Home | History | Annotate | Line # | Download | only in kern
uipc_syscalls.c revision 1.57.2.4
      1 /*	$NetBSD: uipc_syscalls.c,v 1.57.2.4 2001/09/21 22:36:27 nathanw Exp $	*/
      2 
      3 /*
      4  * Copyright (c) 1982, 1986, 1989, 1990, 1993
      5  *	The Regents of the University of California.  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  * 3. All advertising materials mentioning features or use of this software
     16  *    must display the following acknowledgement:
     17  *	This product includes software developed by the University of
     18  *	California, Berkeley and its contributors.
     19  * 4. Neither the name of the University nor the names of its contributors
     20  *    may be used to endorse or promote products derived from this software
     21  *    without specific prior written permission.
     22  *
     23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     33  * SUCH DAMAGE.
     34  *
     35  *	@(#)uipc_syscalls.c	8.6 (Berkeley) 2/14/95
     36  */
     37 
     38 #include "opt_ktrace.h"
     39 #include "opt_new_pipe.h"
     40 
     41 #include <sys/param.h>
     42 #include <sys/systm.h>
     43 #include <sys/filedesc.h>
     44 #include <sys/lwp.h>
     45 #include <sys/proc.h>
     46 #include <sys/file.h>
     47 #include <sys/buf.h>
     48 #include <sys/malloc.h>
     49 #include <sys/mbuf.h>
     50 #include <sys/protosw.h>
     51 #include <sys/socket.h>
     52 #include <sys/socketvar.h>
     53 #include <sys/signalvar.h>
     54 #include <sys/un.h>
     55 #ifdef KTRACE
     56 #include <sys/ktrace.h>
     57 #endif
     58 
     59 #include <sys/mount.h>
     60 #include <sys/syscallargs.h>
     61 
     62 #include <uvm/uvm_extern.h>
     63 
     64 /*
     65  * System call interface to the socket abstraction.
     66  */
     67 extern	struct fileops socketops;
     68 
     69 int
     70 sys_socket(struct lwp *l, void *v, register_t *retval)
     71 {
     72 	struct sys_socket_args /* {
     73 		syscallarg(int)	domain;
     74 		syscallarg(int)	type;
     75 		syscallarg(int)	protocol;
     76 	} */ *uap = v;
     77 
     78 	struct proc	*p;
     79 	struct filedesc	*fdp;
     80 	struct socket	*so;
     81 	struct file	*fp;
     82 	int		fd, error;
     83 
     84 	p = l->l_proc;
     85 	fdp = p->p_fd;
     86 	/* falloc() will use the desciptor for us */
     87 	if ((error = falloc(p, &fp, &fd)) != 0)
     88 		return (error);
     89 	fp->f_flag = FREAD|FWRITE;
     90 	fp->f_type = DTYPE_SOCKET;
     91 	fp->f_ops = &socketops;
     92 	error = socreate(SCARG(uap, domain), &so, SCARG(uap, type),
     93 			 SCARG(uap, protocol));
     94 	if (error) {
     95 		FILE_UNUSE(fp, p);
     96 		fdremove(fdp, fd);
     97 		ffree(fp);
     98 	} else {
     99 		fp->f_data = (caddr_t)so;
    100 		FILE_SET_MATURE(fp);
    101 		FILE_UNUSE(fp, p);
    102 		*retval = fd;
    103 	}
    104 	return (error);
    105 }
    106 
    107 /* ARGSUSED */
    108 int
    109 sys_bind(struct lwp *l, void *v, register_t *retval)
    110 {
    111 	struct sys_bind_args /* {
    112 		syscallarg(int)				s;
    113 		syscallarg(const struct sockaddr *)	name;
    114 		syscallarg(unsigned int)		namelen;
    115 	} */ *uap = v;
    116 	struct proc	*p;
    117 	struct file	*fp;
    118 	struct mbuf	*nam;
    119 	int		error;
    120 
    121 	p = l->l_proc;
    122 	/* getsock() will use the descriptor for us */
    123 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    124 		return (error);
    125 	error = sockargs(&nam, SCARG(uap, name), SCARG(uap, namelen),
    126 	    MT_SONAME);
    127 	if (error) {
    128 		FILE_UNUSE(fp, p);
    129 		return (error);
    130 	}
    131 	error = sobind((struct socket *)fp->f_data, nam, p);
    132 	m_freem(nam);
    133 	FILE_UNUSE(fp, p);
    134 	return (error);
    135 }
    136 
    137 /* ARGSUSED */
    138 int
    139 sys_listen(struct lwp *l, void *v, register_t *retval)
    140 {
    141 	struct sys_listen_args /* {
    142 		syscallarg(int)	s;
    143 		syscallarg(int)	backlog;
    144 	} */ *uap = v;
    145 	struct proc	*p;
    146 	struct file	*fp;
    147 	int		error;
    148 
