Home | History | Annotate | Line # | Download | only in kern
sys_generic.c revision 1.131
      1 /*	$NetBSD: sys_generic.c,v 1.131 2017/06/01 02:45:13 chs Exp $	*/
      2 
      3 /*-
      4  * Copyright (c) 2007, 2008, 2009 The NetBSD Foundation, Inc.
      5  * All rights reserved.
      6  *
      7  * This code is derived from software contributed to The NetBSD Foundation
      8  * by Andrew Doran.
      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) 1982, 1986, 1989, 1993
     34  *	The Regents of the University of California.  All rights reserved.
     35  * (c) UNIX System Laboratories, Inc.
     36  * All or some portions of this file are derived from material licensed
     37  * to the University of California by American Telephone and Telegraph
     38  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
     39  * the permission of UNIX System Laboratories, Inc.
     40  *
     41  * Redistribution and use in source and binary forms, with or without
     42  * modification, are permitted provided that the following conditions
     43  * are met:
     44  * 1. Redistributions of source code must retain the above copyright
     45  *    notice, this list of conditions and the following disclaimer.
     46  * 2. Redistributions in binary form must reproduce the above copyright
     47  *    notice, this list of conditions and the following disclaimer in the
     48  *    documentation and/or other materials provided with the distribution.
     49  * 3. Neither the name of the University nor the names of its contributors
     50  *    may be used to endorse or promote products derived from this software
     51  *    without specific prior written permission.
     52  *
     53  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     54  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     55  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     56  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     57  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     58  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     59  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     60  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     61  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     62  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     63  * SUCH DAMAGE.
     64  *
     65  *	@(#)sys_generic.c	8.9 (Berkeley) 2/14/95
     66  */
     67 
     68 /*
     69  * System calls relating to files.
     70  */
     71 
     72 #include <sys/cdefs.h>
     73 __KERNEL_RCSID(0, "$NetBSD: sys_generic.c,v 1.131 2017/06/01 02:45:13 chs Exp $");
     74 
     75 #include <sys/param.h>
     76 #include <sys/systm.h>
     77 #include <sys/filedesc.h>
     78 #include <sys/ioctl.h>
     79 #include <sys/file.h>
     80 #include <sys/proc.h>
     81 #include <sys/socketvar.h>
     82 #include <sys/signalvar.h>
     83 #include <sys/uio.h>
     84 #include <sys/kernel.h>
     85 #include <sys/stat.h>
     86 #include <sys/kmem.h>
     87 #include <sys/poll.h>
     88 #include <sys/vnode.h>
     89 #include <sys/mount.h>
     90 #include <sys/syscallargs.h>
     91 #include <sys/ktrace.h>
     92 #include <sys/atomic.h>
     93 #include <sys/disklabel.h>
     94 
     95 /*
     96  * Read system call.
     97  */
     98 /* ARGSUSED */
     99 int
    100 sys_read(struct lwp *l, const struct sys_read_args *uap, register_t *retval)
    101 {
    102 	/* {
    103 		syscallarg(int)		fd;
    104 		syscallarg(void *)	buf;
    105 		syscallarg(size_t)	nbyte;
    106 	} */
    107 	file_t *fp;
    108 	int fd;
    109 
    110 	fd = SCARG(uap, fd);
    111 
    112 	if ((fp = fd_getfile(fd)) == NULL)
    113 		return (EBADF);
    114 
    115 	if ((fp->f_flag & FREAD) == 0) {
    116 		fd_putfile(fd);
    117 		return (EBADF);
    118 	}
    119 
    120 	/* dofileread() will unuse the descriptor for us */
    121 	return (dofileread(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
    122 	    &fp->f_offset, FOF_UPDATE_OFFSET, retval));
    123 }
    124 
    125 int
    126 dofileread(int fd, struct file *fp, void *buf, size_t nbyte,
    127 	off_t *offset, int flags, register_t *retval)
    128 {
    129 	struct iovec aiov;
    130 	struct uio auio;
    131 	size_t cnt;
    132 	int error;
    133 	lwp_t *l;
    134 
    135 	l = curlwp;
    136 
    137 	aiov.iov_base = (void *)buf;
    138 	aiov.iov_len = nbyte;
    139 	auio.uio_iov = &aiov;
    140 	auio.uio_iovcnt = 1;
    141 	auio.uio_resid = nbyte;
    142 	auio.uio_rw = UIO_READ;
    143 	auio.uio_vmspace = l->l_proc->p_vmspace;
    144 
    145 	/*
    146 	 * Reads return ssize_t because -1 is returned on error.  Therefore
    147 	 * we must restrict the length to SSIZE_MAX to avoid garbage return
    148 	 * values.
