */
446 	struct msqid_ds msqbuf;
447 	int cmd, error;
448 
449 	cmd = SCARG(uap, cmd);
450 
451 	if (cmd == IPC_SET) {
452 		error = copyin(SCARG(uap, buf), &msqbuf, sizeof(msqbuf));
453 		if (error)
454 			return (error);
455 	}
456 
457 	error = msgctl1(l, SCARG(uap, msqid), cmd,
458 	    (cmd == IPC_SET || cmd == IPC_STAT) ? &msqbuf : NULL);
459 
460 	if (error == 0 && cmd == IPC_STAT)
461 		error = copyout(&msqbuf, SCARG(uap, buf), sizeof(msqbuf));
462 
463 	return (error);
464 }
465 
466 int
467 msgctl1(struct lwp *l, int msqid, int cmd, struct msqid_ds *msqbuf)
468 {
469 	kauth_cred_t cred = l->l_cred;
470 	struct msqid_ds *msqptr;
471 	kmsq_t *msq;
472 	int error = 0, ix;
473 
474 	MSG_PRINTF(("call to msgctl1(%d, %d)\n", msqid, cmd));
475 
476 	ix = IPCID_TO_IX(msqid);
477 
478 	mutex_enter(&msgmutex);
479 
480 	if (ix < 0 || ix >= msginfo.msgmni) {
481 		MSG_PRINTF(("msqid (%d) out of range (0<=msqid<%d)\n", ix,
482 		    msginfo.msgmni));
483 		error = EINVAL;
484 		goto unlock;
485 	}
486 
487 	msq = &msqs[ix];
488 	msqptr = &msq->msq_u;
489 
490 	if (msqptr->msg_qbytes == 0) {
491 		MSG_PRINTF(("no such msqid\n"));
492 		error = EINVAL;
493 		goto unlock;
494 	}
495 	if (msqptr->msg_perm._seq != IPCID_TO_SEQ(msqid)) {
496 		MSG_PRINTF(("wrong sequence number\n"));
497 		error = EINVAL;
498 		goto unlock;
499 	}
500 
501 	switch (cmd) {
502 	case IPC_RMID:
503 	{
504 		struct __msg *msghdr;
505 		if ((error = ipcperm(cred, &msqptr->msg_perm, IPC_M)) != 0)
506 			break;
507 		/* Free the message headers */
508 		msghdr = msqptr->_msg_first;
509 		while (msghdr != NULL) {
510 			struct __msg *msghdr_tmp;
511 
512 			/* Free the segments of each message */
513 			msqptr->_msg_cbytes -= msghdr->msg_ts;
514 			msqptr->msg_qnum--;
515 			msghdr_tmp = msghdr;
516 			msghdr = msghdr->msg_next;
517 			msg_freehdr(msghdr_tmp);
518 		}
519 		KASSERT(msqptr->_msg_cbytes == 0);
520 		KASSERT(msqptr->msg_qnum == 0);
521 
522 		/* Mark it as free */
523 		msqptr->msg_qbytes = 0;
524 		cv_broadcast(&msq->msq_cv);
525 	}
526 		break;
527 
528 	case IPC_SET:
529 		if ((error = ipcperm(cred, &msqptr->msg_perm, IPC_M)))
530 			break;
531 		if (msqbuf->msg_qbytes > msqptr->msg_qbytes &&
532 		    kauth_authorize_system(cred, KAUTH_SYSTEM_SYSVIPC,
533 		    KAUTH_REQ_SYSTEM_SYSVIPC_MSGQ_OVERSIZE,
534 		    KAUTH_ARG(msqbuf->msg_qbytes),
535 		    KAUTH_ARG(msqptr->msg_qbytes), NULL) != 0) {
536 			error = EPERM;
537 			break;
538 		}
539 		if (msqbuf->msg_qbytes > msginfo.msgmnb) {
540 			MSG_PRINTF(("can't increase msg_qbytes beyond %d "
541 			    "(truncating)\n", msginfo.msgmnb));