    149 	p = l->l_proc;
    150 	/* getsock() will use the descriptor for us */
    151 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    152 		return (error);
    153 	error = solisten((struct socket *)fp->f_data, SCARG(uap, backlog));
    154 	FILE_UNUSE(fp, p);
    155 	return (error);
    156 }
    157 
    158 int
    159 sys_accept(struct lwp *l, void *v, register_t *retval)
    160 {
    161 	struct sys_accept_args /* {
    162 		syscallarg(int)			s;
    163 		syscallarg(struct sockaddr *)	name;
    164 		syscallarg(unsigned int *)	anamelen;
    165 	} */ *uap = v;
    166 	struct proc	*p;
    167 	struct filedesc	*fdp;
    168 	struct file	*fp;
    169 	struct mbuf	*nam;
    170 	unsigned int	namelen;
    171 	int		error, s, fd;
    172 	struct socket	*so;
    173 
    174 	p = l->l_proc;
    175 	fdp = p->p_fd;
    176 	if (SCARG(uap, name) && (error = copyin((caddr_t)SCARG(uap, anamelen),
    177 	    (caddr_t)&namelen, sizeof(namelen))))
    178 		return (error);
    179 	if (SCARG(uap, name) != NULL &&
    180 	    uvm_useracc((caddr_t)SCARG(uap, name), sizeof(struct sockaddr),
    181 	     B_WRITE) == FALSE)
    182 		return (EFAULT);
    183 
    184 	/* getsock() will use the descriptor for us */
    185 	if ((error = getsock(fdp, SCARG(uap, s), &fp)) != 0)
    186 		return (error);
    187 	s = splsoftnet();
    188 	so = (struct socket *)fp->f_data;
    189 	FILE_UNUSE(fp, p);
    190 	if (!(so->so_proto->pr_flags & PR_LISTEN)) {
    191 		splx(s);
    192 		return (EOPNOTSUPP);
    193 	}
    194 	if ((so->so_options & SO_ACCEPTCONN) == 0) {
    195 		splx(s);
    196 		return (EINVAL);
    197 	}
    198 	if ((so->so_state & SS_NBIO) && so->so_qlen == 0) {
    199 		splx(s);
    200 		return (EWOULDBLOCK);
    201 	}
    202 	while (so->so_qlen == 0 && so->so_error == 0) {
    203 		if (so->so_state & SS_CANTRCVMORE) {
    204 			so->so_error = ECONNABORTED;
    205 			break;
    206 		}
    207 		error = tsleep((caddr_t)&so->so_timeo, PSOCK | PCATCH,
    208 			       netcon, 0);
    209 		if (error) {
    210 			splx(s);
    211 			return (error);
    212 		}
    213 	}
    214 	if (so->so_error) {
    215 		error = so->so_error;
    216 		so->so_error = 0;
    217 		splx(s);
    218 		return (error);
    219 	}
    220 	/* falloc() will use the descriptor for us */
    221 	if ((error = falloc(p, &fp, &fd)) != 0) {
    222 		splx(s);
    223 		return (error);
    224 	}
    225 	*retval = fd;
    226 	{ struct socket *aso = so->so_q.tqh_first;
    227 	  if (soqremque(aso, 1) == 0)
    228 		panic("accept");
    229 	  so = aso;
    230 	}
    231 	fp->f_type = DTYPE_SOCKET;
    232 	fp->f_flag = FREAD|FWRITE;
    233 	fp->f_ops = &socketops;
    234 	fp->f_data = (caddr_t)so;
    235 	FILE_UNUSE(fp, p);
    236 	nam = m_get(M_WAIT, MT_SONAME);
    237 	if ((error = soaccept(so, nam)) == 0 && SCARG(uap, name)) {
    238 		if (namelen > nam->m_len)
    239 			namelen = nam->m_len;
    240 		/* SHOULD COPY OUT A CHAIN HERE */
    241 		if ((error = copyout(mtod(nam, caddr_t),
    242 		    (caddr_t)SCARG(uap, name), namelen)) == 0)
    243 			error = copyout((caddr_t)&namelen,
    244 			    (caddr_t)SCARG(uap, anamelen),
    245 			    sizeof(*SCARG(uap, anamelen)));
    246 	}
    247 	/* if an error occurred, free the file descriptor */
    248 	if (error) {
    249 		fdremove(fdp, fd);
    250 		ffree(fp);
    251 	}
    252 	m_freem(nam);
    253 	splx(s);
    254 	FILE_SET_MATURE(fp);
    255 	return (error);
    256 }
    257 
    258 /* ARGSUSED */
    259 int
    260 sys_connect(struct lwp *l, void *v, register_t *retval)
    261 {
    262 	struct sys_connect_args /* {
    263 		syscallarg(int)				s;
    264 		syscallarg(const struct sockaddr *)	name;
    265 		syscallarg(unsigned int)		namelen;
    266 	} */ *uap = v;
    267 	struct proc	*p;
    268 	struct file	*fp;
    269 	struct socket	*so;
    270 	struct mbuf	*nam;
    271 	int		error, s;
    272 
    273 	p = l->l_proc;
    274 	/* getsock() will use the descriptor for us */
    275 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    276 		return (error);
    277 	so = (struct socket *)fp->f_data;
