xref: /netbsd-src/sys/kern/sysv_msg.c (revision aaf4ece63a859a04e37cf3a7229b5fab0157cc06)
1 /*	$NetBSD: sysv_msg.c,v 1.40 2005/12/07 06:14:13 thorpej Exp $	*/
2 
3 /*-
4  * Copyright (c) 1999 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.
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  * 3. All advertising materials mentioning features or use of this software
20  *    must display the following acknowledgement:
21  *	This product includes software developed by the NetBSD
22  *	Foundation, Inc. and its contributors.
23  * 4. Neither the name of The NetBSD Foundation nor the names of its
24  *    contributors may be used to endorse or promote products derived
25  *    from this software without specific prior written permission.
26  *
27  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
28  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
29  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
30  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
31  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
32  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
33  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
34  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
35  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
36  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
37  * POSSIBILITY OF SUCH DAMAGE.
38  */
39 
40 /*
41  * Implementation of SVID messages
42  *
43  * Author: Daniel Boulet
44  *
45  * Copyright 1993 Daniel Boulet and RTMX Inc.
46  *
47  * This system call was implemented by Daniel Boulet under contract from RTMX.
48  *
49  * Redistribution and use in source forms, with and without modification,
50  * are permitted provided that this entire comment appears intact.
51  *
52  * Redistribution in binary form may occur without any restrictions.
53  * Obviously, it would be nice if you gave credit where credit is due
54  * but requiring it would be too onerous.
55  *
56  * This software is provided ``AS IS'' without any warranties of any kind.
57  */
58 
59 #include <sys/cdefs.h>
60 __KERNEL_RCSID(0, "$NetBSD: sysv_msg.c,v 1.40 2005/12/07 06:14:13 thorpej Exp $");
61 
62 #define SYSVMSG
63 
64 #include <sys/param.h>
65 #include <sys/kernel.h>
66 #include <sys/msg.h>
67 #include <sys/sysctl.h>
68 #include <sys/mount.h>		/* XXX for <sys/syscallargs.h> */
69 #include <sys/sa.h>
70 #include <sys/syscallargs.h>
71 
72 #define MSG_DEBUG
73 #undef MSG_DEBUG_OK
74 
75 #ifdef MSG_DEBUG_OK
76 #define MSG_PRINTF(a)	printf a
77 #else
78 #define MSG_PRINTF(a)
79 #endif
80 
81 static int	nfree_msgmaps;		/* # of free map entries */
82 static short	free_msgmaps;	/* head of linked list of free map entries */
83 static struct	__msg *free_msghdrs;	/* list of free msg headers */
84 static char	*msgpool;		/* MSGMAX byte long msg buffer pool */
85 static struct	msgmap *msgmaps;	/* MSGSEG msgmap structures */
86 static struct __msg *msghdrs;		/* MSGTQL msg headers */
87 struct	msqid_ds *msqids;		/* MSGMNI msqid_ds struct's */
88 
89 static void msg_freehdr(struct __msg *);
90 
91 void
92 msginit(void)
93 {
94 	int i, sz;
95 	vaddr_t v;
96 
97 	/*
98 	 * msginfo.msgssz should be a power of two for efficiency reasons.
99 	 * It is also pretty silly if msginfo.msgssz is less than 8
100 	 * or greater than about 256 so ...
