Home | History | Annotate | Line # | Download | only in kern
sys_descrip.c revision 1.30
      1 /*	$NetBSD: sys_descrip.c,v 1.30 2014/09/05 09:20:59 matt Exp $	*/
      2 
      3 /*-
      4  * Copyright (c) 2008 The NetBSD Foundation, Inc.
      5  * All rights reserved.
      6  *
      7  * Redistribution and use in source and binary forms, with or without
      8  * modification, are permitted provided that the following conditions
      9  * are met:
     10  * 1. Redistributions of source code must retain the above copyright
     11  *    notice, this list of conditions and the following disclaimer.
     12  * 2. Redistributions in binary form must reproduce the above copyright
     13  *    notice, this list of conditions and the following disclaimer in the
     14  *    documentation and/or other materials provided with the distribution.
     15  *
     16  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
     17  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
     18  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
     19  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
     20  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
     21  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
     22  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
     23  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
     24  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
     25  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
     26  * POSSIBILITY OF SUCH DAMAGE.
     27  */
     28 
     29 /*
     30  * Copyright (c) 1982, 1986, 1989, 1991, 1993
     31  *	The Regents of the University of California.  All rights reserved.
     32  * (c) UNIX System Laboratories, Inc.
     33  * All or some portions of this file are derived from material licensed
     34  * to the University of California by American Telephone and Telegraph
     35  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
     36  * the permission of UNIX System Laboratories, Inc.
     37  *
     38  * Redistribution and use in source and binary forms, with or without
     39  * modification, are permitted provided that the following conditions
     40  * are met:
     41  * 1. Redistributions of source code must retain the above copyright
     42  *    notice, this list of conditions and the following disclaimer.
     43  * 2. Redistributions in binary form must reproduce the above copyright
     44  *    notice, this list of conditions and the following disclaimer in the
     45  *    documentation and/or other materials provided with the distribution.
     46  * 3. Neither the name of the University nor the names of its contributors
     47  *    may be used to endorse or promote products derived from this software
     48  *    without specific prior written permission.
     49  *
     50  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     51  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     52  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     53  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     54  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     55  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     56  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     57  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     58  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     59  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     60  * SUCH DAMAGE.
     61  *
     62  *	@(#)kern_descrip.c	8.8 (Berkeley) 2/14/95
     63  */
     64 
     65 /*
     66  * System calls on descriptors.
     67  */
     68 
     69 #include <sys/cdefs.h>
     70 __KERNEL_RCSID(0, "$NetBSD: sys_descrip.c,v 1.30 2014/09/05 09:20:59 matt Exp $");
     71 
     72 #include <sys/param.h>
     73 #include <sys/systm.h>
     74 #include <sys/filedesc.h>
     75 #include <sys/kernel.h>
     76 #include <sys/vnode.h>
     77 #include <sys/proc.h>
     78 #include <sys/file.h>
     79 #include <sys/namei.h>
     80 #include <sys/socket.h>
     81 #include <sys/socketvar.h>
     82 #include <sys/stat.h>
     83 #include <sys/ioctl.h>
     84 #include <sys/fcntl.h>
     85 #include <sys/kmem.h>
     86 #include <sys/pool.h>
     87 #include <sys/syslog.h>
     88 #include <sys/unistd.h>
     89 #include <sys/resourcevar.h>
     90 #include <sys/conf.h>
     91 #include <sys/event.h>
     92 #include <sys/kauth.h>
     93 #include <sys/atomic.h>
     94 #include <sys/mount.h>
     95 #include <sys/syscallargs.h>
     96 
     97 #include <uvm/uvm_readahead.h>
     98 
     99 /*
    100  * Duplicate a file descriptor.