542 			/* silently restrict qbytes to system limit */
543 			msqbuf->msg_qbytes = msginfo.msgmnb;
544 		}
545 		if (msqbuf->msg_qbytes == 0) {
546 			MSG_PRINTF(("can't reduce msg_qbytes to 0\n"));
547 			error = EINVAL;		/* XXX non-standard errno! */
548 			break;
549 		}
550 		msqptr->msg_perm.uid = msqbuf->msg_perm.uid;
551 		msqptr->msg_perm.gid = msqbuf->msg_perm.gid;
552 		msqptr->msg_perm.mode = (msqptr->msg_perm.mode & ~0777) |
553 		    (msqbuf->msg_perm.mode & 0777);
554 		msqptr->msg_qbytes = msqbuf->msg_qbytes;
555 		msqptr->msg_ctime = time_second;
556 		break;
557 
558 	case IPC_STAT:
559 		if ((error = ipcperm(cred, &msqptr->msg_perm, IPC_R))) {
560 			MSG_PRINTF(("requester doesn't have read access\n"));
561 			break;
562 		}
563 		memcpy(msqbuf, msqptr, sizeof(struct msqid_ds));
564 		break;
565 
566 	default:
567 		MSG_PRINTF(("invalid command %d\n", cmd));
568 		error = EINVAL;
569 		break;
570 	}
571 
572 unlock:
573 	mutex_exit(&msgmutex);
574 	return (error);
575 }
576 
577 int
578 sys_msgget(struct lwp *l, const struct sys_msgget_args *uap, register_t *retval)
579 {
580 	/* {
581 		syscallarg(key_t) key;
582 		syscallarg(int) msgflg;
583 	} */
584 	int msqid, error = 0;
585 	int key = SCARG(uap, key);
586 	int msgflg = SCARG(uap, msgflg);
587 	kauth_cred_t cred = l->l_cred;
588 	struct msqid_ds *msqptr = NULL;
589 	kmsq_t *msq;
590 
591 	mutex_enter(&msgmutex);
592 
593 	MSG_PRINTF(("msgget(0x%x, 0%o)\n", key, msgflg));
594 
595 	if (key != IPC_PRIVATE) {
596 		for (msqid = 0; msqid < msginfo.msgmni; msqid++) {
597 			msq = &msqs[msqid];
598 			msqptr = &msq->msq_u;
599 			if (msqptr->msg_qbytes != 0 &&
600 			    msqptr->msg_perm._key == key)
601 				break;
602 		}
603 		if (msqid < msginfo.msgmni) {
604 			MSG_PRINTF(("found public key\n"));
605 			if ((msgflg & IPC_CREAT) && (msgflg & IPC_EXCL)) {
606 				MSG_PRINTF(("not exclusive\n"));
607 				error = EEXIST;
608 				goto unlock;
609 			}
610 			if ((error = ipcperm(cred, &msqptr->msg_perm,
611 			    msgflg & 0700 ))) {
612 				MSG_PRINTF(("requester doesn't have 0%o access\n",
613 				    msgflg & 0700));
614 				goto unlock;
615 			}
616 			goto found;
617 		}
618 	}
619 
620 	MSG_PRINTF(("need to allocate the msqid_ds\n"));
621 	if (key == IPC_PRIVATE || (msgflg & IPC_CREAT)) {
622 		for (msqid = 0; msqid < msginfo.msgmni; msqid++) {
623 			/*
624 			 * Look for an unallocated and unlocked msqid_ds.
625 			 * msqid_ds's can be locked by msgsnd or msgrcv while
626 			 * they are copying the message in/out.  We can't
627 			 * re-use the entry until they release it.
