xref: /netbsd-src/sys/kern/uipc_sem.c (revision bdc22b2e01993381dcefeff2bc9b56ca75a4235c)
1 /*	$NetBSD: uipc_sem.c,v 1.51 2018/05/06 00:46:09 christos Exp $	*/
2 
3 /*-
4  * Copyright (c) 2011 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Mindaugas Rasiukevicius.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
20  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
21  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
22  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
23  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
24  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
25  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
26  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
27  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
28  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29  * POSSIBILITY OF SUCH DAMAGE.
30  */
31 
32 /*
33  * Copyright (c) 2002 Alfred Perlstein <alfred@FreeBSD.org>
34  * All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  *
45  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
46  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
47  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
48  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
49  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
50  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
51  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
52  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
53  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
54  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
55  * SUCH DAMAGE.
56  */
57 
58 /*
59  * Implementation of POSIX semaphore.
60  */
61 
62 #include <sys/cdefs.h>
63 __KERNEL_RCSID(0, "$NetBSD: uipc_sem.c,v 1.51 2018/05/06 00:46:09 christos Exp $");
64 
65 #include <sys/param.h>
66 #include <sys/kernel.h>
67 
68 #include <sys/atomic.h>
69 #include <sys/proc.h>
70 #include <sys/ksem.h>
71 #include <sys/syscall.h>
72 #include <sys/stat.h>
73 #include <sys/kmem.h>
74 #include <sys/fcntl.h>
75 #include <sys/file.h>
76 #include <sys/filedesc.h>
77 #include <sys/kauth.h>
78 #include <sys/module.h>
79 #include <sys/mount.h>
80 #include <sys/semaphore.h>
81 #include <sys/syscall.h>
82 #include <sys/syscallargs.h>
83 #include <sys/syscallvar.h>
84 #include <sys/sysctl.h>
85 
86 MODULE(MODULE_CLASS_MISC, ksem, NULL);
87 
88 #define	SEM_MAX_NAMELEN		NAME_MAX
89 
90 #define	SEM_NSEMS_MAX		256
91 #define	KS_UNLINKED		0x01
92 
93 static kmutex_t		ksem_lock	__cacheline_aligned;
94 static LIST_HEAD(,ksem)	ksem_head	__cacheline_aligned;
95 static u_int		nsems_total	__cacheline_aligned;
96 static u_int		nsems		__cacheline_aligned;
97 
98 static kauth_listener_t	ksem_listener;
99 
100 static int		ksem_sysinit(void);
101 static int		ksem_sysfini(bool);
102 static int		ksem_modcmd(modcmd_t, void *);
103 static int		ksem_close_fop(file_t *);
104 static int		ksem_stat_fop(file_t *, struct stat *);
105 static int		ksem_read_fop(file_t *, off_t *, struct uio *,
106     kauth_cred_t, int);
107 
108 static const struct fileops semops = {
109 	.fo_name = "sem",
110 	.fo_read = ksem_read_fop,
111 	.fo_write = fbadop_write,
112 	.fo_ioctl = fbadop_ioctl,
113 	.fo_fcntl = fnullop_fcntl,
114 	.fo_poll = fnullop_poll,
115 	.fo_stat = ksem_stat_fop,
116 	.fo_close = ksem_close_fop,
117 	.fo_kqfilter = fnullop_kqfilter,