101 	 */
102 
103 	i = 8;
104 	while (i < 1024 && i != msginfo.msgssz)
105 		i <<= 1;
106     	if (i != msginfo.msgssz) {
107 		MSG_PRINTF(("msginfo.msgssz=%d (0x%x)\n", msginfo.msgssz,
108 		    msginfo.msgssz));
109 		panic("msginfo.msgssz not a small power of 2");
110 	}
111 
112 	if (msginfo.msgseg > 32767) {
113 		MSG_PRINTF(("msginfo.msgseg=%d\n", msginfo.msgseg));
114 		panic("msginfo.msgseg > 32767");
115 	}
116 
117 	/* Allocate pageable memory for our structures */
118 	sz = msginfo.msgmax
119 		+ msginfo.msgseg * sizeof(struct msgmap)
120 		+ msginfo.msgtql * sizeof(struct __msg)
121 		+ msginfo.msgmni * sizeof(struct msqid_ds);
122 	v = uvm_km_alloc(kernel_map, round_page(sz), 0,
123 	    UVM_KMF_WIRED|UVM_KMF_ZERO);
124 	if (v == 0)
125 		panic("sysv_msg: cannot allocate memory");
126 	msgpool = (void *)v;
127 	msgmaps = (void *) (msgpool + msginfo.msgmax);
128 	msghdrs = (void *) (msgmaps + msginfo.msgseg);
129 	msqids = (void *) (msghdrs + msginfo.msgtql);
130 
131 	for (i = 0; i < msginfo.msgseg; i++) {
132 		if (i > 0)
133 			msgmaps[i-1].next = i;
134 		msgmaps[i].next = -1;	/* implies entry is available */
135 	}
136 	free_msgmaps = 0;
137 	nfree_msgmaps = msginfo.msgseg;
138 
139 	if (msghdrs == NULL)
140 		panic("msghdrs is NULL");
141 
142 	for (i = 0; i < msginfo.msgtql; i++) {
143 		msghdrs[i].msg_type = 0;
144 		if (i > 0)
145 			msghdrs[i-1].msg_next = &msghdrs[i];
146 		msghdrs[i].msg_next = NULL;
147     	}
148 	free_msghdrs = &msghdrs[0];
149 
150 	if (msqids == NULL)
151 		panic("msqids is NULL");
152 
153 	for (i = 0; i < msginfo.msgmni; i++) {
154 		msqids[i].msg_qbytes = 0;	/* implies entry is available */
155 		msqids[i].msg_perm._seq = 0;	/* reset to a known value */
156 	}
157 }
158 
159 static void
160 msg_freehdr(struct __msg *msghdr)
161 {
162 	while (msghdr->msg_ts > 0) {
163 		short next;
164 		if (msghdr->msg_spot < 0 || msghdr->msg_spot >= msginfo.msgseg)
165 			panic("msghdr->msg_spot out of range");
166 		next = msgmaps[msghdr->msg_spot].next;
167 		msgmaps[msghdr->msg_spot].next = free_msgmaps;
168 		free_msgmaps = msghdr->msg_spot;
169 		nfree_msgmaps++;
170 		msghdr->msg_spot = next;
171 		if (msghdr->msg_ts >= msginfo.msgssz)
172 			msghdr->msg_ts -= msginfo.msgssz;
173 		else
174 			msghdr->msg_ts = 0;
175 	}
176 	if (msghdr->msg_spot != -1)
177 		panic("msghdr->msg_spot != -1");
178 	msghdr->msg_next = free_msghdrs;
179 	free_msghdrs = msghdr;
180 }
181 
182 int
183 sys___msgctl13(struct lwp *l, void *v, register_t *retval)
184 {
185 	struct sys___msgctl13_args /* {
186 		syscallarg(int) msqid;
187 		syscallarg(int) cmd;
188 		syscallarg(struct msqid_ds *) buf;
189 	} */ *uap = v;
190 	struct proc *p = l->l_proc;
191 	struct msqid_ds msqbuf;
192 	int cmd, error;
193 
194 	cmd = SCARG(uap, cmd);
195 
196 	if (cmd == IPC_SET) {
197 		error = copyin(SCARG(uap, buf), &msqbuf, sizeof(msqbuf));
198 		if (error)
199 			return (error);
200 	}
201 
202 	error = msgctl1(p, SCARG(uap, msqid), cmd,
203 	    (cmd == IPC_SET || cmd == IPC_STAT) ? &msqbuf : NULL);
204 
205 	if (error == 0 && cmd == IPC_STAT)
206 		error = copyout(&msqbuf, SCARG(uap, buf), sizeof(msqbuf));
207 
208 	return (error);
209 }
210 
211 int
212 msgctl1(struct proc *p, int msqid, int cmd, struct msqid_ds *msqbuf)
213 {
214 	struct ucred *cred = p->p_ucred;
215 	struct msqid_ds *msqptr;
216 	int error = 0, ix;
217 
218 	MSG_PRINTF(("call to msgctl1(%d, %d)\n", msqid, cmd));
219 
220 	ix = IPCID_TO_IX(msqid);
221 
222 	if (ix < 0 || ix >= msginfo.msgmni) {
223 		MSG_PRINTF(("msqid (%d) out of range (0<=msqid<%d)\n", ix,
224 		    msginfo.msgmni));
225 		return (EINVAL);
226 	}
227 
228 	msqptr = &msqids[ix];
229 