628 			 */
629 			msq = &msqs[msqid];
630 			msqptr = &msq->msq_u;
631 			if (msqptr->msg_qbytes == 0 &&
632 			    (msqptr->msg_perm.mode & MSG_LOCKED) == 0)
633 				break;
634 		}
635 		if (msqid == msginfo.msgmni) {
636 			MSG_PRINTF(("no more msqid_ds's available\n"));
637 			error = ENOSPC;
638 			goto unlock;
639 		}
640 		MSG_PRINTF(("msqid %d is available\n", msqid));
641 		msqptr->msg_perm._key = key;
642 		msqptr->msg_perm.cuid = kauth_cred_geteuid(cred);
643 		msqptr->msg_perm.uid = kauth_cred_geteuid(cred);
644 		msqptr->msg_perm.cgid = kauth_cred_getegid(cred);
645 		msqptr->msg_perm.gid = kauth_cred_getegid(cred);
646 		msqptr->msg_perm.mode = (msgflg & 0777);
647 		/* Make sure that the returned msqid is unique */
648 		msqptr->msg_perm._seq++;
649 		msqptr->_msg_first = NULL;
650 		msqptr->_msg_last = NULL;
651 		msqptr->_msg_cbytes = 0;
652 		msqptr->msg_qnum = 0;
653 		msqptr->msg_qbytes = msginfo.msgmnb;
654 		msqptr->msg_lspid = 0;
655 		msqptr->msg_lrpid = 0;
656 		msqptr->msg_stime = 0;
657 		msqptr->msg_rtime = 0;
658 		msqptr->msg_ctime = time_second;
659 	} else {
660 		MSG_PRINTF(("didn't find it and wasn't asked to create it\n"));
661 		error = ENOENT;
662 		goto unlock;
663 	}
664 
665 found:
666 	/* Construct the unique msqid */
667 	*retval = IXSEQ_TO_IPCID(msqid, msqptr->msg_perm);
668 
669 unlock:
670 	mutex_exit(&msgmutex);
671 	return (error);
672 }
673 
674 int
675 sys_msgsnd(struct lwp *l, const struct sys_msgsnd_args *uap, register_t *retval)
676 {
677 	/* {
678 		syscallarg(int) msqid;
679 		syscallarg(const void *) msgp;
680 		syscallarg(size_t) msgsz;
681 		syscallarg(int) msgflg;
682 	} */
683 
684 	return msgsnd1(l, SCARG(uap, msqid), SCARG(uap, msgp),
685 	    SCARG(uap, msgsz), SCARG(uap, msgflg), sizeof(long), copyin);
686 }
687 
688 int
689 msgsnd1(struct lwp *l, int msqidr, const char *user_msgp, size_t msgsz,
690     int msgflg, size_t typesz, copyin_t fetch_type)
691 {
692 	int segs_needed, error = 0, msqid;
693 	kauth_cred_t cred = l->l_cred;
694 	struct msqid_ds *msqptr;
695 	struct __msg *msghdr;
696 	kmsq_t *msq;
697 	short next;
698 
699 	MSG_PRINTF(("call to msgsnd(%d, %p, %lld, %d)\n", msqidr,
700 	     user_msgp, (long long)msgsz, msgflg));
701 
702 	if ((ssize_t)msgsz < 0)
703 		return EINVAL;
704 
705 restart:
706 	msqid = IPCID_TO_IX(msqidr);
707 
708 	mutex_enter(&msgmutex);
709 	/* In case of reallocation, we will wait for completion */
710 	while (__predict_false(msg_realloc_state))
711 		cv_wait(&msg_realloc_cv, &msgmutex);
712 
713 	if (msqid < 0 || msqid >= msginfo.msgmni) {
714 		MSG_PRINTF(("msqid (%d) out of range (0<=msqid<%d)\n", msqid,
715 		    msginfo.msgmni));
716 		error = EINVAL;
717 		goto unlock;
718 	}
719 
720 	msq = &msqs[msqid];
721 	msqptr = &msq->msq_u;
722 
723 	if (msqptr->msg_qbytes == 0) {
724 		MSG_PRINTF(("no such message queue id\n"));
725 		error = EINVAL;
726 		goto unlock;
727 	}
728 	if (msqptr->msg_perm._seq != IPCID_TO_SEQ(msqidr)) {