    101  */
    102 int
    103 sys_dup(struct lwp *l, const struct sys_dup_args *uap, register_t *retval)
    104 {
    105 	/* {
    106 		syscallarg(int)	fd;
    107 	} */
    108 	int error, newfd, oldfd;
    109 	file_t *fp;
    110 
    111 	oldfd = SCARG(uap, fd);
    112 
    113 	if ((fp = fd_getfile(oldfd)) == NULL) {
    114 		return EBADF;
    115 	}
    116 	error = fd_dup(fp, 0, &newfd, false);
    117 	fd_putfile(oldfd);
    118 	*retval = newfd;
    119 	return error;
    120 }
    121 
    122 /*
    123  * Duplicate a file descriptor to a particular value.
    124  */
    125 int
    126 dodup(struct lwp *l, int from, int to, int flags, register_t *retval)
    127 {
    128 	int error;
    129 	file_t *fp;
    130 
    131 	if ((fp = fd_getfile(from)) == NULL)
    132 		return EBADF;
    133 	mutex_enter(&fp->f_lock);
    134 	fp->f_count++;
    135 	mutex_exit(&fp->f_lock);
    136 	fd_putfile(from);
    137 
    138 	if ((u_int)to >= curproc->p_rlimit[RLIMIT_NOFILE].rlim_cur ||
    139 	    (u_int)to >= maxfiles)
    140 		error = EBADF;
    141 	else if (from == to)
    142 		error = 0;
    143 	else
    144 		error = fd_dup2(fp, to, flags);
    145 	closef(fp);
    146 	*retval = to;
    147 
    148 	return error;
    149 }
    150 
    151 int
    152 sys_dup3(struct lwp *l, const struct sys_dup3_args *uap, register_t *retval)
    153 {
    154 	/* {
    155 		syscallarg(int)	from;
    156 		syscallarg(int)	to;
    157 		syscallarg(int)	flags;
    158 	} */
    159 	return dodup(l, SCARG(uap, from), SCARG(uap, to), SCARG(uap, flags),
    160 	    retval);
    161 }
    162 
    163 int
    164 sys_dup2(struct lwp *l, const struct sys_dup2_args *uap, register_t *retval)
    165 {
    166 	/* {
    167 		syscallarg(int)	from;
    168 		syscallarg(int)	to;
    169 	} */
    170 	return dodup(l, SCARG(uap, from), SCARG(uap, to), 0, retval);
    171 }
    172 
    173 /*
    174  * fcntl call which is being passed to the file's fs.
    175  */
    176 static int
    177 fcntl_forfs(int fd, file_t *fp, int cmd, void *arg)
    178 {
    179 	int		error;
    180 	u_int		size;
    181 	void		*data, *memp;
    182 #define STK_PARAMS	128
    183 	char		stkbuf[STK_PARAMS];
    184 
    185 	if ((fp->f_flag & (FREAD | FWRITE)) == 0)
    186 		return (EBADF);
    187 
    188 	/*
    189 	 * Interpret high order word to find amount of data to be
    190 	 * copied to/from the user's address space.
    191 	 */
    192 	size = (size_t)F_PARAM_LEN(cmd);
    193 	if (size > F_PARAM_MAX)
    194 		return (EINVAL);
    195 	memp = NULL;
    196 	if (size > sizeof(stkbuf)) {
    197 		memp = kmem_alloc(size, KM_SLEEP);
    198 		data = memp;
    199 	} else
    200 		data = stkbuf;
    201 	if (cmd & F_FSIN) {
    202 		if (size) {
    203 			error = copyin(arg, data, size);
    204 			if (error) {
    205 				if (memp)
    206 					kmem_free(memp, size);
    207 				return (error);
    208 			}
    209 		} else
    210 			*(void **)data = arg;
    211 	} else if ((cmd & F_FSOUT) != 0 && size != 0) {
    212 		/*
    213 		 * Zero the buffer so the user always
    214 		 * gets back something deterministic.
    215 		 */
    216 		memset(data, 0, size);
    217 	} else if (cmd & F_FSVOID)
    218 		*(void **)data = arg;
    219 
    220 
    221 	error = (*fp->f_ops->fo_fcntl)(fp, cmd, data);
    222 
    223 	/*
    224 	 * Copy any data to user, size was
    225 	 * already set and checked above.
