Home | History | Annotate | Line # | Download | only in kern
uipc_syscalls.c revision 1.81.2.2
      1 /*	$NetBSD: uipc_syscalls.c,v 1.81.2.2 2004/08/03 10:52:58 skrll 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. Neither the name of the University nor the names of its contributors
     16  *    may be used to endorse or promote products derived from this software
     17  *    without specific prior written permission.
     18  *
     19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     29  * SUCH DAMAGE.
     30  *
     31  *	@(#)uipc_syscalls.c	8.6 (Berkeley) 2/14/95
     32  */
     33 
     34 #include <sys/cdefs.h>
     35 __KERNEL_RCSID(0, "$NetBSD: uipc_syscalls.c,v 1.81.2.2 2004/08/03 10:52:58 skrll Exp $");
     36 
     37 #include "opt_ktrace.h"
     38 #include "opt_pipe.h"
     39 
     40 #include <sys/param.h>
     41 #include <sys/systm.h>
     42 #include <sys/filedesc.h>
     43 #include <sys/proc.h>
     44 #include <sys/file.h>
     45 #include <sys/buf.h>
     46 #include <sys/malloc.h>
     47 #include <sys/mbuf.h>
     48 #include <sys/protosw.h>
     49 #include <sys/socket.h>
     50 #include <sys/socketvar.h>
     51 #include <sys/signalvar.h>
     52 #include <sys/un.h>
     53 #ifdef KTRACE
     54 #include <sys/ktrace.h>
     55 #endif
     56 #include <sys/event.h>
     57 
     58 #include <sys/mount.h>
     59 #include <sys/sa.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), l);
     94 	if (error) {
     95 		FILE_UNUSE(fp, l);
     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, l);
    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, l);
    129 		return (error);
    130 	}
    131 	MCLAIM(nam, ((struct socket *)fp->f_data)->so_mowner);
    132 	error = sobind((struct socket *)fp->f_data, nam, l);
    133 	m_freem(nam);
    134 	FILE_UNUSE(fp, l);
    135 	return (error);
    136 }
    137 
    138 /* ARGSUSED */
    139 int
    140 sys_listen(struct lwp *l, void *v, register_t *retval)
    141 {
    142 	struct sys_listen_args /* {
    143 		syscallarg(int)	s;
    144 		syscallarg(int)	backlog;
    145 	} */ *uap = v;
    146 	struct proc	*p;
    147 	struct file	*fp;
    148 	int		error;
    149 
    150 	p = l->l_proc;
    151 	/* getsock() will use the descriptor for us */
    152 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    153 		return (error);
    154 	error = solisten((struct socket *)fp->f_data, SCARG(uap, backlog));
    155 	FILE_UNUSE(fp, l);
    156 	return (error);
    157 }
    158 
    159 int
    160 sys_accept(struct lwp *l, void *v, register_t *retval)
    161 {
    162 	struct sys_accept_args /* {
    163 		syscallarg(int)			s;
    164 		syscallarg(struct sockaddr *)	name;
    165 		syscallarg(unsigned int *)	anamelen;
    166 	} */ *uap = v;
    167 	struct proc	*p;
    168 	struct filedesc	*fdp;
    169 	struct file	*fp;
    170 	struct mbuf	*nam;
    171 	unsigned int	namelen;
    172 	int		error, s, fd;
    173 	struct socket	*so;
    174 	int		fflag;
    175 
    176 	p = l->l_proc;