    278 	if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING)) {
    279 		error = EALREADY;
    280 		goto out;
    281 	}
    282 	error = sockargs(&nam, SCARG(uap, name), SCARG(uap, namelen),
    283 	    MT_SONAME);
    284 	if (error)
    285 		goto out;
    286 	error = soconnect(so, nam);
    287 	if (error)
    288 		goto bad;
    289 	if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING)) {
    290 		m_freem(nam);
    291 		error = EINPROGRESS;
    292 		goto out;
    293 	}
    294 	s = splsoftnet();
    295 	while ((so->so_state & SS_ISCONNECTING) && so->so_error == 0) {
    296 		error = tsleep((caddr_t)&so->so_timeo, PSOCK | PCATCH,
    297 			       netcon, 0);
    298 		if (error)
    299 			break;
    300 	}
    301 	if (error == 0) {
    302 		error = so->so_error;
    303 		so->so_error = 0;
    304 	}
    305 	splx(s);
    306  bad:
    307 	so->so_state &= ~SS_ISCONNECTING;
    308 	m_freem(nam);
    309 	if (error == ERESTART)
    310 		error = EINTR;
    311  out:
    312 	FILE_UNUSE(fp, p);
    313 	return (error);
    314 }
    315 
    316 int
    317 sys_socketpair(struct lwp *l, void *v, register_t *retval)
    318 {
    319 	struct sys_socketpair_args /* {
    320 		syscallarg(int)		domain;
    321 		syscallarg(int)		type;
    322 		syscallarg(int)		protocol;
    323 		syscallarg(int *)	rsv;
    324 	} */ *uap = v;
    325 	struct proc *p;
    326 	struct filedesc	*fdp;
    327 	struct file	*fp1, *fp2;
    328 	struct socket	*so1, *so2;
    329 	int		fd, error, sv[2];
    330 
    331 	p = l->l_proc;
    332 	fdp = p->p_fd;
    333 	error = socreate(SCARG(uap, domain), &so1, SCARG(uap, type),
    334 			 SCARG(uap, protocol));
    335 	if (error)
    336 		return (error);
    337 	error = socreate(SCARG(uap, domain), &so2, SCARG(uap, type),
    338 			 SCARG(uap, protocol));
    339 	if (error)
    340 		goto free1;
    341 	/* falloc() will use the descriptor for us */
    342 	if ((error = falloc(p, &fp1, &fd)) != 0)
    343 		goto free2;
    344 	sv[0] = fd;
    345 	fp1->f_flag = FREAD|FWRITE;
    346 	fp1->f_type = DTYPE_SOCKET;
    347 	fp1->f_ops = &socketops;
    348 	fp1->f_data = (caddr_t)so1;
    349 	if ((error = falloc(p, &fp2, &fd)) != 0)
    350 		goto free3;
    351 	fp2->f_flag = FREAD|FWRITE;
    352 	fp2->f_type = DTYPE_SOCKET;
    353 	fp2->f_ops = &socketops;
    354 	fp2->f_data = (caddr_t)so2;
    355 	sv[1] = fd;
    356 	if ((error = soconnect2(so1, so2)) != 0)
    357 		goto free4;
    358 	if (SCARG(uap, type) == SOCK_DGRAM) {
    359 		/*
    360 		 * Datagram socket connection is asymmetric.
    361 		 */
    362 		 if ((error = soconnect2(so2, so1)) != 0)
    363 			goto free4;
    364 	}
    365 	error = copyout((caddr_t)sv, (caddr_t)SCARG(uap, rsv),
    366 	    2 * sizeof(int));
    367 	FILE_SET_MATURE(fp1);
    368 	FILE_SET_MATURE(fp2);
    369 	FILE_UNUSE(fp1, p);
    370 	FILE_UNUSE(fp2, p);
    371 	return (error);
    372  free4:
    373 	FILE_UNUSE(fp2, p);
    374 	ffree(fp2);
    375 	fdremove(fdp, sv[1]);
    376  free3:
    377 	FILE_UNUSE(fp1, p);
    378 	ffree(fp1);
    379 	fdremove(fdp, sv[0]);
    380  free2:
    381 	(void)soclose(so2);
    382  free1:
    383 	(void)soclose(so1);
    384 	return (error);
    385 }
    386 
    387 int
    388 sys_sendto(struct lwp *l, void *v, register_t *retval)
    389 {
    390 	struct sys_sendto_args /* {
    391 		syscallarg(int)				s;
    392 		syscallarg(const void *)		buf;
    393 		syscallarg(size_t)			len;
    394 		syscallarg(int)				flags;
    395 		syscallarg(const struct sockaddr *)	to;
    396 		syscallarg(unsigned int)		tolen;
    397 	} */ *uap = v;
    398 	struct proc	*p;
    399 	struct msghdr	msg;
    400 	struct iovec	aiov;
    401 
    402 	p = l->l_proc;
    403 	msg.msg_name = (caddr_t)SCARG(uap, to);		/* XXX kills const */
    404 	msg.msg_namelen = SCARG(uap, tolen);
    405 	msg.msg_iov = &aiov;
    406 	msg.msg_iovlen = 1;
    407 	msg.msg_control = 0;
    408 	msg.msg_flags = 0;
    409 	aiov.iov_base = (char *)SCARG(uap, buf);	/* XXX kills const */
    410 	aiov.iov_len = SCARG(uap, len);
    411 	return (sendit(p, SCARG(uap, s), &msg, SCARG(uap, flags), retval));