    226 	 */
    227 	if (error == 0 && (cmd & F_FSOUT) && size)
    228 		error = copyout(data, arg, size);
    229 	if (memp)
    230 		kmem_free(memp, size);
    231 	return (error);
    232 }
    233 
    234 int
    235 do_fcntl_lock(int fd, int cmd, struct flock *fl)
    236 {
    237 	file_t *fp;
    238 	vnode_t *vp;
    239 	proc_t *p;
    240 	int error, flg;
    241 
    242 	if ((fp = fd_getfile(fd)) == NULL)
    243 		return EBADF;
    244 	if (fp->f_type != DTYPE_VNODE) {
    245 		fd_putfile(fd);
    246 		return EINVAL;
    247 	}
    248 	vp = fp->f_vnode;
    249 	if (fl->l_whence == SEEK_CUR)
    250 		fl->l_start += fp->f_offset;
    251 
    252 	flg = F_POSIX;
    253 	p = curproc;
    254 
    255 	switch (cmd) {
    256 	case F_SETLKW:
    257 		flg |= F_WAIT;
    258 		/* Fall into F_SETLK */
    259 
    260 	case F_SETLK:
    261 		switch (fl->l_type) {
    262 		case F_RDLCK:
    263 			if ((fp->f_flag & FREAD) == 0) {
    264 				error = EBADF;
    265 				break;
    266 			}
    267 			if ((p->p_flag & PK_ADVLOCK) == 0) {
    268 				mutex_enter(p->p_lock);
    269 				p->p_flag |= PK_ADVLOCK;
    270 				mutex_exit(p->p_lock);
    271 			}
    272 			error = VOP_ADVLOCK(vp, p, F_SETLK, fl, flg);
    273 			break;
    274 
    275 		case F_WRLCK:
    276 			if ((fp->f_flag & FWRITE) == 0) {
    277 				error = EBADF;
    278 				break;
    279 			}
    280 			if ((p->p_flag & PK_ADVLOCK) == 0) {
    281 				mutex_enter(p->p_lock);
    282 				p->p_flag |= PK_ADVLOCK;
    283 				mutex_exit(p->p_lock);
    284 			}
    285 			error = VOP_ADVLOCK(vp, p, F_SETLK, fl, flg);
    286 			break;
    287 
    288 		case F_UNLCK:
    289 			error = VOP_ADVLOCK(vp, p, F_UNLCK, fl, F_POSIX);
    290 			break;
    291 
    292 		default:
    293 			error = EINVAL;
    294 			break;
    295 		}
    296 		break;
    297 
    298 	case F_GETLK:
    299 		if (fl->l_type != F_RDLCK &&
    300 		    fl->l_type != F_WRLCK &&
    301 		    fl->l_type != F_UNLCK) {
    302 			error = EINVAL;
    303 			break;
    304 		}
    305 		error = VOP_ADVLOCK(vp, p, F_GETLK, fl, F_POSIX);
    306 		break;
    307 
    308 	default:
    309 		error = EINVAL;
    310 		break;
    311 	}
    312 
    313 	fd_putfile(fd);
    314 	return error;
    315 }
    316 
    317 /*
    318  * The file control system call.