    177 	fdp = p->p_fd;
    178 	if (SCARG(uap, name) && (error = copyin(SCARG(uap, anamelen),
    179 	    &namelen, sizeof(namelen))))
    180 		return (error);
    181 
    182 	/* getsock() will use the descriptor for us */
    183 	if ((error = getsock(fdp, SCARG(uap, s), &fp)) != 0)
    184 		return (error);
    185 	s = splsoftnet();
    186 	so = (struct socket *)fp->f_data;
    187 	FILE_UNUSE(fp, l);
    188 	if (!(so->so_proto->pr_flags & PR_LISTEN)) {
    189 		splx(s);
    190 		return (EOPNOTSUPP);
    191 	}
    192 	if ((so->so_options & SO_ACCEPTCONN) == 0) {
    193 		splx(s);
    194 		return (EINVAL);
    195 	}
    196 	if ((so->so_state & SS_NBIO) && so->so_qlen == 0) {
    197 		splx(s);
    198 		return (EWOULDBLOCK);
    199 	}
    200 	while (so->so_qlen == 0 && so->so_error == 0) {
    201 		if (so->so_state & SS_CANTRCVMORE) {
    202 			so->so_error = ECONNABORTED;
    203 			break;
    204 		}
    205 		error = tsleep((caddr_t)&so->so_timeo, PSOCK | PCATCH,
    206 			       netcon, 0);
    207 		if (error) {
    208 			splx(s);
    209 			return (error);
    210 		}
    211 	}
    212 	if (so->so_error) {
    213 		error = so->so_error;
    214 		so->so_error = 0;
    215 		splx(s);
    216 		return (error);
    217 	}
    218 	fflag = fp->f_flag;
    219 	/* falloc() will use the descriptor for us */
    220 	if ((error = falloc(p, &fp, &fd)) != 0) {
    221 		splx(s);
    222 		return (error);
    223 	}
    224 	*retval = fd;
    225 
    226 	/* connection has been removed from the listen queue */
    227 	KNOTE(&so->so_rcv.sb_sel.sel_klist, 0);
    228 
    229 	{ struct socket *aso = TAILQ_FIRST(&so->so_q);
    230 	  if (soqremque(aso, 1) == 0)
    231 		panic("accept");
    232 	  so = aso;
    233 	}
    234 	fp->f_type = DTYPE_SOCKET;
    235 	fp->f_flag = fflag;
    236 	fp->f_ops = &socketops;
    237 	fp->f_data = (caddr_t)so;
    238 	FILE_UNUSE(fp, l);
    239 	nam = m_get(M_WAIT, MT_SONAME);
    240 	if ((error = soaccept(so, nam)) == 0 && SCARG(uap, name)) {
    241 		if (namelen > nam->m_len)
    242 			namelen = nam->m_len;
    243 		/* SHOULD COPY OUT A CHAIN HERE */
    244 		if ((error = copyout(mtod(nam, caddr_t),
    245 		    (caddr_t)SCARG(uap, name), namelen)) == 0)
    246 			error = copyout((caddr_t)&namelen,
    247 			    (caddr_t)SCARG(uap, anamelen),
    248 			    sizeof(*SCARG(uap, anamelen)));
    249 	}
    250 	/* if an error occurred, free the file descriptor */
    251 	if (error) {
    252 		fdremove(fdp, fd);
    253 		ffree(fp);
    254 	}
    255 	m_freem(nam);
    256 	splx(s);
    257 	FILE_SET_MATURE(fp);
    258 	return (error);
    259 }
    260 
    261 /* ARGSUSED */
    262 int
    263 sys_connect(struct lwp *l, void *v, register_t *retval)
    264 {
    265 	struct sys_connect_args /* {
    266 		syscallarg(int)				s;
    267 		syscallarg(const struct sockaddr *)	name;
    268 		syscallarg(unsigned int)		namelen;
    269 	} */ *uap = v;
    270 	struct proc	*p;
    271 	struct file	*fp;
    272 	struct socket	*so;