    149 	 */
    150 	if (auio.uio_resid > SSIZE_MAX) {
    151 		error = EINVAL;
    152 		goto out;
    153 	}
    154 
    155 	cnt = auio.uio_resid;
    156 	error = (*fp->f_ops->fo_read)(fp, offset, &auio, fp->f_cred, flags);
    157 	if (error)
    158 		if (auio.uio_resid != cnt && (error == ERESTART ||
    159 		    error == EINTR || error == EWOULDBLOCK))
    160 			error = 0;
    161 	cnt -= auio.uio_resid;
    162 	ktrgenio(fd, UIO_READ, buf, cnt, error);
    163 	*retval = cnt;
    164  out:
    165 	fd_putfile(fd);
    166 	return (error);
    167 }
    168 
    169 /*
    170  * Scatter read system call.
    171  */
    172 int
    173 sys_readv(struct lwp *l, const struct sys_readv_args *uap, register_t *retval)
    174 {
    175 	/* {
    176 		syscallarg(int)				fd;
    177 		syscallarg(const struct iovec *)	iovp;
    178 		syscallarg(int)				iovcnt;
    179 	} */
    180 
    181 	return do_filereadv(SCARG(uap, fd), SCARG(uap, iovp),
    182 	    SCARG(uap, iovcnt), NULL, FOF_UPDATE_OFFSET, retval);
    183 }
    184 
    185 int
    186 do_filereadv(int fd, const struct iovec *iovp, int iovcnt,
    187     off_t *offset, int flags, register_t *retval)
    188 {
    189 	struct uio	auio;
    190 	struct iovec	*iov, *needfree = NULL, aiov[UIO_SMALLIOV];
    191 	int		i, error;
    192 	size_t		cnt;
    193 	u_int		iovlen;
    194 	struct file	*fp;
    195 	struct iovec	*ktriov = NULL;
    196 
    197 	if (iovcnt == 0)
    198 		return EINVAL;
    199 
    200 	if ((fp = fd_getfile(fd)) == NULL)
    201 		return EBADF;
    202 
    203 	if ((fp->f_flag & FREAD) == 0) {
    204 		fd_putfile(fd);
    205 		return EBADF;
    206 	}
    207 
    208 	if (offset == NULL)
    209 		offset = &fp->f_offset;
    210 	else {
    211 		struct vnode *vp = fp->f_vnode;
    212 		if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
    213 			error = ESPIPE;
    214 			goto out;
    215 		}
    216 		/*
    217 		 * Test that the device is seekable ?
    218 		 * XXX This works because no file systems actually
    219 		 * XXX take any action on the seek operation.
    220 		 */
    221 		error = VOP_SEEK(vp, fp->f_offset, *offset, fp->f_cred);
    222 		if (error != 0)
    223 			goto out;
    224 	}
    225 
    226 	iovlen = iovcnt * sizeof(struct iovec);
    227 	if (flags & FOF_IOV_SYSSPACE)
    228 		iov = __UNCONST(iovp);
    229 	else {
    230 		iov = aiov;
    231 		if ((u_int)iovcnt > UIO_SMALLIOV) {
    232 			if ((u_int)iovcnt > IOV_MAX) {
    233 				error = EINVAL;
    234 				goto out;
    235 			}
    236 			iov = kmem_alloc(iovlen, KM_SLEEP);
    237 			needfree = iov;
    238 		}
    239 		error = copyin(iovp, iov, iovlen);
    240 		if (error)
    241 			goto done;
    242 	}
    243 
    244 	auio.uio_iov = iov;
    245 	auio.uio_iovcnt = iovcnt;
    246 	auio.uio_rw = UIO_READ;
    247 	auio.uio_vmspace = curproc->p_vmspace;
    248 
    249 	auio.uio_resid = 0;
    250 	for (i = 0; i < iovcnt; i++, iov++) {
    251 		auio.uio_resid += iov->iov_len;
    252 		/*
    253 		 * Reads return ssize_t because -1 is returned on error.
    254 		 * Therefore we must restrict the length to SSIZE_MAX to
    255 		 * avoid garbage return values.