729 		MSG_PRINTF(("wrong sequence number\n"));
730 		error = EINVAL;
731 		goto unlock;
732 	}
733 
734 	if ((error = ipcperm(cred, &msqptr->msg_perm, IPC_W))) {
735 		MSG_PRINTF(("requester doesn't have write access\n"));
736 		goto unlock;
737 	}
738 
739 	segs_needed = (msgsz + msginfo.msgssz - 1) / msginfo.msgssz;
740 	MSG_PRINTF(("msgsz=%lld, msgssz=%d, segs_needed=%d\n",
741 	    (long long)msgsz, msginfo.msgssz, segs_needed));
742 	for (;;) {
743 		int need_more_resources = 0;
744 
745 		/*
746 		 * check msgsz [cannot be negative since it is unsigned]
747 		 * (inside this loop in case msg_qbytes changes while we sleep)
748 		 */
749 
750 		if (msgsz > msqptr->msg_qbytes) {
751 			MSG_PRINTF(("msgsz > msqptr->msg_qbytes\n"));
752 			error = EINVAL;
753 			goto unlock;
754 		}
755 
756 		if (msqptr->msg_perm.mode & MSG_LOCKED) {
757 			MSG_PRINTF(("msqid is locked\n"));
758 			need_more_resources = 1;
759 		}
760 		if (msgsz + msqptr->_msg_cbytes > msqptr->msg_qbytes) {
761 			MSG_PRINTF(("msgsz + msg_cbytes > msg_qbytes\n"));
762 			need_more_resources = 1;
763 		}
764 		if (segs_needed > nfree_msgmaps) {
765 			MSG_PRINTF(("segs_needed > nfree_msgmaps\n"));
766 			need_more_resources = 1;
767 		}
768 		if (free_msghdrs == NULL) {
769 			MSG_PRINTF(("no more msghdrs\n"));
770 			need_more_resources = 1;
771 		}
772 
773 		if (need_more_resources) {
774 			int we_own_it;
775 
776 			if ((msgflg & IPC_NOWAIT) != 0) {
777 				MSG_PRINTF(("need more resources but caller "
778 				    "doesn't want to wait\n"));
779 				error = EAGAIN;
780 				goto unlock;
781 			}
782 
783 			if ((msqptr->msg_perm.mode & MSG_LOCKED) != 0) {
784 				MSG_PRINTF(("we don't own the msqid_ds\n"));
785 				we_own_it = 0;
786 			} else {
787 				/* Force later arrivals to wait for our
788 				   request */
789 				MSG_PRINTF(("we own the msqid_ds\n"));
790 				msqptr->msg_perm.mode |= MSG_LOCKED;
791 				we_own_it = 1;
792 			}
793 
794 			msg_waiters++;
795 			MSG_PRINTF(("goodnight\n"));
796 			error = cv_wait_sig(&msq->msq_cv, &msgmutex);
797 			MSG_PRINTF(("good morning, error=%d\n", error));
798 			msg_waiters--;
799 
800 			if (we_own_it)
801 				msqptr->msg_perm.mode &= ~MSG_LOCKED;
802 
803 			/*
804 			 * In case of such state, notify reallocator and
805 			 * restart the call.
806 			 */
807 			if (msg_realloc_state) {
808 				cv_broadcast(&msg_realloc_cv);
809 				mutex_exit(&msgmutex);
810 				goto restart;
811 			}
812 
813 			if (error != 0) {
814 				MSG_PRINTF(("msgsnd: interrupted system "
815 				    "call\n"));
816 				error = EINTR;
817 				goto unlock;
818 			}
819 
820 			/*
821 			 * Make sure that the msq queue still exists
822 			 */
823 
824 			if (msqptr->msg_qbytes == 0) {
825 				MSG_PRINTF(("msqid deleted\n"));
826 				error = EIDRM;
827 				goto unlock;
828 			}
829 		} else {
830 			MSG_PRINTF(("got all the resources that we need\n"));
831 			break;
832 		}
833 	}
834 
835 	/*
836 	 * We have the resources that we need.
837 	 * Make sure!