    273 	struct mbuf	*nam;
    274 	int		error, s;
    275 	int		interrupted = 0;
    276 
    277 	p = l->l_proc;
    278 	/* getsock() will use the descriptor for us */
    279 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    280 		return (error);
    281 	so = (struct socket *)fp->f_data;
    282 	if (so->so_state & SS_ISCONNECTING) {
    283 		error = EALREADY;
    284 		goto out;
    285 	}
    286 	error = sockargs(&nam, SCARG(uap, name), SCARG(uap, namelen),
    287 	    MT_SONAME);
    288 	if (error)
    289 		goto out;
    290 	MCLAIM(nam, so->so_mowner);
    291 	error = soconnect(so, nam, l);
    292 	if (error)
    293 		goto bad;
    294 	if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING)) {
    295 		m_freem(nam);
    296 		error = EINPROGRESS;
    297 		goto out;
    298 	}
    299 	s = splsoftnet();
    300 	while ((so->so_state & SS_ISCONNECTING) && so->so_error == 0) {
    301 		error = tsleep((caddr_t)&so->so_timeo, PSOCK | PCATCH,
    302 			       netcon, 0);
    303 		if (error) {
    304 			if (error == EINTR || error == ERESTART)
    305 				interrupted = 1;
    306 			break;
    307 		}
    308 	}
    309 	if (error == 0) {
    310 		error = so->so_error;
    311 		so->so_error = 0;
    312 	}
    313 	splx(s);
    314  bad:
    315 	if (!interrupted)
    316 		so->so_state &= ~SS_ISCONNECTING;
    317 	m_freem(nam);
    318 	if (error == ERESTART)
    319 		error = EINTR;
    320  out:
    321 	FILE_UNUSE(fp, l);
    322 	return (error);
    323 }
    324 
    325 int
    326 sys_socketpair(struct lwp *l, void *v, register_t *retval)
    327 {
    328 	struct sys_socketpair_args /* {
    329 		syscallarg(int)		domain;
    330 		syscallarg(int)		type;
    331 		syscallarg(int)		protocol;
    332 		syscallarg(int *)	rsv;
    333 	} */ *uap = v;
    334 	struct proc *p;
    335 	struct filedesc	*fdp;
    336 	struct file	*fp1, *fp2;
    337 	struct socket	*so1, *so2;
    338 	int		fd, error, sv[2];
    339 
    340 	p = l->l_proc;
    341 	fdp = p->p_fd;
    342 	error = socreate(SCARG(uap, domain), &so1, SCARG(uap, type),
    343 			 SCARG(uap, protocol), l);
    344 	if (error)
    345 		return (error);
    346 	error = socreate(SCARG(uap, domain), &so2, SCARG(uap, type),
    347 			 SCARG(uap, protocol), l);
    348 	if (error)
    349 		goto free1;
    350 	/* falloc() will use the descriptor for us */
    351 	if ((error = falloc(p, &fp1, &fd)) != 0)
    352 		goto free2;
    353 	sv[0] = fd;
    354 	fp1->f_flag = FREAD|FWRITE;
    355 	fp1->f_type = DTYPE_SOCKET;
    356 	fp1->f_ops = &socketops;
    357 	fp1->f_data = (caddr_t)so1;
    358 	if ((error = falloc(p, &fp2, &fd)) != 0)
    359 		goto free3;
    360 	fp2->f_flag = FREAD|FWRITE;
    361 	fp2->f_type = DTYPE_SOCKET;
    362 	fp2->f_ops = &socketops;
    363 	fp2->f_data = (caddr_t)so2;
    364 	sv[1] = fd;
    365 	if ((error = soconnect2(so1, so2)) != 0)
    366 		goto free4;
    367 	if (SCARG(uap, type) == SOCK_DGRAM) {
    368 		/*
    369 		 * Datagram socket connection is asymmetric.