    256 		 */
    257 		if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
    258 			error = EINVAL;
    259 			goto done;
    260 		}
    261 	}
    262 
    263 	/*
    264 	 * if tracing, save a copy of iovec
    265 	 */
    266 	if (ktrpoint(KTR_GENIO))  {
    267 		ktriov = kmem_alloc(iovlen, KM_SLEEP);
    268 		memcpy(ktriov, auio.uio_iov, iovlen);
    269 	}
    270 
    271 	cnt = auio.uio_resid;
    272 	error = (*fp->f_ops->fo_read)(fp, offset, &auio, fp->f_cred, flags);
    273 	if (error)
    274 		if (auio.uio_resid != cnt && (error == ERESTART ||
    275 		    error == EINTR || error == EWOULDBLOCK))
    276 			error = 0;
    277 	cnt -= auio.uio_resid;
    278 	*retval = cnt;
    279 
    280 	if (ktriov != NULL) {
    281 		ktrgeniov(fd, UIO_READ, ktriov, cnt, error);
    282 		kmem_free(ktriov, iovlen);
    283 	}
    284 
    285  done:
    286 	if (needfree)
    287 		kmem_free(needfree, iovlen);
    288  out:
    289 	fd_putfile(fd);
    290 	return (error);
    291 }
    292 
    293 /*
    294  * Write system call
    295  */
    296 int
    297 sys_write(struct lwp *l, const struct sys_write_args *uap, register_t *retval)
    298 {
    299 	/* {
    300 		syscallarg(int)			fd;
    301 		syscallarg(const void *)	buf;
    302 		syscallarg(size_t)		nbyte;
    303 	} */
    304 	file_t *fp;
    305 	int fd;
    306 
    307 	fd = SCARG(uap, fd);
    308 
    309 	if ((fp = fd_getfile(fd)) == NULL)
    310 		return (EBADF);
    311 
    312 	if ((fp->f_flag & FWRITE) == 0) {
    313 		fd_putfile(fd);
    314 		return (EBADF);
    315 	}
    316 
    317 	/* dofilewrite() will unuse the descriptor for us */
    318 	return (dofilewrite(fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
    319 	    &fp->f_offset, FOF_UPDATE_OFFSET, retval));
    320 }
    321 
    322 int
    323 dofilewrite(int fd, struct file *fp, const void *buf,
    324 	size_t nbyte, off_t *offset, int flags, register_t *retval)
    325 {
    326 	struct iovec aiov;
    327 	struct uio auio;
    328 	size_t cnt;
    329 	int error;
    330 
    331 	aiov.iov_base = __UNCONST(buf);		/* XXXUNCONST kills const */
    332 	aiov.iov_len = nbyte;
    333 	auio.uio_iov = &aiov;
    334 	auio.uio_iovcnt = 1;
    335 	auio.uio_resid = nbyte;
    336 	auio.uio_rw = UIO_WRITE;
    337 	auio.uio_vmspace = curproc->p_vmspace;
    338 
    339 	/*
    340 	 * Writes return ssize_t because -1 is returned on error.  Therefore
    341 	 * we must restrict the length to SSIZE_MAX to avoid garbage return
    342 	 * values.