    319  */
    320 int
    321 sys_fcntl(struct lwp *l, const struct sys_fcntl_args *uap, register_t *retval)
    322 {
    323 	/* {
    324 		syscallarg(int)		fd;
    325 		syscallarg(int)		cmd;
    326 		syscallarg(void *)	arg;
    327 	} */
    328 	int fd, i, tmp, error, cmd, newmin;
    329 	filedesc_t *fdp;
    330 	file_t *fp;
    331 	struct flock fl;
    332 	bool cloexec = false;
    333 
    334 	fd = SCARG(uap, fd);
    335 	cmd = SCARG(uap, cmd);
    336 	fdp = l->l_fd;
    337 	error = 0;
    338 
    339 	switch (cmd) {
    340 	case F_CLOSEM:
    341 		if (fd < 0)
    342 			return EBADF;
    343 		while ((i = fdp->fd_lastfile) >= fd) {
    344 			if (fd_getfile(i) == NULL) {
    345 				/* Another thread has updated. */
    346 				continue;
    347 			}
    348 			fd_close(i);
    349 		}
    350 		return 0;
    351 
    352 	case F_MAXFD:
    353 		*retval = fdp->fd_lastfile;
    354 		return 0;
    355 
    356 	case F_SETLKW:
    357 	case F_SETLK:
    358 	case F_GETLK:
    359 		error = copyin(SCARG(uap, arg), &fl, sizeof(fl));
    360 		if (error)
    361 			return error;
    362 		error = do_fcntl_lock(fd, cmd, &fl);
    363 		if (cmd == F_GETLK && error == 0)
    364 			error = copyout(&fl, SCARG(uap, arg), sizeof(fl));
    365 		return error;
    366 
    367 	default:
    368 		/* Handled below */
    369 		break;
    370 	}
    371 
    372 	if ((fp = fd_getfile(fd)) == NULL)
    373 		return (EBADF);
    374 
    375 	if ((cmd & F_FSCTL)) {
    376 		error = fcntl_forfs(fd, fp, cmd, SCARG(uap, arg));
    377 		fd_putfile(fd);
    378 		return error;
    379 	}
    380 
    381 	switch (cmd) {
    382 	case F_DUPFD_CLOEXEC:
    383 		cloexec = true;
    384 		/*FALLTHROUGH*/
    385 	case F_DUPFD:
    386 		newmin = (long)SCARG(uap, arg);
    387 		if ((u_int)newmin >=
    388 		    l->l_proc->p_rlimit[RLIMIT_NOFILE].rlim_cur ||
    389 		    (u_int)newmin >= maxfiles) {
    390 			fd_putfile(fd);
    391 			return EINVAL;
    392 		}
    393 		error = fd_dup(fp, newmin, &i, cloexec);
    394 		*retval = i;
    395 		break;
    396 
    397 	case F_GETFD:
    398 		*retval = fdp->fd_dt->dt_ff[fd]->ff_exclose;
    399 		break;
    400 
    401 	case F_SETFD:
    402 		fd_set_exclose(l, fd,
    403 		    ((long)SCARG(uap, arg) & FD_CLOEXEC) != 0);
    404 		break;
    405 
    406 	case F_GETNOSIGPIPE:
    407 		*retval = (fp->f_flag & FNOSIGPIPE) != 0;
    408 		break;
    409 
    410 	case F_SETNOSIGPIPE:
    411 		if (SCARG(uap, arg))
    412 			atomic_or_uint(&fp->f_flag, FNOSIGPIPE);
    413 		else
    414 			atomic_and_uint(&fp->f_flag, ~FNOSIGPIPE);
    415 		*retval = 0;
    416 		break;
    417 
    418 	case F_GETFL:
    419 		*retval = OFLAGS(fp->f_flag);
    420 		break;
    421 
    422 	case F_SETFL:
    423 		/* XXX not guaranteed to be atomic. */
    424 		tmp = FFLAGS((long)SCARG(uap, arg)) & FCNTLFLAGS;
    425 		error = (*fp->f_ops->fo_fcntl)(fp, F_SETFL, &tmp);
    426 		if (error)
    427 			break;
    428 		i = tmp ^ fp->f_flag;
    429 		if (i & FNONBLOCK) {
    430 			int flgs = tmp & FNONBLOCK;
    431 			error = (*fp->f_ops->fo_ioctl)(fp, FIONBIO, &flgs);
    432 			if (error) {
    433 				(*fp->f_ops->fo_fcntl)(fp, F_SETFL,
    434 				    &fp->f_flag);
    435 				break;
    436 			}
    437 		}
    438 		if (i & FASYNC) {
    439 			int flgs = tmp & FASYNC;
    440 			error = (*fp->f_ops->fo_ioctl)(fp, FIOASYNC, &flgs);
    441 			if (error) {
    442 				if (i & FNONBLOCK) {
    443 					tmp = fp->f_flag & FNONBLOCK;
    444 					(void)(*fp->f_ops->fo_ioctl)(fp,
    445 						FIONBIO, &tmp);
    446 				}
    447 				(*fp->f_ops->fo_fcntl)(fp, F_SETFL,
    448 				    &fp->f_flag);
    449 				break;
    450 			}
    451 		}
    452 		fp->f_flag = (fp->f_flag & ~FCNTLFLAGS) | tmp;
    453 		break;
    454 
    455 	case F_GETOWN:
    456 		error = (*fp->f_ops->fo_ioctl)(fp, FIOGETOWN, &tmp);
    457 		*retval = tmp;
    458 		break;
    459 
    460 	case F_SETOWN:
    461 		tmp = (int)(uintptr_t) SCARG(uap, arg);
    462 		error = (*fp->f_ops->fo_ioctl)(fp, FIOSETOWN, &tmp);
    463 		break;
    464 
    465 	default:
    466 		error = EINVAL;
    467 	}
    468 
    469 	fd_putfile(fd);
    470 	return (error);
    471 }
    472 
    473 /*
    474  * Close a file descriptor.