118 	.fo_restart = fnullop_restart,
119 };
120 
121 static const struct syscall_package ksem_syscalls[] = {
122 	{ SYS__ksem_init, 0, (sy_call_t *)sys__ksem_init },
123 	{ SYS__ksem_open, 0, (sy_call_t *)sys__ksem_open },
124 	{ SYS__ksem_unlink, 0, (sy_call_t *)sys__ksem_unlink },
125 	{ SYS__ksem_close, 0, (sy_call_t *)sys__ksem_close },
126 	{ SYS__ksem_post, 0, (sy_call_t *)sys__ksem_post },
127 	{ SYS__ksem_wait, 0, (sy_call_t *)sys__ksem_wait },
128 	{ SYS__ksem_trywait, 0, (sy_call_t *)sys__ksem_trywait },
129 	{ SYS__ksem_getvalue, 0, (sy_call_t *)sys__ksem_getvalue },
130 	{ SYS__ksem_destroy, 0, (sy_call_t *)sys__ksem_destroy },
131 	{ SYS__ksem_timedwait, 0, (sy_call_t *)sys__ksem_timedwait },
132 	{ 0, 0, NULL },
133 };
134 
135 struct sysctllog *ksem_clog;
136 int ksem_max;
137 
138 static int
139 name_copyin(const char *uname, char **name)
140 {
141 	*name = kmem_alloc(SEM_MAX_NAMELEN, KM_SLEEP);
142 
143 	int error = copyinstr(uname, *name, SEM_MAX_NAMELEN, NULL);
144 	if (error)
145 		kmem_free(*name, SEM_MAX_NAMELEN);
146 
147 	return error;
148 }
149 
150 static void
151 name_destroy(char **name)
152 {
153 	if (!*name)
154 		return;
155 
156 	kmem_free(*name, SEM_MAX_NAMELEN);
157 	*name = NULL;
158 }
159 
160 static int
161 ksem_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
162     void *arg0, void *arg1, void *arg2, void *arg3)
163 {
164 	ksem_t *ks;
165 	mode_t mode;
166 
167 	if (action != KAUTH_SYSTEM_SEMAPHORE)
168 		return KAUTH_RESULT_DEFER;
169 
170 	ks = arg1;
171 	mode = ks->ks_mode;
172 
173 	if ((kauth_cred_geteuid(cred) == ks->ks_uid && (mode & S_IWUSR) != 0) ||
174 	    (kauth_cred_getegid(cred) == ks->ks_gid && (mode & S_IWGRP) != 0) ||
175 	    (mode & S_IWOTH) != 0)
176 		return KAUTH_RESULT_ALLOW;
177 
178 	return KAUTH_RESULT_DEFER;
179 }
180 
181 static int
182 ksem_sysinit(void)
183 {
184 	int error;
185 	const struct sysctlnode *rnode;
186 
187 	mutex_init(&ksem_lock, MUTEX_DEFAULT, IPL_NONE);
188 	LIST_INIT(&ksem_head);
189 	nsems_total = 0;
190 	nsems = 0;
191 
192 	error = syscall_establish(NULL, ksem_syscalls);
193 	if (error) {
194 		(void)ksem_sysfini(false);
195 	}
196 
197 	ksem_listener = kauth_listen_scope(KAUTH_SCOPE_SYSTEM,
198 	    ksem_listener_cb, NULL);
199 
200 	/* Define module-specific sysctl tree */
201 
202 	ksem_max = KSEM_MAX;
203 	ksem_clog = NULL;
204 
205 	sysctl_createv(&ksem_clog, 0, NULL, &rnode,
206 			CTLFLAG_PERMANENT,
207 			CTLTYPE_NODE, "posix",
208 			SYSCTL_DESCR("POSIX options"),
209 			NULL, 0, NULL, 0,
210 			CTL_KERN, CTL_CREATE, CTL_EOL);
211 	sysctl_createv(&ksem_clog, 0, &rnode, NULL,
212 			CTLFLAG_PERMANENT | CTLFLAG_READWRITE,
213 			CTLTYPE_INT, "semmax",
214 			SYSCTL_DESCR("Maximal number of semaphores"),
215 			NULL, 0, &ksem_max, 0,
216 			CTL_CREATE, CTL_EOL);
217 	sysctl_createv(&ksem_clog, 0, &rnode, NULL,
218 			CTLFLAG_PERMANENT | CTLFLAG_READONLY,
219 			CTLTYPE_INT, "semcnt",
220 			SYSCTL_DESCR("Current number of semaphores"),
221 			NULL, 0, &nsems, 0,
222 			CTL_CREATE, CTL_EOL);
223 
224 	return error;
225 }
226 
227 static int
228 ksem_sysfini(bool interface)
229 {
230 	int error;
231 
232 	if (interface) {
233 		error = syscall_disestablish(NULL, ksem_syscalls);
234 		if (error != 0) {
235 			return error;
236 		}
237 		/*
238 		 * Make sure that no semaphores are in use.  Note: semops
239 		 * must be unused at this point.