230 	if (msqptr->msg_qbytes == 0) {
231 		MSG_PRINTF(("no such msqid\n"));
232 		return (EINVAL);
233 	}
234 	if (msqptr->msg_perm._seq != IPCID_TO_SEQ(msqid)) {
235 		MSG_PRINTF(("wrong sequence number\n"));
236 		return (EINVAL);
237 	}
238 
239 	switch (cmd) {
240 	case IPC_RMID:
241 	{
242 		struct __msg *msghdr;
243 		if ((error = ipcperm(cred, &msqptr->msg_perm, IPC_M)) != 0)
244 			return (error);
245 		/* Free the message headers */
246 		msghdr = msqptr->_msg_first;
247 		while (msghdr != NULL) {
248 			struct __msg *msghdr_tmp;
249 
250 			/* Free the segments of each message */
251 			msqptr->_msg_cbytes -= msghdr->msg_ts;
252 			msqptr->msg_qnum--;
253 			msghdr_tmp = msghdr;
254 			msghdr = msghdr->msg_next;
255 			msg_freehdr(msghdr_tmp);
256 		}
257 
258 		if (msqptr->_msg_cbytes != 0)
259 			panic("msg_cbytes is screwed up");
260 		if (msqptr->msg_qnum != 0)
261 			panic("msg_qnum is screwed up");
262 
263 		msqptr->msg_qbytes = 0;	/* Mark it as free */
264 
265 		wakeup(msqptr);
266 	}
267 		break;
268 
269 	case IPC_SET:
270 		if ((error = ipcperm(cred, &msqptr->msg_perm, IPC_M)))
271 			return (error);
272 		if (msqbuf->msg_qbytes > msqptr->msg_qbytes && cred->cr_uid != 0)
273 			return (EPERM);
274 		if (msqbuf->msg_qbytes > msginfo.msgmnb) {
275 			MSG_PRINTF(("can't increase msg_qbytes beyond %d "
276 			    "(truncating)\n", msginfo.msgmnb));
277 			/* silently restrict qbytes to system limit */
278 			msqbuf->msg_qbytes = msginfo.msgmnb;
279 		}
280 		if (msqbuf->msg_qbytes == 0) {
281 			MSG_PRINTF(("can't reduce msg_qbytes to 0\n"));
282 			return (EINVAL);	/* XXX non-standard errno! */
283 		}
284 		msqptr->msg_perm.uid = msqbuf->msg_perm.uid;
285 		msqptr->msg_perm.gid = msqbuf->msg_perm.gid;
286 		msqptr->msg_perm.mode = (msqptr->msg_perm.mode & ~0777) |
287 		    (msqbuf->msg_perm.mode & 0777);
288 		msqptr->msg_qbytes = msqbuf->msg_qbytes;
289 		msqptr->msg_ctime = time.tv_sec;
290 		break;
291 
292 	case IPC_STAT:
293 		if ((error = ipcperm(cred, &msqptr->msg_perm, IPC_R))) {
294 			MSG_PRINTF(("requester doesn't have read access\n"));
295 			return (error);
296 		}
297 		memcpy(msqbuf, msqptr, sizeof(struct msqid_ds));
298 		break;
299 
300 	default:
301 		MSG_PRINTF(("invalid command %d\n", cmd));
302 		return (EINVAL);
303 	}
304 
305 	return (error);
306 }
307 
308 int
309 sys_msgget(struct lwp *l, void *v, register_t *retval)
310 {
311 	struct sys_msgget_args /* {
312 		syscallarg(key_t) key;
313 		syscallarg(int) msgflg;
314 	} */ *uap = v;
315 	struct proc *p = l->l_proc;
316 	int msqid, error;
317 	int key = SCARG(uap, key);
318 	int msgflg = SCARG(uap, msgflg);
319 	struct ucred *cred = p->p_ucred;
320 	struct msqid_ds *msqptr = NULL;
321 
322 	MSG_PRINTF(("msgget(0x%x, 0%o)\n", key, msgflg));
323 
324 	if (key != IPC_PRIVATE) {
325 		for (msqid = 0; msqid < msginfo.msgmni; msqid++) {
326 			msqptr = &msqids[msqid];
327 			if (msqptr->msg_qbytes != 0 &&
328 			    msqptr->msg_perm._key == key)
329 				break;
330 		}
331 		if (msqid < msginfo.msgmni) {
332 			MSG_PRINTF(("found public key\n"));
333 			if ((msgflg & IPC_CREAT) && (msgflg & IPC_EXCL)) {
334 				MSG_PRINTF(("not exclusive\n"));
335 				return(EEXIST);
336 			}
337 			if ((error = ipcperm(cred, &msqptr->msg_perm,
338 			    msgflg & 0700 ))) {
339 				MSG_PRINTF(("requester doesn't have 0%o access\n",
340 				    msgflg & 0700));
341 				return (error);
342 			}
343 			goto found;
344 		}
345 	}
346 
347 	MSG_PRINTF(("need to allocate the msqid_ds\n"));
348 	if (key == IPC_PRIVATE || (msgflg & IPC_CREAT)) {
349 		for (msqid = 0; msqid < msginfo.msgmni; msqid++) {
350 			/*
351 			 * Look for an unallocated and unlocked msqid_ds.