    475  */
    476 int
    477 sys_close(struct lwp *l, const struct sys_close_args *uap, register_t *retval)
    478 {
    479 	/* {
    480 		syscallarg(int)	fd;
    481 	} */
    482 	int error;
    483 
    484 	if (fd_getfile(SCARG(uap, fd)) == NULL) {
    485 		return EBADF;
    486 	}
    487 
    488 	error = fd_close(SCARG(uap, fd));
    489 	if (error == ERESTART) {
    490 #ifdef DIAGNOSTIC
    491 		printf("pid %d: close returned ERESTART\n",
    492 		    (int)l->l_proc->p_pid);
    493 #endif
    494 		error = EINTR;
    495 	}
    496 
    497 	return error;
    498 }
    499 
    500 /*
    501  * Return status information about a file descriptor.
    502  * Common function for compat code.
    503  */
    504 int
    505 do_sys_fstat(int fd, struct stat *sb)
    506 {
    507 	file_t *fp;
    508 	int error;
    509 
    510 	if ((fp = fd_getfile(fd)) == NULL) {
    511 		return EBADF;
    512 	}
    513 	error = (*fp->f_ops->fo_stat)(fp, sb);
    514 	fd_putfile(fd);
    515 
    516 	return error;
    517 }
    518 
    519 /*
    520  * Return status information about a file descriptor.
    521  */
    522 int
    523 sys___fstat50(struct lwp *l, const struct sys___fstat50_args *uap,
    524 	      register_t *retval)
    525 {
    526 	/* {
    527 		syscallarg(int)			fd;
    528 		syscallarg(struct stat *)	sb;
    529 	} */
    530 	struct stat sb;
    531 	int error;
    532 
    533 	error = do_sys_fstat(SCARG(uap, fd), &sb);
    534 	if (error == 0) {
    535 		error = copyout(&sb, SCARG(uap, sb), sizeof(sb));
    536 	}
    537 	return error;
    538 }
    539 
    540 /*
    541  * Return pathconf information about a file descriptor.
    542  */
    543 int
    544 sys_fpathconf(struct lwp *l, const struct sys_fpathconf_args *uap,
    545 	      register_t *retval)
    546 {
    547 	/* {
    548 		syscallarg(int)	fd;
    549 		syscallarg(int)	name;
    550 	} */
    551 	int fd, error;
    552 	file_t *fp;
    553 
    554 	fd = SCARG(uap, fd);
    555 	error = 0;
    556 
    557 	if ((fp = fd_getfile(fd)) == NULL) {
    558 		return (EBADF);
    559 	}
    560 	switch (fp->f_type) {
    561 	case DTYPE_SOCKET:
    562 	case DTYPE_PIPE:
    563 		if (SCARG(uap, name) != _PC_PIPE_BUF)
    564 			error = EINVAL;
    565 		else
    566 			*retval = PIPE_BUF;
    567 		break;
    568 
    569 	case DTYPE_VNODE:
    570 		error = VOP_PATHCONF(fp->f_vnode, SCARG(uap, name), retval);
    571 		break;
    572 
    573 	case DTYPE_KQUEUE:
    574 		error = EINVAL;
    575 		break;
    576 
    577 	default:
    578 		error = EOPNOTSUPP;
    579 		break;
    580 	}
    581 
    582 	fd_putfile(fd);
    583 	return (error);
    584 }
    585 
    586 /*
    587  * Apply an advisory lock on a file descriptor.