    370 		 */
    371 		 if ((error = soconnect2(so2, so1)) != 0)
    372 			goto free4;
    373 	}
    374 	error = copyout((caddr_t)sv, (caddr_t)SCARG(uap, rsv),
    375 	    2 * sizeof(int));
    376 	FILE_SET_MATURE(fp1);
    377 	FILE_SET_MATURE(fp2);
    378 	FILE_UNUSE(fp1, l);
    379 	FILE_UNUSE(fp2, l);
    380 	return (error);
    381  free4:
    382 	FILE_UNUSE(fp2, l);
    383 	ffree(fp2);
    384 	fdremove(fdp, sv[1]);
    385  free3:
    386 	FILE_UNUSE(fp1, l);
    387 	ffree(fp1);
    388 	fdremove(fdp, sv[0]);
    389  free2:
    390 	(void)soclose(so2);
    391  free1:
    392 	(void)soclose(so1);
    393 	return (error);
    394 }
    395 
    396 int
    397 sys_sendto(struct lwp *l, void *v, register_t *retval)
    398 {
    399 	struct sys_sendto_args /* {
    400 		syscallarg(int)				s;
    401 		syscallarg(const void *)		buf;
    402 		syscallarg(size_t)			len;
    403 		syscallarg(int)				flags;
    404 		syscallarg(const struct sockaddr *)	to;
    405 		syscallarg(unsigned int)		tolen;
    406 	} */ *uap = v;
    407 	struct msghdr	msg;
    408 	struct iovec	aiov;
    409 
    410 	msg.msg_name = (caddr_t)SCARG(uap, to);		/* XXX kills const */
    411 	msg.msg_namelen = SCARG(uap, tolen);
    412 	msg.msg_iov = &aiov;
    413 	msg.msg_iovlen = 1;
    414 	msg.msg_control = 0;
    415 	msg.msg_flags = 0;
    416 	aiov.iov_base = (char *)SCARG(uap, buf);	/* XXX kills const */
    417 	aiov.iov_len = SCARG(uap, len);
    418 	return (sendit(l, SCARG(uap, s), &msg, SCARG(uap, flags), retval));
    419 }
    420 
    421 int
    422 sys_sendmsg(struct lwp *l, void *v, register_t *retval)
    423 {
    424 	struct sys_sendmsg_args /* {
    425 		syscallarg(int)				s;
    426 		syscallarg(const struct msghdr *)	msg;
    427 		syscallarg(int)				flags;
    428 	} */ *uap = v;
    429 	struct msghdr	msg;
    430 	struct iovec	aiov[UIO_SMALLIOV], *iov;
    431 	int		error;
    432 
    433 	error = copyin(SCARG(uap, msg), (caddr_t)&msg, sizeof(msg));
    434 	if (error)
    435 		return (error);
    436 	if ((unsigned int)msg.msg_iovlen > UIO_SMALLIOV) {
    437 		if ((unsigned int)msg.msg_iovlen > IOV_MAX)
    438 			return (EMSGSIZE);
    439 		iov = malloc(sizeof(struct iovec) * msg.msg_iovlen,
    440 		    M_IOV, M_WAITOK);
    441 	} else
    442 		iov = aiov;
    443 	if ((unsigned int)msg.msg_iovlen > 0) {
    444 		error = copyin((caddr_t)msg.msg_iov, (caddr_t)iov,
    445 		    (size_t)(msg.msg_iovlen * sizeof(struct iovec)));
    446 		if (error)
    447 			goto done;
    448 	}
    449 	msg.msg_iov = iov;
    450 	msg.msg_flags = 0;
    451 	error = sendit(l, SCARG(uap, s), &msg, SCARG(uap, flags), retval);
    452 done:
    453 	if (iov != aiov)
    454 		free(iov, M_IOV);
    455 	return (error);
    456 }
    457 
    458 int
    459 sendit(struct lwp *l, int s, struct msghdr *mp, int flags, register_t *retsize)
    460 {
    461 	struct proc	*p;
    462 	struct file	*fp;
    463 	struct uio	auio;
    464 	struct iovec	*iov;
    465 	int		i, len, error;
    466 	struct mbuf	*to, *control;
    467 	struct socket	*so;
    468 #ifdef KTRACE
    469 	struct iovec	*ktriov;
    470 #endif
    471 
    472 #ifdef KTRACE
    473 	ktriov = NULL;
    474 #endif
    475 	p = l->l_proc;
    476 	/* getsock() will use the descriptor for us */
    477 	if ((error = getsock(p->p_fd, s, &fp)) != 0)
    478 		return (error);
    479 	so = (struct socket *)fp->f_data;
    480 	auio.uio_iov = mp->msg_iov;
    481 	auio.uio_iovcnt = mp->msg_iovlen;
    482 	auio.uio_segflg = UIO_USERSPACE;
    483 	auio.uio_rw = UIO_WRITE;
    484 	auio.uio_lwp = l;
    485 	auio.uio_offset = 0;			/* XXX */
    486 	auio.uio_resid = 0;
    487 	iov = mp->msg_iov;
    488 	for (i = 0; i < mp->msg_iovlen; i++, iov++) {
    489 #if 0
    490 		/* cannot happen; iov_len is unsigned */
    491 		if (iov->iov_len < 0) {
    492 			error = EINVAL;
    493 			goto out;
    494 		}
    495 #endif
    496 		/*
    497 		 * Writes return ssize_t because -1 is returned on error.