838 	 */
839 
840 	KASSERT((msqptr->msg_perm.mode & MSG_LOCKED) == 0);
841 	KASSERT(segs_needed <= nfree_msgmaps);
842 	KASSERT(msgsz + msqptr->_msg_cbytes <= msqptr->msg_qbytes);
843 	KASSERT(free_msghdrs != NULL);
844 
845 	/*
846 	 * Re-lock the msqid_ds in case we page-fault when copying in the
847 	 * message
848 	 */
849 
850 	KASSERT((msqptr->msg_perm.mode & MSG_LOCKED) == 0);
851 	msqptr->msg_perm.mode |= MSG_LOCKED;
852 
853 	/*
854 	 * Allocate a message header
855 	 */
856 
857 	msghdr = free_msghdrs;
858 	free_msghdrs = msghdr->msg_next;
859 	msghdr->msg_spot = -1;
860 	msghdr->msg_ts = msgsz;
861 
862 	/*
863 	 * Allocate space for the message
864 	 */
865 
866 	while (segs_needed > 0) {
867 		KASSERT(nfree_msgmaps > 0);
868 		KASSERT(free_msgmaps != -1);
869 		KASSERT(free_msgmaps < msginfo.msgseg);
870 
871 		next = free_msgmaps;
872 		MSG_PRINTF(("allocating segment %d to message\n", next));
873 		free_msgmaps = msgmaps[next].next;
874 		nfree_msgmaps--;
875 		msgmaps[next].next = msghdr->msg_spot;
876 		msghdr->msg_spot = next;
877 		segs_needed--;
878 	}
879 
880 	/*
881 	 * Copy in the message type
882 	 */
883 	mutex_exit(&msgmutex);
884 	error = (*fetch_type)(user_msgp, &msghdr->msg_type, typesz);
885 	mutex_enter(&msgmutex);
886 	if (error != 0) {
887 		MSG_PRINTF(("error %d copying the message type\n", error));
888 		msg_freehdr(msghdr);
889 		msqptr->msg_perm.mode &= ~MSG_LOCKED;
890 		cv_broadcast(&msq->msq_cv);
891 		goto unlock;
892 	}
893 	user_msgp += typesz;
894 
895 	/*
896 	 * Validate the message type
897 	 */
898 
899 	if (msghdr->msg_type < 1) {
900 		msg_freehdr(msghdr);
901 		msqptr->msg_perm.mode &= ~MSG_LOCKED;
902 		cv_broadcast(&msq->msq_cv);
903 		MSG_PRINTF(("mtype (%ld) < 1\n", msghdr->msg_type));
904 		error = EINVAL;
905 		goto unlock;
906 	}
907 
908 	/*
909 	 * Copy in the message body
910 	 */
911 
912 	next = msghdr->msg_spot;
913 	while (msgsz > 0) {
914 		size_t tlen;
915 		KASSERT(next > -1);
916 		KASSERT(next < msginfo.msgseg);
917 
918 		if (msgsz > msginfo.msgssz)
919 			tlen = msginfo.msgssz;
920 		else
921 			tlen = msgsz;
922 		mutex_exit(&msgmutex);
923 		error = copyin(user_msgp, &msgpool[next * msginfo.msgssz], tlen);
924 		mutex_enter(&msgmutex);
925 		if (error != 0) {
926 			MSG_PRINTF(("error %d copying in message segment\n",
927 			    error));
928 			msg_freehdr(msghdr);
929 			msqptr->msg_perm.mode &= ~MSG_LOCKED;
930 			cv_broadcast(&msq->msq_cv);
931 			goto unlock;
932 		}
933 		msgsz -= tlen;
934 		user_msgp += tlen;
935 		next = msgmaps[next].next;
936 	}
937 	KASSERT(next == -1);
938 
939 	/*
940 	 * We've got the message.  Unlock the msqid_ds.
941 	 */
942 
943 	msqptr->msg_perm.mode &= ~MSG_LOCKED;
944 
945 	/*
946 	 * Make sure that the msqid_ds is still allocated.
947 	 */
948 
949 	if (msqptr->msg_qbytes == 0) {
950 		msg_freehdr(msghdr);
951 		cv_broadcast(&msq->msq_cv);
952 		error = EIDRM;
953 		goto unlock;
954 	}
955 
956 	/*
957 	 * Put the message into the queue
958 	 */
959 
960 	if (msqptr->_msg_first == NULL) {
961 		msqptr->_msg_first = msghdr;
962 		msqptr->_msg_last = msghdr;
963 	} else {
964 		msqptr->_msg_last->msg_next = msghdr;
965 		msqptr->_msg_last = msghdr;
966 	}
967 	msqptr->_msg_last->msg_next = NULL;
968 
969 	msqptr->_msg_cbytes += msghdr->msg_ts;
970 	msqptr->msg_qnum++;
971 	msqptr->msg_lspid = l->l_proc->p_pid;