352 			 * msqid_ds's can be locked by msgsnd or msgrcv while
353 			 * they are copying the message in/out.  We can't
354 			 * re-use the entry until they release it.
355 			 */
356 			msqptr = &msqids[msqid];
357 			if (msqptr->msg_qbytes == 0 &&
358 			    (msqptr->msg_perm.mode & MSG_LOCKED) == 0)
359 				break;
360 		}
361 		if (msqid == msginfo.msgmni) {
362 			MSG_PRINTF(("no more msqid_ds's available\n"));
363 			return (ENOSPC);
364 		}
365 		MSG_PRINTF(("msqid %d is available\n", msqid));
366 		msqptr->msg_perm._key = key;
367 		msqptr->msg_perm.cuid = cred->cr_uid;
368 		msqptr->msg_perm.uid = cred->cr_uid;
369 		msqptr->msg_perm.cgid = cred->cr_gid;
370 		msqptr->msg_perm.gid = cred->cr_gid;
371 		msqptr->msg_perm.mode = (msgflg & 0777);
372 		/* Make sure that the returned msqid is unique */
373 		msqptr->msg_perm._seq++;
374 		msqptr->_msg_first = NULL;
375 		msqptr->_msg_last = NULL;
376 		msqptr->_msg_cbytes = 0;
377 		msqptr->msg_qnum = 0;
378 		msqptr->msg_qbytes = msginfo.msgmnb;
379 		msqptr->msg_lspid = 0;
380 		msqptr->msg_lrpid = 0;
381 		msqptr->msg_stime = 0;
382 		msqptr->msg_rtime = 0;
383 		msqptr->msg_ctime = time.tv_sec;
384 	} else {
385 		MSG_PRINTF(("didn't find it and wasn't asked to create it\n"));
386 		return (ENOENT);
387 	}
388 
389  found:
390 	/* Construct the unique msqid */
391 	*retval = IXSEQ_TO_IPCID(msqid, msqptr->msg_perm);
392 	return (0);
393 }
394 
395 int
396 sys_msgsnd(struct lwp *l, void *v, register_t *retval)
397 {
398 	struct sys_msgsnd_args /* {
399 		syscallarg(int) msqid;
400 		syscallarg(const void *) msgp;
401 		syscallarg(size_t) msgsz;
402 		syscallarg(int) msgflg;
403 	} */ *uap = v;
404 	struct proc *p = l->l_proc;
405 	int msqid = SCARG(uap, msqid);
406 	const char *user_msgp = SCARG(uap, msgp);
407 	size_t msgsz = SCARG(uap, msgsz);
408 	int msgflg = SCARG(uap, msgflg);
409 	int segs_needed, error;
410 	struct ucred *cred = p->p_ucred;
411 	struct msqid_ds *msqptr;
412 	struct __msg *msghdr;
413 	short next;
414 
415 	MSG_PRINTF(("call to msgsnd(%d, %p, %lld, %d)\n", msqid, user_msgp,
416 	    (long long)msgsz, msgflg));
417 
418 	msqid = IPCID_TO_IX(msqid);
419 
420 	if (msqid < 0 || msqid >= msginfo.msgmni) {
421 		MSG_PRINTF(("msqid (%d) out of range (0<=msqid<%d)\n", msqid,
422 		    msginfo.msgmni));
423 		return (EINVAL);
424 	}
425 
426 	msqptr = &msqids[msqid];
427 	if (msqptr->msg_qbytes == 0) {
428 		MSG_PRINTF(("no such message queue id\n"));
429 		return (EINVAL);
430 	}
431 	if (msqptr->msg_perm._seq != IPCID_TO_SEQ(SCARG(uap, msqid))) {
432 		MSG_PRINTF(("wrong sequence number\n"));