    412 }
    413 
    414 int
    415 sys_sendmsg(struct lwp *l, void *v, register_t *retval)
    416 {
    417 	struct sys_sendmsg_args /* {
    418 		syscallarg(int)				s;
    419 		syscallarg(const struct msghdr *)	msg;
    420 		syscallarg(int)				flags;
    421 	} */ *uap = v;
    422 	struct proc	*p;
    423 	struct msghdr	msg;
    424 	struct iovec	aiov[UIO_SMALLIOV], *iov;
    425 	int		error;
    426 
    427 	error = copyin(SCARG(uap, msg), (caddr_t)&msg, sizeof(msg));
    428 	if (error)
    429 		return (error);
    430 	if ((unsigned int)msg.msg_iovlen > UIO_SMALLIOV) {
    431 		if ((unsigned int)msg.msg_iovlen > IOV_MAX)
    432 			return (EMSGSIZE);
    433 		iov = malloc(sizeof(struct iovec) * msg.msg_iovlen,
    434 		    M_IOV, M_WAITOK);
    435 	} else
    436 		iov = aiov;
    437 	if ((unsigned int)msg.msg_iovlen > 0) {
    438 		error = copyin((caddr_t)msg.msg_iov, (caddr_t)iov,
    439 		    (size_t)(msg.msg_iovlen * sizeof(struct iovec)));
    440 		if (error)
    441 			goto done;
    442 	}
    443 	msg.msg_iov = iov;
    444 	msg.msg_flags = 0;
    445 	p = l->l_proc;
    446 	error = sendit(p, SCARG(uap, s), &msg, SCARG(uap, flags), retval);
    447 done:
    448 	if (iov != aiov)
    449 		free(iov, M_IOV);
    450 	return (error);
    451 }
    452 
    453 int
    454 sendit(struct proc *p, int s, struct msghdr *mp, int flags, register_t *retsize)
    455 {
    456 	struct file	*fp;
    457 	struct uio	auio;
    458 	struct iovec	*iov;
    459 	int		i, len, error;
    460 	struct mbuf	*to, *control;
    461 	struct socket	*so;
    462 #ifdef KTRACE
    463 	struct iovec	*ktriov;
    464 #endif
    465 
    466 #ifdef KTRACE
    467 	ktriov = NULL;
    468 #endif
    469 	/* getsock() will use the descriptor for us */
    470 	if ((error = getsock(p->p_fd, s, &fp)) != 0)
    471 		return (error);
    472 	auio.uio_iov = mp->msg_iov;
    473 	auio.uio_iovcnt = mp->msg_iovlen;
    474 	auio.uio_segflg = UIO_USERSPACE;
    475 	auio.uio_rw = UIO_WRITE;
    476 	auio.uio_procp = p;
    477 	auio.uio_offset = 0;			/* XXX */
    478 	auio.uio_resid = 0;
    479 	iov = mp->msg_iov;
    480 	for (i = 0; i < mp->msg_iovlen; i++, iov++) {
    481 #if 0
    482 		/* cannot happen; iov_len is unsigned */
    483 		if (iov->iov_len < 0) {
    484 			error = EINVAL;
    485 			goto out;
    486 		}
    487 #endif
    488 		/*
    489 		 * Writes return ssize_t because -1 is returned on error.
    490 		 * Therefore, we must restrict the length to SSIZE_MAX to
    491 		 * avoid garbage return values.
    492 		 */
    493 		auio.uio_resid += iov->iov_len;
    494 		if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
    495 			error = EINVAL;
    496 			goto out;
    497 		}
    498 	}
    499 	if (mp->msg_name) {
    500 		error = sockargs(&to, mp->msg_name, mp->msg_namelen,
    501 				 MT_SONAME);
    502 		if (error)
    503 			goto out;
    504 	} else
    505 		to = 0;
    506 	if (mp->msg_control) {
    507 		if (mp->msg_controllen < sizeof(struct cmsghdr)) {
    508 			error = EINVAL;
    509 			goto bad;
    510 		}
    511 		error = sockargs(&control, mp->msg_control,
    512 				 mp->msg_controllen, MT_CONTROL);
    513 		if (error)
    514 			goto bad;
    515 	} else
    516 		control = 0;
    517 #ifdef KTRACE
    518 	if (KTRPOINT(p, KTR_GENIO)) {
    519 		int iovlen = auio.uio_iovcnt * sizeof(struct iovec);
    520 
    521 		ktriov = malloc(iovlen, M_TEMP, M_WAITOK);
    522 		memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen);
    523 	}
    524 #endif
    525 	len = auio.uio_resid;
    526 	so = (struct socket *)fp->f_data;
    527 	error = (*so->so_send)(so, to, &auio, NULL, control, flags);
    528 	if (error) {
    529 		if (auio.uio_resid != len && (error == ERESTART ||
    530 		    error == EINTR || error == EWOULDBLOCK))
    531 			error = 0;
    532 		if (error == EPIPE)
    533 			psignal(p, SIGPIPE);
    534 	}
    535 	if (error == 0)
    536 		*retsize = len - auio.uio_resid;
    537 #ifdef KTRACE
    538 	if (ktriov != NULL) {
    539 		if (error == 0)
    540 			ktrgenio(p, s, UIO_WRITE, ktriov, *retsize, error);
    541 		free(ktriov, M_TEMP);