    343 	 */
    344 	if (auio.uio_resid > SSIZE_MAX) {
    345 		error = EINVAL;
    346 		goto out;
    347 	}
    348 
    349 	cnt = auio.uio_resid;
    350 	error = (*fp->f_ops->fo_write)(fp, offset, &auio, fp->f_cred, flags);
    351 	if (error) {
    352 		if (auio.uio_resid != cnt && (error == ERESTART ||
    353 		    error == EINTR || error == EWOULDBLOCK))
    354 			error = 0;
    355 		if (error == EPIPE && !(fp->f_flag & FNOSIGPIPE)) {
    356 			mutex_enter(proc_lock);
    357 			psignal(curproc, SIGPIPE);
    358 			mutex_exit(proc_lock);
    359 		}
    360 	}
    361 	cnt -= auio.uio_resid;
    362 	ktrgenio(fd, UIO_WRITE, buf, cnt, error);
    363 	*retval = cnt;
    364  out:
    365 	fd_putfile(fd);
    366 	return (error);
    367 }
    368 
    369 /*
    370  * Gather write system call
    371  */
    372 int
    373 sys_writev(struct lwp *l, const struct sys_writev_args *uap, register_t *retval)
    374 {
    375 	/* {
    376 		syscallarg(int)				fd;
    377 		syscallarg(const struct iovec *)	iovp;
    378 		syscallarg(int)				iovcnt;
    379 	} */
    380 
    381 	return do_filewritev(SCARG(uap, fd), SCARG(uap, iovp),
    382 	    SCARG(uap, iovcnt), NULL, FOF_UPDATE_OFFSET, retval);
    383 }
    384 
    385 int
    386 do_filewritev(int fd, const struct iovec *iovp, int iovcnt,
    387     off_t *offset, int flags, register_t *retval)
    388 {
    389 	struct uio	auio;
    390 	struct iovec	*iov, *needfree = NULL, aiov[UIO_SMALLIOV];
    391 	int		i, error;
    392 	size_t		cnt;
    393 	u_int		iovlen;
    394 	struct file	*fp;
    395 	struct iovec	*ktriov = NULL;
    396 
    397 	if (iovcnt == 0)
    398 		return EINVAL;
    399 
    400 	if ((fp = fd_getfile(fd)) == NULL)
    401 		return EBADF;
    402 
    403 	if ((fp->f_flag & FWRITE) == 0) {
    404 		fd_putfile(fd);
    405 		return EBADF;
    406 	}
    407 
    408 	if (offset == NULL)
    409 		offset = &fp->f_offset;
    410 	else {
    411 		struct vnode *vp = fp->f_vnode;
    412 		if (fp->f_type != DTYPE_VNODE || vp->v_type == VFIFO) {
    413 			error = ESPIPE;
    414 			goto out;
    415 		}
    416 		/*
    417 		 * Test that the device is seekable ?
    418 		 * XXX This works because no file systems actually
    419 		 * XXX take any action on the seek operation.
    420 		 */
    421 		error = VOP_SEEK(vp, fp->f_offset, *offset, fp->f_cred);
    422 		if (error != 0)
    423 			goto out;
    424 	}
    425 
    426 	iovlen = iovcnt * sizeof(struct iovec);
    427 	if (flags & FOF_IOV_SYSSPACE)
    428 		iov = __UNCONST(iovp);
    429 	else {
    430 		iov = aiov;
    431 		if ((u_int)iovcnt > UIO_SMALLIOV) {
    432 			if ((u_int)iovcnt > IOV_MAX) {
    433 				error = EINVAL;
    434 				goto out;
    435 			}
    436 			iov = kmem_alloc(iovlen, KM_SLEEP);
    437 			needfree = iov;
    438 		}
    439 		error = copyin(iovp, iov, iovlen);
    440 		if (error)
    441 			goto done;
    442 	}
    443 
    444 	auio.uio_iov = iov;
    445 	auio.uio_iovcnt = iovcnt;
    446 	auio.uio_rw = UIO_WRITE;
    447 	auio.uio_vmspace = curproc->p_vmspace;
    448 
    449 	auio.uio_resid = 0;
    450 	for (i = 0; i < iovcnt; i++, iov++) {
    451 		auio.uio_resid += iov->iov_len;
    452 		/*
    453 		 * Writes return ssize_t because -1 is returned on error.
    454 		 * Therefore we must restrict the length to SSIZE_MAX to
    455 		 * avoid garbage return values.