    498 		 * Therefore, we must restrict the length to SSIZE_MAX to
    499 		 * avoid garbage return values.
    500 		 */
    501 		auio.uio_resid += iov->iov_len;
    502 		if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
    503 			error = EINVAL;
    504 			goto out;
    505 		}
    506 	}
    507 	if (mp->msg_name) {
    508 		error = sockargs(&to, mp->msg_name, mp->msg_namelen,
    509 				 MT_SONAME);
    510 		if (error)
    511 			goto out;
    512 		MCLAIM(to, so->so_mowner);
    513 	} else
    514 		to = 0;
    515 	if (mp->msg_control) {
    516 		if (mp->msg_controllen < sizeof(struct cmsghdr)) {
    517 			error = EINVAL;
    518 			goto bad;
    519 		}
    520 		error = sockargs(&control, mp->msg_control,
    521 				 mp->msg_controllen, MT_CONTROL);
    522 		if (error)
    523 			goto bad;
    524 		MCLAIM(control, so->so_mowner);
    525 	} else
    526 		control = 0;
    527 #ifdef KTRACE
    528 	if (KTRPOINT(p, KTR_GENIO)) {
    529 		int iovlen = auio.uio_iovcnt * sizeof(struct iovec);
    530 
    531 		ktriov = malloc(iovlen, M_TEMP, M_WAITOK);
    532 		memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen);
    533 	}
    534 #endif
    535 	len = auio.uio_resid;
    536 	error = (*so->so_send)(so, to, &auio, NULL, control, flags, l);
    537 	if (error) {
    538 		if (auio.uio_resid != len && (error == ERESTART ||
    539 		    error == EINTR || error == EWOULDBLOCK))
    540 			error = 0;
    541 		if (error == EPIPE)
    542 			psignal(p, SIGPIPE);
    543 	}
    544 	if (error == 0)
    545 		*retsize = len - auio.uio_resid;
    546 #ifdef KTRACE
    547 	if (ktriov != NULL) {
    548 		if (error == 0)
    549 			ktrgenio(l, s, UIO_WRITE, ktriov, *retsize, error);
    550 		free(ktriov, M_TEMP);
    551 	}
    552 #endif
    553  bad:
    554 	if (to)
    555 		m_freem(to);
    556  out:
    557 	FILE_UNUSE(fp, l);
    558 	return (error);
    559 }
    560 
    561 int
    562 sys_recvfrom(struct lwp *l, void *v, register_t *retval)
    563 {
    564 	struct sys_recvfrom_args /* {
    565 		syscallarg(int)			s;
    566 		syscallarg(void *)		buf;
    567 		syscallarg(size_t)		len;
    568 		syscallarg(int)			flags;
    569 		syscallarg(struct sockaddr *)	from;
    570 		syscallarg(unsigned int *)	fromlenaddr;
    571 	} */ *uap = v;
    572 	struct msghdr	msg;
    573 	struct iovec	aiov;
    574 	int		error;
    575 
    576 	if (SCARG(uap, fromlenaddr)) {
    577 		error = copyin((caddr_t)SCARG(uap, fromlenaddr),
    578 			       (caddr_t)&msg.msg_namelen,
    579 			       sizeof(msg.msg_namelen));
    580 		if (error)
    581 			return (error);
    582 	} else
    583 		msg.msg_namelen = 0;
    584 	msg.msg_name = (caddr_t)SCARG(uap, from);
    585 	msg.msg_iov = &aiov;
    586 	msg.msg_iovlen = 1;
    587 	aiov.iov_base = SCARG(uap, buf);
    588 	aiov.iov_len = SCARG(uap, len);
    589 	msg.msg_control = 0;