972 	msqptr->msg_stime = time_second;
973 
974 	cv_broadcast(&msq->msq_cv);
975 
976 unlock:
977 	mutex_exit(&msgmutex);
978 	return error;
979 }
980 
981 int
982 sys_msgrcv(struct lwp *l, const struct sys_msgrcv_args *uap, register_t *retval)
983 {
984 	/* {
985 		syscallarg(int) msqid;
986 		syscallarg(void *) msgp;
987 		syscallarg(size_t) msgsz;
988 		syscallarg(long) msgtyp;
989 		syscallarg(int) msgflg;
990 	} */
991 
992 	return msgrcv1(l, SCARG(uap, msqid), SCARG(uap, msgp),
993 	    SCARG(uap, msgsz), SCARG(uap, msgtyp), SCARG(uap, msgflg),
994 	    sizeof(long), copyout, retval);
995 }
996 
997 int
998 msgrcv1(struct lwp *l, int msqidr, char *user_msgp, size_t msgsz, long msgtyp,
999     int msgflg, size_t typesz, copyout_t put_type, register_t *retval)
1000 {
1001 	size_t len;
1002 	kauth_cred_t cred = l->l_cred;
1003 	struct msqid_ds *msqptr;
1004 	struct __msg *msghdr;
1005 	int error = 0, msqid;
1006 	kmsq_t *msq;
1007 	short next;
1008 
1009 	MSG_PRINTF(("call to msgrcv(%d, %p, %lld, %ld, %d)\n", msqidr,
1010 	    user_msgp, (long long)msgsz, msgtyp, msgflg));
1011 
1012 	if ((ssize_t)msgsz < 0)
1013 		return EINVAL;
1014 
1015 restart:
1016 	msqid = IPCID_TO_IX(msqidr);
1017 
1018 	mutex_enter(&msgmutex);
1019 	/* In case of reallocation, we will wait for completion */
1020 	while (__predict_false(msg_realloc_state))
1021 		cv_wait(&msg_realloc_cv, &msgmutex);
1022 
1023 	if (msqid < 0 || msqid >= msginfo.msgmni) {
1024 		MSG_PRINTF(("msqid (%d) out of range (0<=msqid<%d)\n", msqid,
1025 		    msginfo.msgmni));
1026 		error = EINVAL;
1027 		goto unlock;
1028 	}
1029 
1030 	msq = &msqs[msqid];
1031 	msqptr = &msq->msq_u;
1032 
1033 	if (msqptr->msg_qbytes == 0) {
1034 		MSG_PRINTF(("no such message queue id\n"));
1035 		error = EINVAL;
1036 		goto unlock;
1037 	}
1038 	if (msqptr->msg_perm._seq != IPCID_TO_SEQ(msqidr)) {
1039 		MSG_PRINTF(("wrong sequence number\n"));
1040 		error = EINVAL;
1041 		goto unlock;
1042 	}
1043 
1044 	if ((error = ipcperm(cred, &msqptr->msg_perm, IPC_R))) {
1045 		MSG_PRINTF(("requester doesn't have read access\n"));
1046 		goto unlock;
1047 	}
1048 
1049 	msghdr = NULL;
1050 	while (msghdr == NULL) {
1051 		if (msgtyp == 0) {
1052 			msghdr = msqptr->_msg_first;
1053 			if (msghdr != NULL) {
1054 				if (msgsz < msghdr->msg_ts &&
1055 				    (msgflg & MSG_NOERROR) == 0) {
1056 					MSG_PRINTF(("first msg on the queue "
1057 					    "is too big (want %lld, got %d)\n",
1058 					    (long long)msgsz, msghdr->msg_ts));
1059 					error = E2BIG;
1060 					goto unlock;
1061 				}
1062 				if (msqptr->_msg_first == msqptr->_msg_last) {
1063 					msqptr->_msg_first = NULL;
1064 					msqptr->_msg_last = NULL;
1065 				} else {
1066 					msqptr->_msg_first = msghdr->msg_next;
1067 					KASSERT(msqptr->_msg_first != NULL);
1068 				}
1069 			}
1070 		} else {
1071 			struct __msg *previous;
1072 			struct __msg **prev;
1073 
1074 			for (previous = NULL, prev = &msqptr->_msg_first;
1075 			     (msghdr = *prev) != NULL;
1076 			     previous = msghdr, prev = &msghdr->msg_next) {
1077 				/*
1078 				 * Is this message's type an exact match or is
1079 				 * this message's type less than or equal to
1080 				 * the absolute value of a negative msgtyp?
1081 				 * Note that the second half of this test can
1082 				 * NEVER be true if msgtyp is positive since
1083 				 * msg_type is always positive!