433 		return (EINVAL);
434 	}
435 
436 	if ((error = ipcperm(cred, &msqptr->msg_perm, IPC_W))) {
437 		MSG_PRINTF(("requester doesn't have write access\n"));
438 		return (error);
439 	}
440 
441 	segs_needed = (msgsz + msginfo.msgssz - 1) / msginfo.msgssz;
442 	MSG_PRINTF(("msgsz=%lld, msgssz=%d, segs_needed=%d\n",
443 	    (long long)msgsz, msginfo.msgssz, segs_needed));
444 	for (;;) {
445 		int need_more_resources = 0;
446 
447 		/*
448 		 * check msgsz [cannot be negative since it is unsigned]
449 		 * (inside this loop in case msg_qbytes changes while we sleep)
450 		 */
451 
452 		if (msgsz > msqptr->msg_qbytes) {
453 			MSG_PRINTF(("msgsz > msqptr->msg_qbytes\n"));
454 			return (EINVAL);
455 		}
456 
457 		if (msqptr->msg_perm.mode & MSG_LOCKED) {
458 			MSG_PRINTF(("msqid is locked\n"));
459 			need_more_resources = 1;
460 		}
461 		if (msgsz + msqptr->_msg_cbytes > msqptr->msg_qbytes) {
462 			MSG_PRINTF(("msgsz + msg_cbytes > msg_qbytes\n"));
463 			need_more_resources = 1;
464 		}
465 		if (segs_needed > nfree_msgmaps) {
466 			MSG_PRINTF(("segs_needed > nfree_msgmaps\n"));
467 			need_more_resources = 1;
468 		}
469 		if (free_msghdrs == NULL) {
470 			MSG_PRINTF(("no more msghdrs\n"));
471 			need_more_resources = 1;
472 		}
473 
474 		if (need_more_resources) {
475 			int we_own_it;
476 
477 			if ((msgflg & IPC_NOWAIT) != 0) {
478 				MSG_PRINTF(("need more resources but caller "
479 				    "doesn't want to wait\n"));
480 				return (EAGAIN);
481 			}
482 
483 			if ((msqptr->msg_perm.mode & MSG_LOCKED) != 0) {
484 				MSG_PRINTF(("we don't own the msqid_ds\n"));
485 				we_own_it = 0;
486 			} else {
487 				/* Force later arrivals to wait for our
488 				   request */
489 				MSG_PRINTF(("we own the msqid_ds\n"));
490 				msqptr->msg_perm.mode |= MSG_LOCKED;
491 				we_own_it = 1;
492 			}
493 			MSG_PRINTF(("goodnight\n"));
494 			error = tsleep(msqptr, (PZERO - 4) | PCATCH,
495 			    "msgwait", 0);
496 			MSG_PRINTF(("good morning, error=%d\n", error));
497 			if (we_own_it)
498 				msqptr->msg_perm.mode &= ~MSG_LOCKED;
499 			if (error != 0) {
500 				MSG_PRINTF(("msgsnd: interrupted system "
501 				    "call\n"));
502 				return (EINTR);
503 			}
504 
505 			/*
506 			 * Make sure that the msq queue still exists
507 			 */
508 
509 			if (msqptr->msg_qbytes == 0) {
510 				MSG_PRINTF(("msqid deleted\n"));
511 				return (EIDRM);
512 			}
513 		} else {
514 			MSG_PRINTF(("got all the resources that we need\n"));
515 			break;
516 		}
517 	}
518 
519 	/*
520 	 * We have the resources that we need.
521 	 * Make sure!