    542 	}
    543 #endif
    544  bad:
    545 	if (to)
    546 		m_freem(to);
    547  out:
    548 	FILE_UNUSE(fp, p);
    549 	return (error);
    550 }
    551 
    552 int
    553 sys_recvfrom(struct lwp *l, void *v, register_t *retval)
    554 {
    555 	struct sys_recvfrom_args /* {
    556 		syscallarg(int)			s;
    557 		syscallarg(void *)		buf;
    558 		syscallarg(size_t)		len;
    559 		syscallarg(int)			flags;
    560 		syscallarg(struct sockaddr *)	from;
    561 		syscallarg(unsigned int *)	fromlenaddr;
    562 	} */ *uap = v;
    563 	struct proc	*p;
    564 	struct msghdr	msg;
    565 	struct iovec	aiov;
    566 	int		error;
    567 
    568 	if (SCARG(uap, fromlenaddr)) {
    569 		error = copyin((caddr_t)SCARG(uap, fromlenaddr),
    570 			       (caddr_t)&msg.msg_namelen,
    571 			       sizeof(msg.msg_namelen));
    572 		if (error)
    573 			return (error);
    574 	} else
    575 		msg.msg_namelen = 0;
    576 	msg.msg_name = (caddr_t)SCARG(uap, from);
    577 	msg.msg_iov = &aiov;
    578 	msg.msg_iovlen = 1;
    579 	aiov.iov_base = SCARG(uap, buf);
    580 	aiov.iov_len = SCARG(uap, len);
    581 	msg.msg_control = 0;
    582 	msg.msg_flags = SCARG(uap, flags);
    583 	p = l->l_proc;
    584 	return (recvit(p, SCARG(uap, s), &msg,
    585 		       (caddr_t)SCARG(uap, fromlenaddr), retval));
    586 }
    587 
    588 int
    589 sys_recvmsg(struct lwp *l, void *v, register_t *retval)
    590 {
    591 	struct sys_recvmsg_args /* {
    592 		syscallarg(int)			s;
    593 		syscallarg(struct msghdr *)	msg;
    594 		syscallarg(int)			flags;
    595 	} */ *uap = v;
    596 	struct proc	*p;
    597 	struct msghdr	msg;
    598 	struct iovec	aiov[UIO_SMALLIOV], *uiov, *iov;
    599 	int		error;
    600 
    601 	error = copyin((caddr_t)SCARG(uap, msg), (caddr_t)&msg,
    602 		       sizeof(msg));
    603 	if (error)
    604 		return (error);
    605 	if ((unsigned int)msg.msg_iovlen > UIO_SMALLIOV) {
    606 		if ((unsigned int)msg.msg_iovlen > IOV_MAX)
    607 			return (EMSGSIZE);
    608 		iov = malloc(sizeof(struct iovec) * msg.msg_iovlen,
    609 		    M_IOV, M_WAITOK);
    610 	} else
    611 		iov = aiov;
    612 	if ((unsigned int)msg.msg_iovlen > 0) {
    613 		error = copyin((caddr_t)msg.msg_iov, (caddr_t)iov,
    614 		    (size_t)(msg.msg_iovlen * sizeof(struct iovec)));
    615 		if (error)
    616 			goto done;
    617 	}
    618 	uiov = msg.msg_iov;
    619 	msg.msg_iov = iov;
    620 	msg.msg_flags = SCARG(uap, flags);
    621 	p = l->l_proc;
    622 	if ((error = recvit(p, SCARG(uap, s), &msg, (caddr_t)0, retval)) == 0) {
    623 		msg.msg_iov = uiov;
    624 		error = copyout((caddr_t)&msg, (caddr_t)SCARG(uap, msg),
    625 		    sizeof(msg));
    626 	}
    627 done:
    628 	if (iov != aiov)
    629 		free(iov, M_IOV);
    630 	return (error);
    631 }
    632 
    633 int
    634 recvit(struct proc *p, int s, struct msghdr *mp, caddr_t namelenp,
    635 	register_t *retsize)
    636 {
    637 	struct file	*fp;
    638 	struct uio	auio;
    639 	struct iovec	*iov;
    640 	int		i, len, error;
    641 	struct mbuf	*from, *control;
    642 	struct socket	*so;
    643 #ifdef KTRACE
    644 	struct iovec	*ktriov;
    645 #endif
    646 
    647 	from = 0;
    648 	control = 0;
    649 #ifdef KTRACE
    650 	ktriov = NULL;
    651 #endif
    652 
    653 	/* getsock() will use the descriptor for us */
    654 	if ((error = getsock(p->p_fd, s, &fp)) != 0)
    655 		return (error);
    656 	auio.uio_iov = mp->msg_iov;
    657 	auio.uio_iovcnt = mp->msg_iovlen;
    658 	auio.uio_segflg = UIO_USERSPACE;
    659 	auio.uio_rw = UIO_READ;
    660 	auio.uio_procp = p;
    661 	auio.uio_offset = 0;			/* XXX */
    662 	auio.uio_resid = 0;
    663 	iov = mp->msg_iov;
    664 	for (i = 0; i < mp->msg_iovlen; i++, iov++) {
    665 #if 0
    666 		/* cannot happen iov_len is unsigned */
    667 		if (iov->iov_len < 0) {
    668 			error = EINVAL;
    669 			goto out1;
    670 		}
    671 #endif
    672 		/*
    673 		 * Reads return ssize_t because -1 is returned on error.
    674 		 * Therefore we must restrict the length to SSIZE_MAX to
    675 		 * avoid garbage return values.