    456 		 */
    457 		if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
    458 			error = EINVAL;
    459 			goto done;
    460 		}
    461 	}
    462 
    463 	/*
    464 	 * if tracing, save a copy of iovec
    465 	 */
    466 	if (ktrpoint(KTR_GENIO))  {
    467 		ktriov = kmem_alloc(iovlen, KM_SLEEP);
    468 		memcpy(ktriov, auio.uio_iov, iovlen);
    469 	}
    470 
    471 	cnt = auio.uio_resid;
    472 	error = (*fp->f_ops->fo_write)(fp, offset, &auio, fp->f_cred, flags);
    473 	if (error) {
    474 		if (auio.uio_resid != cnt && (error == ERESTART ||
    475 		    error == EINTR || error == EWOULDBLOCK))
    476 			error = 0;
    477 		if (error == EPIPE && !(fp->f_flag & FNOSIGPIPE)) {
    478 			mutex_enter(proc_lock);
    479 			psignal(curproc, SIGPIPE);
    480 			mutex_exit(proc_lock);
    481 		}
    482 	}
    483 	cnt -= auio.uio_resid;
    484 	*retval = cnt;
    485 
    486 	if (ktriov != NULL) {
    487 		ktrgeniov(fd, UIO_WRITE, ktriov, cnt, error);
    488 		kmem_free(ktriov, iovlen);
    489 	}
    490 
    491  done:
    492 	if (needfree)
    493 		kmem_free(needfree, iovlen);
    494  out:
    495 	fd_putfile(fd);
    496 	return (error);
    497 }
    498 
    499 /*
    500  * Ioctl system call
    501  */
    502 /* ARGSUSED */
    503 int
    504 sys_ioctl(struct lwp *l, const struct sys_ioctl_args *uap, register_t *retval)
    505 {
    506 	/* {
    507 		syscallarg(int)		fd;
    508 		syscallarg(u_long)	com;
    509 		syscallarg(void *)	data;
    510 	} */
    511 	struct file	*fp;
    512 	proc_t		*p;
    513 	u_long		com;
    514 	int		error;
    515 	size_t		size, alloc_size;
    516 	void 		*data, *memp;
    517 #define	STK_PARAMS	128
    518 	u_long		stkbuf[STK_PARAMS/sizeof(u_long)];
    519 #if  __TMPBIGMAXPARTITIONS > MAXPARTITIONS
    520 	size_t		zero_last = 0;
    521 #define	zero_size(SZ)	((SZ)+zero_last)
    522 #else
    523 #define	zero_size(SZ)	(SZ)
    524 #endif
    525 
    526 	memp = NULL;
    527 	alloc_size = 0;
    528 	error = 0;
    529 	p = l->l_proc;
    530 
    531 	if ((fp = fd_getfile(SCARG(uap, fd))) == NULL)
    532 		return (EBADF);
    533 
    534 	if ((fp->f_flag & (FREAD | FWRITE)) == 0) {
    535 		error = EBADF;
    536 		com = 0;
    537 		goto out;
    538 	}
    539 
    540 	switch (com = SCARG(uap, com)) {
    541 	case FIONCLEX:
    542 	case FIOCLEX:
    543 		fd_set_exclose(l, SCARG(uap, fd), com == FIOCLEX);
    544 		goto out;
    545 	}
    546 
    547 	/*
    548 	 * Interpret high order word to find amount of data to be
    549 	 * copied to/from the user's address space.
    550 	 */
    551 	size = IOCPARM_LEN(com);
    552 	alloc_size = size;
    553 
    554 	/*
    555 	 * The disklabel is now padded to a multiple of 8 bytes however the old
    556 	 * disklabel on 32bit platforms wasn't.  This leaves a difference in
    557 	 * size of 4 bytes between the two but are otherwise identical.
    558 	 * To deal with this, we allocate enough space for the new disklabel
    559 	 * but only copyin/out the smaller amount.
    560 	 */
    561 	if (IOCGROUP(com) == 'd') {
    562 #if  __TMPBIGMAXPARTITIONS > MAXPARTITIONS
    563 		u_long ocom = com;
    564 #endif
    565 		u_long ncom = com ^ (DIOCGDINFO ^ DIOCGDINFO32);
    566 
    567 #if  __TMPBIGMAXPARTITIONS > MAXPARTITIONS
    568 	/*
    569 	 * Userland might use struct disklabel that is bigger than the
    570 	 * the kernel version (historic accident) - alloc userland
    571 	 * size and zero unused part on copyout.