    588  *
    589  * Just attempt to get a record lock of the requested type on
    590  * the entire file (l_whence = SEEK_SET, l_start = 0, l_len = 0).
    591  */
    592 /* ARGSUSED */
    593 int
    594 sys_flock(struct lwp *l, const struct sys_flock_args *uap, register_t *retval)
    595 {
    596 	/* {
    597 		syscallarg(int)	fd;
    598 		syscallarg(int)	how;
    599 	} */
    600 	int fd, how, error;
    601 	file_t *fp;
    602 	vnode_t	*vp;
    603 	struct flock lf;
    604 
    605 	fd = SCARG(uap, fd);
    606 	how = SCARG(uap, how);
    607 	error = 0;
    608 
    609 	if ((fp = fd_getfile(fd)) == NULL) {
    610 		return EBADF;
    611 	}
    612 	if (fp->f_type != DTYPE_VNODE) {
    613 		fd_putfile(fd);
    614 		return EOPNOTSUPP;
    615 	}
    616 
    617 	vp = fp->f_vnode;
    618 	lf.l_whence = SEEK_SET;
    619 	lf.l_start = 0;
    620 	lf.l_len = 0;
    621 
    622 	switch (how & ~LOCK_NB) {
    623 	case LOCK_UN:
    624 		lf.l_type = F_UNLCK;
    625 		atomic_and_uint(&fp->f_flag, ~FHASLOCK);
    626 		error = VOP_ADVLOCK(vp, fp, F_UNLCK, &lf, F_FLOCK);
    627 		fd_putfile(fd);
    628 		return error;
    629 	case LOCK_EX:
    630 		lf.l_type = F_WRLCK;
    631 		break;
    632 	case LOCK_SH:
    633 		lf.l_type = F_RDLCK;
    634 		break;
    635 	default:
    636 		fd_putfile(fd);
    637 		return EINVAL;
    638 	}
    639 
    640 	atomic_or_uint(&fp->f_flag, FHASLOCK);
    641 	if (how & LOCK_NB) {
    642 		error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, F_FLOCK);
    643 	} else {
    644 		error = VOP_ADVLOCK(vp, fp, F_SETLK, &lf, F_FLOCK|F_WAIT);
    645 	}
    646 	fd_putfile(fd);
    647 	return error;
    648 }
    649 
    650 int
    651 do_posix_fadvise(int fd, off_t offset, off_t len, int advice)
    652 {
    653 	file_t *fp;
    654 	vnode_t *vp;
    655 	off_t endoffset;
    656 	int error;
    657 
    658 	CTASSERT(POSIX_FADV_NORMAL == UVM_ADV_NORMAL);
    659 	CTASSERT(POSIX_FADV_RANDOM == UVM_ADV_RANDOM);
    660 	CTASSERT(POSIX_FADV_SEQUENTIAL == UVM_ADV_SEQUENTIAL);
    661 
    662 	if (len == 0) {
    663 		endoffset = INT64_MAX;
    664 	} else if (len > 0 && (INT64_MAX - offset) >= len) {
    665 		endoffset = offset + len;
    666 	} else {
    667 		return EINVAL;
    668 	}
    669 	if ((fp = fd_getfile(fd)) == NULL) {
    670 		return EBADF;
    671 	}
    672 	if (fp->f_type != DTYPE_VNODE) {
    673 		if (fp->f_type == DTYPE_PIPE || fp->f_type == DTYPE_SOCKET) {
    674 			error = ESPIPE;
    675 		} else {
    676 			error = EOPNOTSUPP;
    677 		}
    678 		fd_putfile(fd);
    679 		return error;
    680 	}
    681 
    682 	switch (advice) {
    683 	case POSIX_FADV_WILLNEED:
    684 	case POSIX_FADV_DONTNEED:
    685 		vp = fp->f_vnode;
    686 		if (vp->v_type != VREG && vp->v_type != VBLK) {
    687 			fd_putfile(fd);
    688 			return 0;
    689 		}
    690 		break;
    691 	}
    692 
    693 	switch (advice) {
    694 	case POSIX_FADV_NORMAL:
    695 	case POSIX_FADV_RANDOM:
    696 	case POSIX_FADV_SEQUENTIAL:
    697 		/*
    698 		 * We ignore offset and size.  Must lock the file to
    699 		 * do this, as f_advice is sub-word sized.