    676 		 */
    677 		auio.uio_resid += iov->iov_len;
    678 		if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
    679 			error = EINVAL;
    680 			goto out1;
    681 		}
    682 	}
    683 #ifdef KTRACE
    684 	if (KTRPOINT(p, KTR_GENIO)) {
    685 		int iovlen = auio.uio_iovcnt * sizeof(struct iovec);
    686 
    687 		ktriov = malloc(iovlen, M_TEMP, M_WAITOK);
    688 		memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen);
    689 	}
    690 #endif
    691 	len = auio.uio_resid;
    692 	so = (struct socket *)fp->f_data;
    693 	error = (*so->so_receive)(so, &from, &auio, NULL,
    694 			  mp->msg_control ? &control : NULL, &mp->msg_flags);
    695 	if (error) {
    696 		if (auio.uio_resid != len && (error == ERESTART ||
    697 		    error == EINTR || error == EWOULDBLOCK))
    698 			error = 0;
    699 	}
    700 #ifdef KTRACE
    701 	if (ktriov != NULL) {
    702 		if (error == 0)
    703 			ktrgenio(p, s, UIO_READ, ktriov,
    704 			    len - auio.uio_resid, error);
    705 		free(ktriov, M_TEMP);
    706 	}
    707 #endif
    708 	if (error)
    709 		goto out;
    710 	*retsize = len - auio.uio_resid;
    711 	if (mp->msg_name) {
    712 		len = mp->msg_namelen;
    713 		if (len <= 0 || from == 0)
    714 			len = 0;
    715 		else {
    716 			if (len > from->m_len)
    717 				len = from->m_len;
    718 			/* else if len < from->m_len ??? */
    719 			error = copyout(mtod(from, caddr_t),
    720 					(caddr_t)mp->msg_name, (unsigned)len);
    721 			if (error)
    722 				goto out;
    723 		}
    724 		mp->msg_namelen = len;
    725 		if (namelenp &&
    726 		    (error = copyout((caddr_t)&len, namelenp, sizeof(int))))
    727 			goto out;
    728 	}
    729 	if (mp->msg_control) {
    730 		len = mp->msg_controllen;
    731 		if (len <= 0 || control == 0)
    732 			len = 0;
    733 		else {
    734 			struct mbuf *m = control;
    735 			caddr_t p = (caddr_t)mp->msg_control;
    736 
    737 			do {
    738 				i = m->m_len;
    739 				if (len < i) {
    740 					mp->msg_flags |= MSG_CTRUNC;
    741 					i = len;
    742 				}
    743 				error = copyout(mtod(m, caddr_t), p,
    744 				    (unsigned)i);
    745 				if (m->m_next)
    746 					i = ALIGN(i);
    747 				p += i;
    748 				len -= i;
    749 				if (error != 0 || len <= 0)
    750 					break;
    751 			} while ((m = m->m_next) != NULL);
    752 			len = p - (caddr_t)mp->msg_control;
    753 		}
    754 		mp->msg_controllen = len;
    755 	}
    756  out:
    757 	if (from)
    758 		m_freem(from);
    759 	if (control)
    760 		m_freem(control);
    761  out1:
    762 	FILE_UNUSE(fp, p);
    763 	return (error);
    764 }
    765 
    766 /* ARGSUSED */
    767 int
    768 sys_shutdown(struct lwp *l, void *v, register_t *retval)
    769 {
    770 	struct sys_shutdown_args /* {
    771 		syscallarg(int)	s;
    772 		syscallarg(int)	how;
    773 	} */ *uap = v;
    774 	struct proc	*p;
    775 	struct file	*fp;
    776 	int		error;
    777 
    778 	p = l->l_proc;
    779 	/* getsock() will use the descriptor for us */
    780 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    781 		return (error);
    782 	error = soshutdown((struct socket *)fp->f_data, SCARG(uap, how));
    783 	FILE_UNUSE(fp, p);
    784 	return (error);
    785 }
    786 
    787 /* ARGSUSED */
    788 int
    789 sys_setsockopt(struct lwp *l, void *v, register_t *retval)
    790 {
    791 	struct sys_setsockopt_args /* {
    792 		syscallarg(int)			s;
    793 		syscallarg(int)			level;
    794 		syscallarg(int)			name;
    795 		syscallarg(const void *)	val;
    796 		syscallarg(unsigned int)	valsize;
    797 	} */ *uap = v;
    798 	struct proc	*p;
    799 	struct file	*fp;
    800 	struct mbuf	*m;
    801 	int		error;
    802 
    803 	p = l->l_proc;
    804 	m = NULL;
    805 	/* getsock() will use the descriptor for us */
    806 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    807 		return (error);
    808 	if (SCARG(uap, valsize) > MLEN) {
    809 		error = EINVAL;
    810 		goto out;
    811 	}
    812 	if (SCARG(uap, val)) {
    813 		m = m_get(M_WAIT, MT_SOOPTS);
    814 		error = copyin(SCARG(uap, val), mtod(m, caddr_t),
    815 			       SCARG(uap, valsize));
    816 		if (error) {
    817 			(void) m_free(m);
    818 			goto out;
    819 		}
    820 		m->m_len = SCARG(uap, valsize);
    821 	}
    822 	error = sosetopt((struct socket *)fp->f_data, SCARG(uap, level),
    823 			 SCARG(uap, name), m);
    824  out:
    825 	FILE_UNUSE(fp, p);
    826 	return (error);
    827 }
    828 
    829 /* ARGSUSED */
    830 int
    831 sys_getsockopt(struct lwp *l, void *v, register_t *retval)
    832 {
    833 	struct sys_getsockopt_args /* {
    834 		syscallarg(int)			s;