522 	 */
523 
524 	if (msqptr->msg_perm.mode & MSG_LOCKED)
525 		panic("msg_perm.mode & MSG_LOCKED");
526 	if (segs_needed > nfree_msgmaps)
527 		panic("segs_needed > nfree_msgmaps");
528 	if (msgsz + msqptr->_msg_cbytes > msqptr->msg_qbytes)
529 		panic("msgsz + msg_cbytes > msg_qbytes");
530 	if (free_msghdrs == NULL)
531 		panic("no more msghdrs");
532 
533 	/*
534 	 * Re-lock the msqid_ds in case we page-fault when copying in the
535 	 * message
536 	 */
537 
538 	if ((msqptr->msg_perm.mode & MSG_LOCKED) != 0)
539 		panic("msqid_ds is already locked");
540 	msqptr->msg_perm.mode |= MSG_LOCKED;
541 
542 	/*
543 	 * Allocate a message header
544 	 */
545 
546 	msghdr = free_msghdrs;
547 	free_msghdrs = msghdr->msg_next;
548 	msghdr->msg_spot = -1;
549 	msghdr->msg_ts = msgsz;
550 
551 	/*
552 	 * Allocate space for the message
553 	 */
554 
555 	while (segs_needed > 0) {
556 		if (nfree_msgmaps <= 0)
557 			panic("not enough msgmaps");
558 		if (free_msgmaps == -1)
559 			panic("nil free_msgmaps");
560 		next = free_msgmaps;
561 		if (next <= -1)
562 			panic("next too low #1");
563 		if (next >= msginfo.msgseg)
564 			panic("next out of range #1");
565 		MSG_PRINTF(("allocating segment %d to message\n", next));
566 		free_msgmaps = msgmaps[next].next;
567 		nfree_msgmaps--;
568 		msgmaps[next].next = msghdr->msg_spot;
569 		msghdr->msg_spot = next;
570 		segs_needed--;
571 	}
572 
573 	/*
574 	 * Copy in the message type
575 	 */
576 
577 	if ((error = copyin(user_msgp, &msghdr->msg_type,
578 	    sizeof(msghdr->msg_type))) != 0) {
579 		MSG_PRINTF(("error %d copying the message type\n", error));
580 		msg_freehdr(msghdr);
581 		msqptr->msg_perm.mode &= ~MSG_LOCKED;
582 		wakeup(msqptr);
583 		return (error);
584 	}
585 	user_msgp += sizeof(msghdr->msg_type);
586 
587 	/*
588 	 * Validate the message type
589 	 */
590 
591 	if (msghdr->msg_type < 1) {
592 		msg_freehdr(msghdr);
593 		msqptr->msg_perm.mode &= ~MSG_LOCKED;
594 		wakeup(msqptr);
595 		MSG_PRINTF(("mtype (%ld) < 1\n", msghdr->msg_type));
596 		return (EINVAL);
597 	}
598 
599 	/*
600 	 * Copy in the message body
601 	 */
602 
603 	next = msghdr->msg_spot;
604 	while (msgsz > 0) {
605 		size_t tlen;
606 		if (msgsz > msginfo.msgssz)
607 			tlen = msginfo.msgssz;
608 		else
609 			tlen = msgsz;
610 		if (next <= -1)
611 			panic("next too low #2");
612 		if (next >= msginfo.msgseg)
613 			panic("next out of range #2");
614 		if ((error = copyin(user_msgp, &msgpool[next * msginfo.msgssz],
615 		    tlen)) != 0) {
616 			MSG_PRINTF(("error %d copying in message segment\n",
617 			    error));
618 			msg_freehdr(msghdr);
619 			msqptr->msg_perm.mode &= ~MSG_LOCKED;
620 			wakeup(msqptr);
621 			return (error);
622 		}
623 		msgsz -= tlen;
624 		user_msgp += tlen;
625 		next = msgmaps[next].next;
626 	}
627 	if (next != -1)
628 		panic("didn't use all the msg segments");
629 
630 	/*
631 	 * We've got the message.  Unlock the msqid_ds.
632 	 */
633 
634 	msqptr->msg_perm.mode &= ~MSG_LOCKED;
635 
636 	/*
637 	 * Make sure that the msqid_ds is still allocated.
638 	 */
639 
640 	if (msqptr->msg_qbytes == 0) {
641 		msg_freehdr(msghdr);
642 		wakeup(msqptr);
643 		return (EIDRM);
644 	}
645 
646 	/*
647 	 * Put the message into the queue
648 	 */
649 
650 	if (msqptr->_msg_first == NULL) {
651 		msqptr->_msg_first = msghdr;
652 		msqptr->_msg_last = msghdr;
653 	} else {
654 		msqptr->_msg_last->msg_next = msghdr;
655 		msqptr->_msg_last = msghdr;
656 	}
657 	msqptr->_msg_last->msg_next = NULL;
658 
659 	msqptr->_msg_cbytes += msghdr->msg_ts;
660 	msqptr->msg_qnum++;
661 	msqptr->msg_lspid = p->p_pid;
662 	msqptr->msg_stime = time.tv_sec;
663 
664 	wakeup(msqptr);
665 	return (0);
666 }
667 
668 int
669 sys_msgrcv(struct lwp *l, void *v, register_t *retval)
670 {
671 	struct sys_msgrcv_args /* {
672 		syscallarg(int) msqid;
673 		syscallarg(void *) msgp;
674 		syscallarg(size_t) msgsz;