    700 		 */
    701 		mutex_enter(&fp->f_lock);
    702 		fp->f_advice = (u_char)advice;
    703 		mutex_exit(&fp->f_lock);
    704 		error = 0;
    705 		break;
    706 
    707 	case POSIX_FADV_WILLNEED:
    708 		vp = fp->f_vnode;
    709 		error = uvm_readahead(&vp->v_uobj, offset, endoffset - offset);
    710 		break;
    711 
    712 	case POSIX_FADV_DONTNEED:
    713 		vp = fp->f_vnode;
    714 		/*
    715 		 * Align the region to page boundaries as VOP_PUTPAGES expects
    716 		 * by shrinking it.  We shrink instead of expand because we
    717 		 * do not want to deactivate cache outside of the requested
    718 		 * region.  It means that if the specified region is smaller
    719 		 * than PAGE_SIZE, we do nothing.
    720 		 */
    721 		if (round_page(offset) < trunc_page(endoffset) &&
    722 		    offset <= round_page(offset)) {
    723 			mutex_enter(vp->v_interlock);
    724 			error = VOP_PUTPAGES(vp,
    725 			    round_page(offset), trunc_page(endoffset),
    726 			    PGO_DEACTIVATE | PGO_CLEANIT);
    727 		} else {
    728 			error = 0;
    729 		}
    730 		break;
    731 
    732 	case POSIX_FADV_NOREUSE:
    733 		/* Not implemented yet. */
    734 		error = 0;
    735 		break;
    736 	default:
    737 		error = EINVAL;
    738 		break;
    739 	}
    740 
    741 	fd_putfile(fd);
    742 	return error;
    743 }
    744 
    745 int
    746 sys___posix_fadvise50(struct lwp *l,
    747 		      const struct sys___posix_fadvise50_args *uap,
    748 		      register_t *retval)
    749 {
    750 	/* {
    751 		syscallarg(int) fd;
    752 		syscallarg(int) pad;
    753 		syscallarg(off_t) offset;
    754 		syscallarg(off_t) len;
    755 		syscallarg(int) advice;
    756 	} */
    757 
    758 	*retval = do_posix_fadvise(SCARG(uap, fd), SCARG(uap, offset),
    759 	    SCARG(uap, len), SCARG(uap, advice));
    760 
    761 	return 0;
    762 }
    763 
    764 int
    765 sys_pipe(struct lwp *l, const void *v, register_t *retval)
    766 {
    767 	return pipe1(l, retval, 0);
    768 }
    769 
    770 int
    771 sys_pipe2(struct lwp *l, const struct sys_pipe2_args *uap, register_t *retval)
    772 {
    773 	/* {
    774 		syscallarg(int[2]) fildes;
    775 		syscallarg(int) flags;
    776 	} */
    777 	int fd[2], error;
    778 
    779 	if ((error = pipe1(l, retval, SCARG(uap, flags))) != 0)
    780 		return error;
    781 	fd[0] = retval[0];
    782 	fd[1] = retval[1];
    783 	if ((error = copyout(fd, SCARG(uap, fildes), sizeof(fd))) != 0)
    784 		return error;
    785 	retval[0] = 0;
    786 	return 0;
    787 }
    788