    835 		syscallarg(int)			level;
    836 		syscallarg(int)			name;
    837 		syscallarg(void *)		val;
    838 		syscallarg(unsigned int *)	avalsize;
    839 	} */ *uap = v;
    840 	struct proc	*p;
    841 	struct file	*fp;
    842 	struct mbuf	*m, *m0;
    843 	unsigned int	op, i, valsize;
    844 	int		error;
    845 
    846 	p = l->l_proc;
    847 	m = NULL;
    848 	/* getsock() will use the descriptor for us */
    849 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    850 		return (error);
    851 	if (SCARG(uap, val)) {
    852 		error = copyin((caddr_t)SCARG(uap, avalsize),
    853 			       (caddr_t)&valsize, sizeof(valsize));
    854 		if (error)
    855 			goto out;
    856 	} else
    857 		valsize = 0;
    858 	if ((error = sogetopt((struct socket *)fp->f_data, SCARG(uap, level),
    859 	    SCARG(uap, name), &m)) == 0 && SCARG(uap, val) && valsize &&
    860 	    m != NULL) {
    861 		op = 0;
    862 		while (m && !error && op < valsize) {
    863 			i = min(m->m_len, (valsize - op));
    864 			error = copyout(mtod(m, caddr_t), SCARG(uap, val), i);
    865 			op += i;
    866 			SCARG(uap, val) = ((u_int8_t *)SCARG(uap, val)) + i;
    867 			m0 = m;
    868 			MFREE(m0, m);
    869 		}
    870 		valsize = op;
    871 		if (error == 0)
    872 			error = copyout(&valsize,
    873 					SCARG(uap, avalsize), sizeof(valsize));
    874 	}
    875 	if (m != NULL)
    876 		(void) m_free(m);
    877  out:
    878 	FILE_UNUSE(fp, p);
    879 	return (error);
    880 }
    881 
    882 #ifndef NEW_PIPE
    883 /* ARGSUSED */
    884 int
    885 sys_pipe(struct lwp *l, void *v, register_t *retval)
    886 {
    887 	struct proc	*p;
    888 	struct filedesc	*fdp;
    889 	struct file	*rf, *wf;
    890 	struct socket	*rso, *wso;
    891 	int		fd, error;
    892 
    893 	p = l->l_proc;
    894 	fdp = p->p_fd;
    895 	if ((error = socreate(AF_LOCAL, &rso, SOCK_STREAM, 0)) != 0)
    896 		return (error);
    897 	if ((error = socreate(AF_LOCAL, &wso, SOCK_STREAM, 0)) != 0)
    898 		goto free1;
    899 	/* remember this socket pair implements a pipe */
    900 	wso->so_state |= SS_ISAPIPE;
    901 	rso->so_state |= SS_ISAPIPE;
    902 	/* falloc() will use the descriptor for us */
    903 	if ((error = falloc(p, &rf, &fd)) != 0)
    904 		goto free2;
    905 	retval[0] = fd;
    906 	rf->f_flag = FREAD;
    907 	rf->f_type = DTYPE_SOCKET;
    908 	rf->f_ops = &socketops;
    909 	rf->f_data = (caddr_t)rso;
    910 	if ((error = falloc(p, &wf, &fd)) != 0)
    911 		goto free3;
    912 	wf->f_flag = FWRITE;
    913 	wf->f_type = DTYPE_SOCKET;
    914 	wf->f_ops = &socketops;
    915 	wf->f_data = (caddr_t)wso;
    916 	retval[1] = fd;
    917 	if ((error = unp_connect2(wso, rso)) != 0)
    918 		goto free4;
    919 	FILE_SET_MATURE(rf);
    920 	FILE_SET_MATURE(wf);
    921 	FILE_UNUSE(rf, p);
    922 	FILE_UNUSE(wf, p);
    923 	return (0);
    924  free4:
    925 	FILE_UNUSE(wf, p);
    926 	ffree(wf);
    927 	fdremove(fdp, retval[1]);
    928  free3:
    929 	FILE_UNUSE(rf, p);
    930 	ffree(rf);
    931 	fdremove(fdp, retval[0]);
    932  free2:
    933 	(void)soclose(wso);
    934  free1:
    935 	(void)soclose(rso);
    936 	return (error);
    937 }
    938 #endif /* !NEW_PIPE */
    939 
    940 /*
    941  * Get socket name.
    942  */
    943 /* ARGSUSED */
    944 int
    945 sys_getsockname(struct lwp *l, void *v, register_t *retval)
    946 {
    947 	struct sys_getsockname_args /* {
    948 		syscallarg(int)			fdes;
    949 		syscallarg(struct sockaddr *)	asa;
    950 		syscallarg(unsigned int *)	alen;
    951 	} */ *uap = v;
    952 	struct proc	*p;
    953 	struct file	*fp;
    954 	struct socket	*so;
    955 	struct mbuf	*m;
    956 	unsigned int	len;
    957 	int		error;
    958 
    959 	p = l->l_proc;
    960 	/* getsock() will use the descriptor for us */
    961 	if ((error = getsock(p->p_fd, SCARG(uap, fdes), &fp)) != 0)
    962 		return (error);
    963 	error = copyin((caddr_t)SCARG(uap, alen), (caddr_t)&len, sizeof(len));
    964 	if (error)
    965 		goto out;
    966 	so = (struct socket *)fp->f_data;
    967 	m = m_getclr(M_WAIT, MT_SONAME);
    968 	error = (*so->so_proto->pr_usrreq)(so, PRU_SOCKADDR, (struct mbuf *)0,
    969 	    m, (struct mbuf *)0, (struct proc *)0);
    970 	if (error)
    971 		goto bad;
    972 	if (len > m->m_len)
    973 		len = m->m_len;
    974 	error = copyout(mtod(m, caddr_t), (caddr_t)SCARG(uap, asa), len);
    975 	if (error == 0)
    976 		error = copyout((caddr_t)&len, (caddr_t)SCARG(uap, alen),
    977 		    sizeof(len));
    978  bad:
    979 	m_freem(m);
    980  out:
    981 	FILE_UNUSE(fp, p);
    982 	return (error);
    983 }
    984 
    985 /*
    986  * Get name of peer for connected socket.