240 		 */
241 		if (nsems_total) {
242 			error = syscall_establish(NULL, ksem_syscalls);
243 			KASSERT(error == 0);
244 			return EBUSY;
245 		}
246 	}
247 	kauth_unlisten_scope(ksem_listener);
248 	mutex_destroy(&ksem_lock);
249 	sysctl_teardown(&ksem_clog);
250 	return 0;
251 }
252 
253 static int
254 ksem_modcmd(modcmd_t cmd, void *arg)
255 {
256 
257 	switch (cmd) {
258 	case MODULE_CMD_INIT:
259 		return ksem_sysinit();
260 
261 	case MODULE_CMD_FINI:
262 		return ksem_sysfini(true);
263 
264 	default:
265 		return ENOTTY;
266 	}
267 }
268 
269 static ksem_t *
270 ksem_lookup(const char *name)
271 {
272 	ksem_t *ks;
273 
274 	KASSERT(mutex_owned(&ksem_lock));
275 
276 	LIST_FOREACH(ks, &ksem_head, ks_entry) {
277 		if (strcmp(ks->ks_name, name) == 0) {
278 			mutex_enter(&ks->ks_lock);
279 			return ks;
280 		}
281 	}
282 	return NULL;
283 }
284 
285 static int
286 ksem_perm(lwp_t *l, ksem_t *ks)
287 {
288 	kauth_cred_t uc = l->l_cred;
289 
290 	KASSERT(mutex_owned(&ks->ks_lock));
291 
292 	if (kauth_authorize_system(uc, KAUTH_SYSTEM_SEMAPHORE, 0, ks, NULL, NULL) != 0)
293 		return EACCES;
294 
295 	return 0;
296 }
297 
298 /*
299  * ksem_get: get the semaphore from the descriptor.
300  *
301  * => locks the semaphore, if found.
302  * => holds a reference on the file descriptor.
303  */
304 static int
305 ksem_get(int fd, ksem_t **ksret)
306 {
307 	ksem_t *ks;
308 	file_t *fp;
309 
310 	fp = fd_getfile(fd);
311 	if (__predict_false(fp == NULL))
312 		return EINVAL;
313 	if (__predict_false(fp->f_type != DTYPE_SEM)) {
314 		fd_putfile(fd);
315 		return EINVAL;
316 	}
317 	ks = fp->f_ksem;
318 	mutex_enter(&ks->ks_lock);
319 
320 	*ksret = ks;
321 	return 0;
322 }
323 
324 /*
325  * ksem_create: allocate and setup a new semaphore structure.
326  */
327 static int
328 ksem_create(lwp_t *l, const char *name, ksem_t **ksret, mode_t mode, u_int val)
329 {
330 	ksem_t *ks;
331 	kauth_cred_t uc;
332 	char *kname;
333 	size_t len;
334 
335 	/* Pre-check for the limit. */
336 	if (nsems >= ksem_max) {
337 		return ENFILE;
338 	}
339 
340 	if (val > SEM_VALUE_MAX) {
341 		return EINVAL;
342 	}
343 
344 	if (name != NULL) {
345 		len = strlen(name);
346 		if (len > SEM_MAX_NAMELEN) {
347 			return ENAMETOOLONG;
348 		}
349 		/* Name must start with a '/' but not contain one. */
350 		if (*name != '/' || len < 2 || strchr(name + 1, '/') != NULL) {
351 			return EINVAL;
352 		}
353 		kname = kmem_alloc(++len, KM_SLEEP);
354 		strlcpy(kname, name, len);
355 	} else {
356 		kname = NULL;
357 		len = 0;
358 	}
359 
360 	if (atomic_inc_uint_nv(&l->l_proc->p_nsems) > SEM_NSEMS_MAX) {
361 		atomic_dec_uint(&l->l_proc->p_nsems);
362 		if (kname != NULL)
363 			kmem_free(kname, len);
364 		return -1;
365 	}
366 
367 	ks = kmem_zalloc(sizeof(ksem_t), KM_SLEEP);
368 	mutex_init(&ks->ks_lock, MUTEX_DEFAULT, IPL_NONE);