675 		syscallarg(long) msgtyp;
676 		syscallarg(int) msgflg;
677 	} */ *uap = v;
678 	struct proc *p = l->l_proc;
679 	int msqid = SCARG(uap, msqid);
680 	char *user_msgp = SCARG(uap, msgp);
681 	size_t msgsz = SCARG(uap, msgsz);
682 	long msgtyp = SCARG(uap, msgtyp);
683 	int msgflg = SCARG(uap, msgflg);
684 	size_t len;
685 	struct ucred *cred = p->p_ucred;
686 	struct msqid_ds *msqptr;
687 	struct __msg *msghdr;
688 	int error;
689 	short next;
690 
691 	MSG_PRINTF(("call to msgrcv(%d, %p, %lld, %ld, %d)\n", msqid,
692 	    user_msgp, (long long)msgsz, msgtyp, msgflg));
693 
694 	msqid = IPCID_TO_IX(msqid);
695 
696 	if (msqid < 0 || msqid >= msginfo.msgmni) {
697 		MSG_PRINTF(("msqid (%d) out of range (0<=msqid<%d)\n", msqid,
698 		    msginfo.msgmni));
699 		return (EINVAL);
700 	}
701 
702 	msqptr = &msqids[msqid];
703 	if (msqptr->msg_qbytes == 0) {
704 		MSG_PRINTF(("no such message queue id\n"));
705 		return (EINVAL);
706 	}
707 	if (msqptr->msg_perm._seq != IPCID_TO_SEQ(SCARG(uap, msqid))) {
708 		MSG_PRINTF(("wrong sequence number\n"));
709 		return (EINVAL);
710 	}
711 
712 	if ((error = ipcperm(cred, &msqptr->msg_perm, IPC_R))) {
713 		MSG_PRINTF(("requester doesn't have read access\n"));
714 		return (error);
715 	}
716 
717 #if 0
718 	/* cannot happen, msgsz is unsigned */
719 	if (msgsz < 0) {
720 		MSG_PRINTF(("msgsz < 0\n"));
721 		return (EINVAL);
722 	}
723 #endif
724 
725 	msghdr = NULL;
726 	while (msghdr == NULL) {
727 		if (msgtyp == 0) {
728 			msghdr = msqptr->_msg_first;
729 			if (msghdr != NULL) {
730 				if (msgsz < msghdr->msg_ts &&
731 				    (msgflg & MSG_NOERROR) == 0) {
732 					MSG_PRINTF(("first message on the "
733 					    "queue is too big "
734 					    "(want %lld, got %d)\n",
735 					    (long long)msgsz, msghdr->msg_ts));
736 					return (E2BIG);
737 				}
738 				if (msqptr->_msg_first == msqptr->_msg_last) {
739 					msqptr->_msg_first = NULL;
740 					msqptr->_msg_last = NULL;
741 				} else {
742 					msqptr->_msg_first = msghdr->msg_next;
743 					if (msqptr->_msg_first == NULL)
744 						panic("msg_first/last screwed "
745 						    "up #1");
746 				}
747 			}
748 		} else {
749 			struct __msg *previous;
750 			struct __msg **prev;
751 
752 			for (previous = NULL, prev = &msqptr->_msg_first;
753 			     (msghdr = *prev) != NULL;
754 			     previous = msghdr, prev = &msghdr->msg_next) {
755 				/*
756 				 * Is this message's type an exact match or is
757 				 * this message's type less than or equal to
758 				 * the absolute value of a negative msgtyp?
759 				 * Note that the second half of this test can
760 				 * NEVER be true if msgtyp is positive since
761 				 * msg_type is always positive!
762 				 */
763 
764 				if (msgtyp == msghdr->msg_type ||
765 				    msghdr->msg_type <= -msgtyp) {
766 					MSG_PRINTF(("found message type %ld, "
767 					    "requested %ld\n",
768 					    msghdr->msg_type, msgtyp));
769 					if (msgsz < msghdr->msg_ts &&
770 					    (msgflg & MSG_NOERROR) == 0) {
771 						MSG_PRINTF(("requested message "
772 						    "on the queue is too big "
773 						    "(want %lld, got %d)\n",
774 						    (long long)msgsz,
775 						    msghdr->msg_ts));
776 						return (E2BIG);
777 					}
778 					*prev = msghdr->msg_next;
779 					if (msghdr == msqptr->_msg_last) {
780 						if (previous == NULL) {
781 							if (prev !=
782 							    &msqptr->_msg_first)
783 								panic("msg_first/last screwed up #2");
784 							msqptr->_msg_first =
785 							    NULL;
786 							msqptr->_msg_last =
787 							    NULL;
788 						} else {
789 							if (prev ==
790 							    &msqptr->_msg_first)
791 								panic("msg_first/last screwed up #3");
792 							msqptr->_msg_last =
793 							    previous;
794 						}
795 					}
796 					break;
797 				}
798 			}
799 		}
800 
801 		/*
802 		 * We've either extracted the msghdr for the appropriate
803 		 * message or there isn't one.