    987  */
    988 /* ARGSUSED */
    989 int
    990 sys_getpeername(struct lwp *l, void *v, register_t *retval)
    991 {
    992 	struct sys_getpeername_args /* {
    993 		syscallarg(int)			fdes;
    994 		syscallarg(struct sockaddr *)	asa;
    995 		syscallarg(unsigned int *)	alen;
    996 	} */ *uap = v;
    997 	struct proc	*p;
    998 	struct file	*fp;
    999 	struct socket	*so;
   1000 	struct mbuf	*m;
   1001 	unsigned int	len;
   1002 	int		error;
   1003 
   1004 	p = l->l_proc;
   1005 	/* getsock() will use the descriptor for us */
   1006 	if ((error = getsock(p->p_fd, SCARG(uap, fdes), &fp)) != 0)
   1007 		return (error);
   1008 	so = (struct socket *)fp->f_data;
   1009 	if ((so->so_state & (SS_ISCONNECTED|SS_ISCONFIRMING)) == 0) {
   1010 		error = ENOTCONN;
   1011 		goto out;
   1012 	}
   1013 	error = copyin((caddr_t)SCARG(uap, alen), (caddr_t)&len, sizeof(len));
   1014 	if (error)
   1015 		goto out;
   1016 	m = m_getclr(M_WAIT, MT_SONAME);
   1017 	error = (*so->so_proto->pr_usrreq)(so, PRU_PEERADDR, (struct mbuf *)0,
   1018 	    m, (struct mbuf *)0, (struct proc *)0);
   1019 	if (error)
   1020 		goto bad;
   1021 	if (len > m->m_len)
   1022 		len = m->m_len;
   1023 	error = copyout(mtod(m, caddr_t), (caddr_t)SCARG(uap, asa), len);
   1024 	if (error)
   1025 		goto bad;
   1026 	error = copyout((caddr_t)&len, (caddr_t)SCARG(uap, alen), sizeof(len));
   1027  bad:
   1028 	m_freem(m);
   1029  out:
   1030 	FILE_UNUSE(fp, p);
   1031 	return (error);
   1032 }
   1033 
   1034 /*
   1035  * XXX In a perfect world, we wouldn't pass around socket control
   1036  * XXX arguments in mbufs, and this could go away.
   1037  */
   1038 int
   1039 sockargs(struct mbuf **mp, const void *buf, size_t buflen, int type)
   1040 {
   1041 	struct sockaddr	*sa;
   1042 	struct mbuf	*m;
   1043 	int		error;
   1044 
   1045 	/*
   1046 	 * We can't allow socket names > UCHAR_MAX in length, since that
   1047 	 * will overflow sa_len.  Control data more than a page size in
   1048 	 * length is just too much.
   1049 	 */
   1050 	if (buflen > (type == MT_SONAME ? UCHAR_MAX : PAGE_SIZE))
   1051 		return (EINVAL);
   1052 
   1053 	/* Allocate an mbuf to hold the arguments. */
   1054 	m = m_get(M_WAIT, type);
   1055 	if (buflen > MLEN) {
   1056 		/*
   1057 		 * Won't fit into a regular mbuf, so we allocate just
   1058 		 * enough external storage to hold the argument.
   1059 		 */
   1060 		MEXTMALLOC(m, buflen, M_WAITOK);
   1061 	}
   1062 	m->m_len = buflen;
   1063 	error = copyin(buf, mtod(m, caddr_t), buflen);
   1064 	if (error) {
   1065 		(void) m_free(m);
   1066 		return (error);
   1067 	}
   1068 	*mp = m;
   1069 	if (type == MT_SONAME) {
   1070 		sa = mtod(m, struct sockaddr *);
   1071 #if BYTE_ORDER != BIG_ENDIAN
   1072 		/*
   1073 		 * 4.3BSD compat thing - need to stay, since bind(2),
   1074 		 * connect(2), sendto(2) were not versioned for COMPAT_43.
   1075 		 */
   1076 		if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
   1077 			sa->sa_family = sa->sa_len;
   1078 #endif
   1079 		sa->sa_len = buflen;
   1080 	}
   1081 	return (0);
   1082 }
   1083 
   1084 int
   1085 getsock(struct filedesc *fdp, int fdes, struct file **fpp)
   1086 {
   1087 	struct file	*fp;
   1088 
   1089 	if ((fp = fd_getfile(fdp, fdes)) == NULL)
   1090 		return (EBADF);
   1091 
   1092 	FILE_USE(fp);
   1093 
   1094 	if (fp->f_type != DTYPE_SOCKET) {
   1095 		FILE_UNUSE(fp, NULL);
   1096 		return (ENOTSOCK);
   1097 	}
   1098 	*fpp = fp;
   1099 	return (0);
   1100 }
   1101