369 	cv_init(&ks->ks_cv, "psem");
370 	ks->ks_name = kname;
371 	ks->ks_namelen = len;
372 	ks->ks_mode = mode;
373 	ks->ks_value = val;
374 	ks->ks_ref = 1;
375 
376 	uc = l->l_cred;
377 	ks->ks_uid = kauth_cred_geteuid(uc);
378 	ks->ks_gid = kauth_cred_getegid(uc);
379 
380 	atomic_inc_uint(&nsems_total);
381 	*ksret = ks;
382 	return 0;
383 }
384 
385 static void
386 ksem_free(ksem_t *ks)
387 {
388 
389 	KASSERT(!cv_has_waiters(&ks->ks_cv));
390 
391 	if (ks->ks_name) {
392 		KASSERT(ks->ks_namelen > 0);
393 		kmem_free(ks->ks_name, ks->ks_namelen);
394 	}
395 	mutex_destroy(&ks->ks_lock);
396 	cv_destroy(&ks->ks_cv);
397 	kmem_free(ks, sizeof(ksem_t));
398 
399 	atomic_dec_uint(&nsems_total);
400  	atomic_dec_uint(&curproc->p_nsems);
401 }
402 
403 int
404 sys__ksem_init(struct lwp *l, const struct sys__ksem_init_args *uap,
405     register_t *retval)
406 {
407 	/* {
408 		unsigned int value;
409 		intptr_t *idp;
410 	} */
411 
412 	return do_ksem_init(l, SCARG(uap, value), SCARG(uap, idp), copyout);
413 }
414 
415 int
416 do_ksem_init(lwp_t *l, u_int val, intptr_t *idp, copyout_t docopyout)
417 {
418 	proc_t *p = l->l_proc;
419 	ksem_t *ks;
420 	file_t *fp;
421 	intptr_t id;
422 	int fd, error;
423 
424 	error = fd_allocfile(&fp, &fd);
425 	if (error) {
426 		return error;
427 	}
428 	fp->f_type = DTYPE_SEM;
429 	fp->f_flag = FREAD | FWRITE;
430 	fp->f_ops = &semops;
431 
432 	id = (intptr_t)fd;
433 	error = (*docopyout)(&id, idp, sizeof(*idp));
434 	if (error) {
435 		fd_abort(p, fp, fd);
436 		return error;
437 	}
438 
439 	/* Note the mode does not matter for anonymous semaphores. */
440 	error = ksem_create(l, NULL, &ks, 0, val);
441 	if (error) {
442 		fd_abort(p, fp, fd);
443 		return error;
444 	}
445 	fp->f_ksem = ks;
446 	fd_affix(p, fp, fd);
447 	return error;
448 }
449 
450 int
451 sys__ksem_open(struct lwp *l, const struct sys__ksem_open_args *uap,
452     register_t *retval)
453 {
454 	/* {
455 		const char *name;
456 		int oflag;
457 		mode_t mode;
458 		unsigned int value;
459 		intptr_t *idp;
460 	} */
461 
462 	return do_ksem_open(l, SCARG(uap, name), SCARG(uap, oflag),
463 	    SCARG(uap, mode), SCARG(uap, value), SCARG(uap, idp), copyout);
464 }
465 
466 int
467 do_ksem_open(struct lwp *l, const char *semname, int oflag, mode_t mode,
468      unsigned int value, intptr_t *idp, copyout_t docopyout)
469 {
470 	char *name;
471 	proc_t *p = l->l_proc;
472 	ksem_t *ksnew = NULL, *ks;
473 	file_t *fp;
474 	intptr_t id;
475 	int fd, error;
476 
477 	error = name_copyin(semname, &name);
478 	if (error) {
479 		return error;
480 	}
481 	error = fd_allocfile(&fp, &fd);
482 	if (error) {
483 		name_destroy(&name);
484 		return error;
485 	}
486 	fp->f_type = DTYPE_SEM;
487 	fp->f_flag = FREAD | FWRITE;
488 	fp->f_ops = &semops;
489 
490 	/*
491 	 * The ID (file descriptor number) can be stored early.