    572 	 */
    573 #define	DISKLABELLENDIFF	(sizeof(struct partition)	\
    574 				       *(__TMPBIGMAXPARTITIONS-MAXPARTITIONS))
    575 #define	IOCFIXUP(NIOC)	((NIOC&~(IOCPARM_MASK<<IOCPARM_SHIFT))	| \
    576 			   (IOCPARM_LEN(NIOC)-DISKLABELLENDIFF)<<IOCPARM_SHIFT)
    577 
    578 		switch (IOCFIXUP(ocom)) {
    579 		case DIOCGDINFO:
    580 		case DIOCWDINFO:
    581 		case DIOCSDINFO:
    582 		case DIOCGDEFLABEL:
    583 			com = ncom = IOCFIXUP(ocom);
    584 			zero_last = DISKLABELLENDIFF;
    585 			size -= DISKLABELLENDIFF;
    586 			goto done;
    587 		}
    588 #endif
    589 
    590 		switch (ncom) {
    591 		case DIOCGDINFO:
    592 		case DIOCWDINFO:
    593 		case DIOCSDINFO:
    594 		case DIOCGDEFLABEL:
    595 			com = ncom;
    596 			if (IOCPARM_LEN(DIOCGDINFO32) < IOCPARM_LEN(DIOCGDINFO))
    597 				alloc_size = IOCPARM_LEN(DIOCGDINFO);
    598 			break;
    599 		}
    600 #if  __TMPBIGMAXPARTITIONS > MAXPARTITIONS
    601 		done: ;
    602 #endif
    603 	}
    604 	if (size > IOCPARM_MAX) {
    605 		error = ENOTTY;
    606 		goto out;
    607 	}
    608 	memp = NULL;
    609 	if ((com >> IOCPARM_SHIFT) == 0)  {
    610 		/* UNIX-style ioctl. */
    611 		data = SCARG(uap, data);
    612 	} else {
    613 		if (alloc_size > sizeof(stkbuf)) {
    614 			memp = kmem_alloc(alloc_size, KM_SLEEP);
    615 			data = memp;
    616 		} else {
    617 			data = (void *)stkbuf;
    618 		}
    619 		if (com&IOC_IN) {
    620 			if (size) {
    621 				error = copyin(SCARG(uap, data), data, size);
    622 				if (error) {
    623 					goto out;
    624 				}
    625 				/*
    626 				 * The data between size and alloc_size has
    627 				 * not been overwritten.  It shouldn't matter
    628 				 * but let's clear that anyway.
    629 				 */
    630 				if (__predict_false(size < alloc_size)) {
    631 					memset((char *)data+size, 0,
    632 					    alloc_size - size);
    633 				}
    634 				ktrgenio(SCARG(uap, fd), UIO_WRITE,
    635 				    SCARG(uap, data), size, 0);
    636 			} else {
    637 				*(void **)data = SCARG(uap, data);
    638 			}
    639 		} else if ((com&IOC_OUT) && size) {
    640 			/*
    641 			 * Zero the buffer so the user always
    642 			 * gets back something deterministic.
    643 			 */
    644 			memset(data, 0, zero_size(size));
    645 		} else if (com&IOC_VOID) {
    646 			*(void **)data = SCARG(uap, data);
    647 		}
    648 	}
    649 
    650 	switch (com) {
    651 
    652 	case FIONBIO:
    653 		/* XXX Code block is not atomic */
    654 		if (*(int *)data != 0)
    655 			atomic_or_uint(&fp->f_flag, FNONBLOCK);
    656 		else
    657 			atomic_and_uint(&fp->f_flag, ~FNONBLOCK);
    658 		error = (*fp->f_ops->fo_ioctl)(fp, FIONBIO, data);
    659 		break;
    660 
    661 	case FIOASYNC:
    662 		/* XXX Code block is not atomic */
    663 		if (*(int *)data != 0)
    664 			atomic_or_uint(&fp->f_flag, FASYNC);
    665 		else
    666 			atomic_and_uint(&fp->f_flag, ~FASYNC);
    667 		error = (*fp->f_ops->fo_ioctl)(fp, FIOASYNC, data);
    668 		break;
    669 
    670 	default:
    671 		error = (*fp->f_ops->fo_ioctl)(fp, com, data);
    672 		/*
    673 		 * Copy any data to user, size was
    674 		 * already set and checked above.
    675 		 */
    676 		if (error == 0 && (com&IOC_OUT) && size) {
    677 			error = copyout(data, SCARG(uap, data),
    678 			    zero_size(size));
    679 			ktrgenio(SCARG(uap, fd), UIO_READ, SCARG(uap, data),
    680 			    size, error);
    681 		}
    682 		break;
    683 	}
    684  out:
    685 	if (memp)
    686 		kmem_free(memp, alloc_size);
    687 	fd_putfile(SCARG(uap, fd));
    688 	switch (error) {
    689 	case -1:
    690 		printf("sys_ioctl: _IO%s%s('%c', %lu, %lu) returned -1: "
    691 		    "pid=%d comm=%s\n",
    692 		    (com & IOC_IN) ? "W" : "", (com & IOC_OUT) ? "R" : "",
    693 		    (char)IOCGROUP(com), (com & 0xff), IOCPARM_LEN(com),
    694 		    p->p_pid, p->p_comm);
    695 		/* FALLTHROUGH */
    696 	case EPASSTHROUGH:
    697 		error = ENOTTY;
    698 		/* FALLTHROUGH */
    699 	default:
    700 		return (error);
    701 	}
    702 }
    703