    590 	msg.msg_flags = SCARG(uap, flags);
    591 	return (recvit(l, SCARG(uap, s), &msg,
    592 		       (caddr_t)SCARG(uap, fromlenaddr), retval));
    593 }
    594 
    595 int
    596 sys_recvmsg(struct lwp *l, void *v, register_t *retval)
    597 {
    598 	struct sys_recvmsg_args /* {
    599 		syscallarg(int)			s;
    600 		syscallarg(struct msghdr *)	msg;
    601 		syscallarg(int)			flags;
    602 	} */ *uap = v;
    603 	struct iovec	aiov[UIO_SMALLIOV], *uiov, *iov;
    604 	struct msghdr	msg;
    605 	int		error;
    606 
    607 	error = copyin((caddr_t)SCARG(uap, msg), (caddr_t)&msg,
    608 		       sizeof(msg));
    609 	if (error)
    610 		return (error);
    611 	if ((unsigned int)msg.msg_iovlen > UIO_SMALLIOV) {
    612 		if ((unsigned int)msg.msg_iovlen > IOV_MAX)
    613 			return (EMSGSIZE);
    614 		iov = malloc(sizeof(struct iovec) * msg.msg_iovlen,
    615 		    M_IOV, M_WAITOK);
    616 	} else
    617 		iov = aiov;
    618 	if ((unsigned int)msg.msg_iovlen > 0) {
    619 		error = copyin((caddr_t)msg.msg_iov, (caddr_t)iov,
    620 		    (size_t)(msg.msg_iovlen * sizeof(struct iovec)));
    621 		if (error)
    622 			goto done;
    623 	}
    624 	uiov = msg.msg_iov;
    625 	msg.msg_iov = iov;
    626 	msg.msg_flags = SCARG(uap, flags);
    627 	if ((error = recvit(l, SCARG(uap, s), &msg, (caddr_t)0, retval)) == 0) {
    628 		msg.msg_iov = uiov;
    629 		error = copyout((caddr_t)&msg, (caddr_t)SCARG(uap, msg),
    630 		    sizeof(msg));
    631 	}
    632 done:
    633 	if (iov != aiov)
    634 		free(iov, M_IOV);
    635 	return (error);
    636 }
    637 
    638 int
    639 recvit(struct lwp *l, int s, struct msghdr *mp, caddr_t namelenp,
    640 	register_t *retsize)
    641 {
    642 	struct proc	*p;
    643 	struct file	*fp;
    644 	struct uio	auio;
    645 	struct iovec	*iov;
    646 	int		i, len, error;
    647 	struct mbuf	*from, *control;
    648 	struct socket	*so;
    649 #ifdef KTRACE
    650 	struct iovec	*ktriov;
    651 #endif
    652 
    653 	p = l->l_proc;
    654 	from = 0;
    655 	control = 0;
    656 #ifdef KTRACE
    657 	ktriov = NULL;
    658 #endif
    659 
    660 	/* getsock() will use the descriptor for us */
    661 	if ((error = getsock(p->p_fd, s, &fp)) != 0)
    662 		return (error);
    663 	so = (struct socket *)fp->f_data;
    664 	auio.uio_iov = mp->msg_iov;
    665 	auio.uio_iovcnt = mp->msg_iovlen;
    666 	auio.uio_segflg = UIO_USERSPACE;
    667 	auio.uio_rw = UIO_READ;
    668 	auio.uio_lwp = l;
    669 	auio.uio_offset = 0;			/* XXX */
    670 	auio.uio_resid = 0;
    671 	iov = mp->msg_iov;
    672 	for (i = 0; i < mp->msg_iovlen; i++, iov++) {
    673 #if 0
    674 		/* cannot happen iov_len is unsigned */
    675 		if (iov->iov_len < 0) {
    676 			error = EINVAL;
    677 			goto out1;
    678 		}
    679 #endif
    680 		/*
    681 		 * Reads return ssize_t because -1 is returned on error.