492 	 * Note that zero is a special value for libpthread.
493 	 */
494 	id = (intptr_t)fd;
495 	error = (*docopyout)(&id, idp, sizeof(*idp));
496 	if (error) {
497 		goto err;
498 	}
499 
500 	if (oflag & O_CREAT) {
501 		/* Create a new semaphore. */
502 		error = ksem_create(l, name, &ksnew, mode, value);
503 		if (error) {
504 			goto err;
505 		}
506 		KASSERT(ksnew != NULL);
507 	}
508 
509 	/* Lookup for a semaphore with such name. */
510 	mutex_enter(&ksem_lock);
511 	ks = ksem_lookup(name);
512 	name_destroy(&name);
513 	if (ks) {
514 		KASSERT(mutex_owned(&ks->ks_lock));
515 		mutex_exit(&ksem_lock);
516 
517 		/* Check for exclusive create. */
518 		if (oflag & O_EXCL) {
519 			mutex_exit(&ks->ks_lock);
520 			error = EEXIST;
521 			goto err;
522 		}
523 		/*
524 		 * Verify permissions.  If we can access it,
525 		 * add the reference of this thread.
526 		 */
527 		error = ksem_perm(l, ks);
528 		if (error == 0) {
529 			ks->ks_ref++;
530 		}
531 		mutex_exit(&ks->ks_lock);
532 		if (error) {
533 			goto err;
534 		}
535 	} else {
536 		/* Fail if not found and not creating. */
537 		if ((oflag & O_CREAT) == 0) {
538 			mutex_exit(&ksem_lock);
539 			KASSERT(ksnew == NULL);
540 			error = ENOENT;
541 			goto err;
542 		}
543 
544 		/* Check for the limit locked. */
545 		if (nsems >= ksem_max) {
546 			mutex_exit(&ksem_lock);
547 			error = ENFILE;
548 			goto err;
549 		}
550 
551 		/*
552 		 * Finally, insert semaphore into the list.
553 		 * Note: it already has the initial reference.
554 		 */
555 		ks = ksnew;
556 		LIST_INSERT_HEAD(&ksem_head, ks, ks_entry);
557 		nsems++;
558 		mutex_exit(&ksem_lock);
559 
560 		ksnew = NULL;
561 	}
562 	KASSERT(ks != NULL);
563 	fp->f_ksem = ks;
564 	fd_affix(p, fp, fd);
565 err:
566 	name_destroy(&name);
567 	if (error) {
568 		fd_abort(p, fp, fd);
569 	}
570 	if (ksnew) {
571 		ksem_free(ksnew);
572 	}
573 	return error;
574 }
575 
576 int
577 sys__ksem_close(struct lwp *l, const struct sys__ksem_close_args *uap,
578     register_t *retval)
579 {
580 	/* {
581 		intptr_t id;
582 	} */
583 	int fd = (int)SCARG(uap, id);
584 
585 	if (fd_getfile(fd) == NULL) {
586 		return EBADF;
587 	}
588 	return fd_close(fd);
589 }
590 
591 static int
592 ksem_read_fop(file_t *fp, off_t *offset, struct uio *uio, kauth_cred_t cred,
593     int flags)
594 {
595 	size_t len;
596 	char *name;
597 	ksem_t *ks = fp->f_ksem;
598 
599 	mutex_enter(&ks->ks_lock);
600 	len = ks->ks_namelen;
601 	name = ks->ks_name;
602 	mutex_exit(&ks->ks_lock);
603 	if (name == NULL || len == 0)
604 		return 0;
605 	return uiomove(name, len, uio);
606 }
607 
608 static int
609 ksem_stat_fop(file_t *fp, struct stat *ub)
610 {
611 	ksem_t *ks = fp->f_ksem;
612 
613 	mutex_enter(&ks->ks_lock);
614 
615 	memset(ub, 0, sizeof(*ub));
616 
617 	ub->st_mode = ks->ks_mode | ((ks->ks_name && ks->ks_namelen)
618 	    ? _S_IFLNK : _S_IFREG);
619 	ub->st_uid = ks->ks_uid;
620 	ub->st_gid = ks->ks_gid;
621 	ub->st_size = ks->ks_value;
622 	ub->st_blocks = (ub->st_size) ? 1 : 0;
623 	ub->st_nlink = ks->ks_ref;
624 	ub->st_blksize = 4096;
625 
626 	nanotime(&ub->st_atimespec);
627 	ub->st_mtimespec = ub->st_ctimespec = ub->st_birthtimespec =
628 	    ub->st_atimespec;
629 
630 	/*
631 	 * Left as 0: st_dev, st_ino, st_rdev, st_flags, st_gen.