1084 				 */
1085 
1086 				if (msgtyp != msghdr->msg_type &&
1087 				    msghdr->msg_type > -msgtyp)
1088 					continue;
1089 
1090 				MSG_PRINTF(("found message type %ld, requested %ld\n",
1091 				    msghdr->msg_type, msgtyp));
1092 				if (msgsz < msghdr->msg_ts &&
1093 				     (msgflg & MSG_NOERROR) == 0) {
1094 					MSG_PRINTF(("requested message on the queue "
1095 					    "is too big (want %lld, got %d)\n",
1096 					    (long long)msgsz, msghdr->msg_ts));
1097 					error = E2BIG;
1098 					goto unlock;
1099 				}
1100 				*prev = msghdr->msg_next;
1101 				if (msghdr != msqptr->_msg_last)
1102 					break;
1103 				if (previous == NULL) {
1104 					KASSERT(prev == &msqptr->_msg_first);
1105 					msqptr->_msg_first = NULL;
1106 					msqptr->_msg_last = NULL;
1107 				} else {
1108 					KASSERT(prev != &msqptr->_msg_first);
1109 					msqptr->_msg_last = previous;
1110 				}
1111 				break;
1112 			}
1113 		}
1114 
1115 		/*
1116 		 * We've either extracted the msghdr for the appropriate
1117 		 * message or there isn't one.
1118 		 * If there is one then bail out of this loop.
1119 		 */
1120 		if (msghdr != NULL)
1121 			break;
1122 
1123 		/*
1124 		 * Hmph!  No message found.  Does the user want to wait?
1125 		 */
1126 
1127 		if ((msgflg & IPC_NOWAIT) != 0) {
1128 			MSG_PRINTF(("no appropriate message found (msgtyp=%ld)\n",
1129 			    msgtyp));
1130 			error = ENOMSG;
1131 			goto unlock;
1132 		}
1133 
1134 		/*
1135 		 * Wait for something to happen
1136 		 */
1137 
1138 		msg_waiters++;
1139 		MSG_PRINTF(("msgrcv:  goodnight\n"));
1140 		error = cv_wait_sig(&msq->msq_cv, &msgmutex);
1141 		MSG_PRINTF(("msgrcv: good morning (error=%d)\n", error));
1142 		msg_waiters--;
1143 
1144 		/*
1145 		 * In case of such state, notify reallocator and
1146 		 * restart the call.
1147 		 */
1148 		if (msg_realloc_state) {
1149 			cv_broadcast(&msg_realloc_cv);
1150 			mutex_exit(&msgmutex);
1151 			goto restart;
1152 		}
1153 
1154 		if (error != 0) {
1155 			MSG_PRINTF(("msgsnd: interrupted system call\n"));
1156 			error = EINTR;
1157 			goto unlock;
1158 		}
1159 
1160 		/*
1161 		 * Make sure that the msq queue still exists
1162 		 */
1163 
1164 		if (msqptr->msg_qbytes == 0 ||
1165 		    msqptr->msg_perm._seq != IPCID_TO_SEQ(msqidr)) {
1166 			MSG_PRINTF(("msqid deleted\n"));
1167 			error = EIDRM;
1168 			goto unlock;
1169 		}
1170 	}
1171 
1172 	/*
1173 	 * Return the message to the user.
1174 	 *
1175 	 * First, do the bookkeeping (before we risk being interrupted).
1176 	 */
1177 
1178 	msqptr->_msg_cbytes -= msghdr->msg_ts;
1179 	msqptr->msg_qnum--;
1180 	msqptr->msg_lrpid = l->l_proc->p_pid;
1181 	msqptr->msg_rtime = time_second;
1182 
1183 	/*
1184 	 * Make msgsz the actual amount that we'll be returning.
1185 	 * Note that this effectively truncates the message if it is too long
1186 	 * (since msgsz is never increased).
1187 	 */
1188 
1189 	MSG_PRINTF(("found a message, msgsz=%lld, msg_ts=%d\n",
1190 	    (long long)msgsz, msghdr->msg_ts));
1191 	if (msgsz > msghdr->msg_ts)
1192 		msgsz = msghdr->msg_ts;
1193 
1194 	/*
1195 	 * Return the type to the user.