804 		 * If there is one then bail out of this loop.
805 		 */
806 
807 		if (msghdr != NULL)
808 			break;
809 
810 		/*
811 		 * Hmph!  No message found.  Does the user want to wait?
812 		 */
813 
814 		if ((msgflg & IPC_NOWAIT) != 0) {
815 			MSG_PRINTF(("no appropriate message found (msgtyp=%ld)\n",
816 			    msgtyp));
817 			/* The SVID says to return ENOMSG. */
818 #ifdef ENOMSG
819 			return (ENOMSG);
820 #else
821 			/* Unfortunately, BSD doesn't define that code yet! */
822 			return (EAGAIN);
823 #endif
824 		}
825 
826 		/*
827 		 * Wait for something to happen
828 		 */
829 
830 		MSG_PRINTF(("msgrcv:  goodnight\n"));
831 		error = tsleep(msqptr, (PZERO - 4) | PCATCH, "msgwait",
832 		    0);
833 		MSG_PRINTF(("msgrcv: good morning (error=%d)\n", error));
834 
835 		if (error != 0) {
836 			MSG_PRINTF(("msgsnd: interrupted system call\n"));
837 			return (EINTR);
838 		}
839 
840 		/*
841 		 * Make sure that the msq queue still exists
842 		 */
843 
844 		if (msqptr->msg_qbytes == 0 ||
845 		    msqptr->msg_perm._seq != IPCID_TO_SEQ(SCARG(uap, msqid))) {
846 			MSG_PRINTF(("msqid deleted\n"));
847 			return (EIDRM);
848 		}
849 	}
850 
851 	/*
852 	 * Return the message to the user.
853 	 *
854 	 * First, do the bookkeeping (before we risk being interrupted).
855 	 */
856 
857 	msqptr->_msg_cbytes -= msghdr->msg_ts;
858 	msqptr->msg_qnum--;
859 	msqptr->msg_lrpid = p->p_pid;
860 	msqptr->msg_rtime = time.tv_sec;
861 
862 	/*
863 	 * Make msgsz the actual amount that we'll be returning.
864 	 * Note that this effectively truncates the message if it is too long
865 	 * (since msgsz is never increased).
866 	 */
867 
868 	MSG_PRINTF(("found a message, msgsz=%lld, msg_ts=%d\n",
869 	    (long long)msgsz, msghdr->msg_ts));
870 	if (msgsz > msghdr->msg_ts)
871 		msgsz = msghdr->msg_ts;
872 
873 	/*
874 	 * Return the type to the user.
875 	 */
876 
877 	error = copyout(&msghdr->msg_type, user_msgp, sizeof(msghdr->msg_type));
878 	if (error != 0) {
879 		MSG_PRINTF(("error (%d) copying out message type\n", error));
880 		msg_freehdr(msghdr);
881 		wakeup(msqptr);
882 		return (error);
883 	}
884 	user_msgp += sizeof(msghdr->msg_type);
885 
886 	/*
887 	 * Return the segments to the user
888 	 */
889 
890 	next = msghdr->msg_spot;
891 	for (len = 0; len < msgsz; len += msginfo.msgssz) {
892 		size_t tlen;
893 
894 		if (msgsz - len > msginfo.msgssz)
895 			tlen = msginfo.msgssz;
896 		else
897 			tlen = msgsz - len;
898 		if (next <= -1)
899 			panic("next too low #3");
900 		if (next >= msginfo.msgseg)
901 			panic("next out of range #3");
902 		error = copyout(&msgpool[next * msginfo.msgssz],
903 		    user_msgp, tlen);
904 		if (error != 0) {
905 			MSG_PRINTF(("error (%d) copying out message segment\n",
906 			    error));
907 			msg_freehdr(msghdr);
908 			wakeup(msqptr);
909 			return (error);
910 		}
911 		user_msgp += tlen;
912 		next = msgmaps[next].next;
913 	}
914 
915 	/*
916 	 * Done, return the actual number of bytes copied out.
917 	 */
918 
919 	msg_freehdr(msghdr);
920 	wakeup(msqptr);
921 	*retval = msgsz;
922 	return (0);
923 }
924