632 	 * XXX (st_dev, st_ino) should be unique.
633 	 */
634 	mutex_exit(&ks->ks_lock);
635 	return 0;
636 }
637 
638 static int
639 ksem_close_fop(file_t *fp)
640 {
641 	ksem_t *ks = fp->f_ksem;
642 	bool destroy = false;
643 
644 	mutex_enter(&ks->ks_lock);
645 	KASSERT(ks->ks_ref > 0);
646 	if (--ks->ks_ref == 0) {
647 		/*
648 		 * Destroy if the last reference and semaphore is unnamed,
649 		 * or unlinked (for named semaphore).
650 		 */
651 		destroy = (ks->ks_flags & KS_UNLINKED) || (ks->ks_name == NULL);
652 	}
653 	mutex_exit(&ks->ks_lock);
654 
655 	if (destroy) {
656 		ksem_free(ks);
657 	}
658 	return 0;
659 }
660 
661 int
662 sys__ksem_unlink(struct lwp *l, const struct sys__ksem_unlink_args *uap,
663     register_t *retval)
664 {
665 	/* {
666 		const char *name;
667 	} */
668 	char *name;
669 	ksem_t *ks;
670 	u_int refcnt;
671 	int error;
672 
673 	error = name_copyin(SCARG(uap, name), &name);
674 	if (error)
675 		return error;
676 
677 	mutex_enter(&ksem_lock);
678 	ks = ksem_lookup(name);
679 	name_destroy(&name);
680 	if (ks == NULL) {
681 		mutex_exit(&ksem_lock);
682 		return ENOENT;
683 	}
684 	KASSERT(mutex_owned(&ks->ks_lock));
685 
686 	/* Verify permissions. */
687 	error = ksem_perm(l, ks);
688 	if (error) {
689 		mutex_exit(&ks->ks_lock);
690 		mutex_exit(&ksem_lock);
691 		return error;
692 	}
693 
694 	/* Remove from the global list. */
695 	LIST_REMOVE(ks, ks_entry);
696 	nsems--;
697 	mutex_exit(&ksem_lock);
698 
699 	refcnt = ks->ks_ref;
700 	if (refcnt) {
701 		/* Mark as unlinked, if there are references. */
702 		ks->ks_flags |= KS_UNLINKED;
703 	}
704 	mutex_exit(&ks->ks_lock);
705 
706 	if (refcnt == 0) {
707 		ksem_free(ks);
708 	}
709 	return 0;
710 }
711 
712 int
713 sys__ksem_post(struct lwp *l, const struct sys__ksem_post_args *uap,
714     register_t *retval)
715 {
716 	/* {
717 		intptr_t id;
718 	} */
719 	int fd = (int)SCARG(uap, id), error;
720 	ksem_t *ks;
721 
722 	error = ksem_get(fd, &ks);
723 	if (error) {
724 		return error;
725 	}
726 	KASSERT(mutex_owned(&ks->ks_lock));
727 	if (ks->ks_value == SEM_VALUE_MAX) {
728 		error = EOVERFLOW;
729 		goto out;
730 	}
731 	ks->ks_value++;
732 	if (ks->ks_waiters) {
733 		cv_broadcast(&ks->ks_cv);
734 	}
735 out:
736 	mutex_exit(&ks->ks_lock);
737 	fd_putfile(fd);
738 	return error;
739 }
740 
741 int
742 do_ksem_wait(lwp_t *l, intptr_t id, bool try_p, struct timespec *abstime)
743 {
744 	int fd = (int)id, error, timeo;
745 	ksem_t *ks;
746 
747 	error = ksem_get(fd, &ks);
748 	if (error) {
749 		return error;
750 	}
751 	KASSERT(mutex_owned(&ks->ks_lock));
752 	while (ks->ks_value == 0) {
753 		ks->ks_waiters++;
754 		if (!try_p && abstime != NULL) {
755 			error = ts2timo(CLOCK_REALTIME, TIMER_ABSTIME, abstime,