1196 	 */
1197 	mutex_exit(&msgmutex);
1198 	error = (*put_type)(&msghdr->msg_type, user_msgp, typesz);
1199 	mutex_enter(&msgmutex);
1200 	if (error != 0) {
1201 		MSG_PRINTF(("error (%d) copying out message type\n", error));
1202 		msg_freehdr(msghdr);
1203 		cv_broadcast(&msq->msq_cv);
1204 		goto unlock;
1205 	}
1206 	user_msgp += typesz;
1207 
1208 	/*
1209 	 * Return the segments to the user
1210 	 */
1211 
1212 	next = msghdr->msg_spot;
1213 	for (len = 0; len < msgsz; len += msginfo.msgssz) {
1214 		size_t tlen;
1215 		KASSERT(next > -1);
1216 		KASSERT(next < msginfo.msgseg);
1217 
1218 		if (msgsz - len > msginfo.msgssz)
1219 			tlen = msginfo.msgssz;
1220 		else
1221 			tlen = msgsz - len;
1222 		mutex_exit(&msgmutex);
1223 		error = copyout(&msgpool[next * msginfo.msgssz],
1224 		    user_msgp, tlen);
1225 		mutex_enter(&msgmutex);
1226 		if (error != 0) {
1227 			MSG_PRINTF(("error (%d) copying out message segment\n",
1228 			    error));
1229 			msg_freehdr(msghdr);
1230 			cv_broadcast(&msq->msq_cv);
1231 			goto unlock;
1232 		}
1233 		user_msgp += tlen;
1234 		next = msgmaps[next].next;
1235 	}
1236 
1237 	/*
1238 	 * Done, return the actual number of bytes copied out.
1239 	 */
1240 
1241 	msg_freehdr(msghdr);
1242 	cv_broadcast(&msq->msq_cv);
1243 	*retval = msgsz;
1244 
1245 unlock:
1246 	mutex_exit(&msgmutex);
1247 	return error;
1248 }
1249 
1250 /*
1251  * Sysctl initialization and nodes.
1252  */
1253 
1254 static int
1255 sysctl_ipc_msgmni(SYSCTLFN_ARGS)
1256 {
1257 	int newsize, error;
1258 	struct sysctlnode node;
1259 	node = *rnode;
1260 	node.sysctl_data = &newsize;
1261 
1262 	newsize = msginfo.msgmni;
1263 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
1264 	if (error || newp == NULL)
1265 		return error;
1266 
1267 	sysctl_unlock();
1268 	error = msgrealloc(newsize, msginfo.msgseg);
1269 	sysctl_relock();
1270 	return error;
1271 }
1272 
1273 static int
1274 sysctl_ipc_msgseg(SYSCTLFN_ARGS)
1275 {
1276 	int newsize, error;
1277 	struct sysctlnode node;
1278 	node = *rnode;
1279 	node.sysctl_data = &newsize;
1280 
1281 	newsize = msginfo.msgseg;
1282 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
1283 	if (error || newp == NULL)
1284 		return error;
1285 
1286 	sysctl_unlock();
1287 	error = msgrealloc(msginfo.msgmni, newsize);
1288 	sysctl_relock();
1289 	return error;
1290 }
1291 
1292 SYSCTL_SETUP(sysctl_ipc_msg_setup, "sysctl kern.ipc subtree setup")
1293 {
1294 	const struct sysctlnode *node = NULL;
1295 
1296 	sysctl_createv(clog, 0, NULL, &node,
1297 		CTLFLAG_PERMANENT,
1298 		CTLTYPE_NODE, "ipc",
1299 		SYSCTL_DESCR("SysV IPC options"),
1300 		NULL, 0, NULL, 0,
1301 		CTL_KERN, KERN_SYSVIPC, CTL_EOL);
1302 
1303 	if (node == NULL)
1304 		return;
1305 
1306 	sysctl_createv(clog, 0, &node, NULL,
1307 		CTLFLAG_PERMANENT | CTLFLAG_READWRITE,
1308 		CTLTYPE_INT, "msgmni",
1309 		SYSCTL_DESCR("Max number of message queue identifiers"),
1310 		sysctl_ipc_msgmni, 0, &msginfo.msgmni, 0,
1311 		CTL_CREATE, CTL_EOL);
1312 	sysctl_createv(clog, 0, &node, NULL,
1313 		CTLFLAG_PERMANENT | CTLFLAG_READWRITE,
1314 		CTLTYPE_INT, "msgseg",
1315 		SYSCTL_DESCR("Max number of number of message segments"),
1316 		sysctl_ipc_msgseg, 0, &msginfo.msgseg, 0,
1317 		CTL_CREATE, CTL_EOL);
1318 }
1319