756 			    &timeo, NULL);
757 			if (error != 0)
758 				goto out;
759 		} else {
760 			timeo = 0;
761 		}
762 		error = try_p ? EAGAIN : cv_timedwait_sig(&ks->ks_cv,
763 		    &ks->ks_lock, timeo);
764 		ks->ks_waiters--;
765 		if (error)
766 			goto out;
767 	}
768 	ks->ks_value--;
769 out:
770 	mutex_exit(&ks->ks_lock);
771 	fd_putfile(fd);
772 	return error;
773 }
774 
775 int
776 sys__ksem_wait(struct lwp *l, const struct sys__ksem_wait_args *uap,
777     register_t *retval)
778 {
779 	/* {
780 		intptr_t id;
781 	} */
782 
783 	return do_ksem_wait(l, SCARG(uap, id), false, NULL);
784 }
785 
786 int
787 sys__ksem_timedwait(struct lwp *l, const struct sys__ksem_timedwait_args *uap,
788     register_t *retval)
789 {
790 	/* {
791 		intptr_t id;
792 		const struct timespec *abstime;
793 	} */
794 	struct timespec ts;
795 	int error;
796 
797 	error = copyin(SCARG(uap, abstime), &ts, sizeof(ts));
798 	if (error != 0)
799 		return error;
800 
801 	if (ts.tv_sec < 0 || ts.tv_nsec < 0 || ts.tv_nsec >= 1000000000)
802 		return EINVAL;
803 
804 	error = do_ksem_wait(l, SCARG(uap, id), false, &ts);
805 	if (error == EWOULDBLOCK)
806 		error = ETIMEDOUT;
807 	return error;
808 }
809 
810 int
811 sys__ksem_trywait(struct lwp *l, const struct sys__ksem_trywait_args *uap,
812     register_t *retval)
813 {
814 	/* {
815 		intptr_t id;
816 	} */
817 
818 	return do_ksem_wait(l, SCARG(uap, id), true, NULL);
819 }
820 
821 int
822 sys__ksem_getvalue(struct lwp *l, const struct sys__ksem_getvalue_args *uap,
823     register_t *retval)
824 {
825 	/* {
826 		intptr_t id;
827 		unsigned int *value;
828 	} */
829 	int fd = (int)SCARG(uap, id), error;
830 	ksem_t *ks;
831 	unsigned int val;
832 
833 	error = ksem_get(fd, &ks);
834 	if (error) {
835 		return error;
836 	}
837 	KASSERT(mutex_owned(&ks->ks_lock));
838 	val = ks->ks_value;
839 	mutex_exit(&ks->ks_lock);
840 	fd_putfile(fd);
841 
842 	return copyout(&val, SCARG(uap, value), sizeof(val));
843 }
844 
845 int
846 sys__ksem_destroy(struct lwp *l, const struct sys__ksem_destroy_args *uap,
847     register_t *retval)
848 {
849 	/* {
850 		intptr_t id;
851 	} */
852 	int fd = (int)SCARG(uap, id), error;
853 	ksem_t *ks;
854 
855 	error = ksem_get(fd, &ks);
856 	if (error) {
857 		return error;
858 	}
859 	KASSERT(mutex_owned(&ks->ks_lock));
860 
861 	/* Operation is only for unnamed semaphores. */
862 	if (ks->ks_name != NULL) {
863 		error = EINVAL;
864 		goto out;
865 	}
866 	/* Cannot destroy if there are waiters. */
867 	if (ks->ks_waiters) {
868 		error = EBUSY;
869 		goto out;
870 	}
871 out:
872 	mutex_exit(&ks->ks_lock);
873 	if (error) {
874 		fd_putfile(fd);
875 		return error;
876 	}
877 	return fd_close(fd);
878 }
879