    682 		 * Therefore we must restrict the length to SSIZE_MAX to
    683 		 * avoid garbage return values.
    684 		 */
    685 		auio.uio_resid += iov->iov_len;
    686 		if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
    687 			error = EINVAL;
    688 			goto out1;
    689 		}
    690 	}
    691 #ifdef KTRACE
    692 	if (KTRPOINT(p, KTR_GENIO)) {
    693 		int iovlen = auio.uio_iovcnt * sizeof(struct iovec);
    694 
    695 		ktriov = malloc(iovlen, M_TEMP, M_WAITOK);
    696 		memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen);
    697 	}
    698 #endif
    699 	len = auio.uio_resid;
    700 	error = (*so->so_receive)(so, &from, &auio, NULL,
    701 			  mp->msg_control ? &control : NULL, &mp->msg_flags);
    702 	if (error) {
    703 		if (auio.uio_resid != len && (error == ERESTART ||
    704 		    error == EINTR || error == EWOULDBLOCK))
    705 			error = 0;
    706 	}
    707 #ifdef KTRACE
    708 	if (ktriov != NULL) {
    709 		if (error == 0)
    710 			ktrgenio(l, s, UIO_READ, ktriov,
    711 			    len - auio.uio_resid, error);
    712 		free(ktriov, M_TEMP);
    713 	}
    714 #endif
    715 	if (error)
    716 		goto out;
    717 	*retsize = len - auio.uio_resid;
    718 	if (mp->msg_name) {
    719 		len = mp->msg_namelen;
    720 		if (len <= 0 || from == 0)
    721 			len = 0;
    722 		else {
    723 			if (len > from->m_len)
    724 				len = from->m_len;
    725 			/* else if len < from->m_len ??? */
    726 			error = copyout(mtod(from, caddr_t),
    727 					(caddr_t)mp->msg_name, (unsigned)len);
    728 			if (error)
    729 				goto out;
    730 		}
    731 		mp->msg_namelen = len;
    732 		if (namelenp &&
    733 		    (error = copyout((caddr_t)&len, namelenp, sizeof(int))))
    734 			goto out;
    735 	}
    736 	if (mp->msg_control) {
    737 		len = mp->msg_controllen;
    738 		if (len <= 0 || control == 0)
    739 			len = 0;
    740 		else {
    741 			struct mbuf *m = control;
    742 			caddr_t p = (caddr_t)mp->msg_control;
    743 
    744 			do {
    745 				i = m->m_len;
    746 				if (len < i) {
    747 					mp->msg_flags |= MSG_CTRUNC;
    748 					i = len;
    749 				}
    750 				error = copyout(mtod(m, caddr_t), p,
    751 				    (unsigned)i);
    752 				if (m->m_next)
    753 					i = ALIGN(i);
    754 				p += i;
    755 				len -= i;
    756 				if (error != 0 || len <= 0)
    757 					break;
    758 			} while ((m = m->m_next) != NULL);
    759 			len = p - (caddr_t)mp->msg_control;
    760 		}
    761 		mp->msg_controllen = len;
    762 	}
    763  out:
    764 	if (from)
    765 		m_freem(from);
    766 	if (control)
    767 		m_freem(control);
    768  out1:
    769 	FILE_UNUSE(fp, l);
    770 	return (error);
    771 }
    772 
    773 /* ARGSUSED */
    774 int
    775 sys_shutdown(struct lwp *l, void *v, register_t *retval)
    776 {
    777 	struct sys_shutdown_args /* {
    778 		syscallarg(int)	s;
    779 		syscallarg(int)	how;
    780 	} */ *uap = v;
    781 	struct proc	*p;
    782 	struct file	*fp;
    783 	int		error;
    784 
    785 	p = l->l_proc;
    786 	/* getsock() will use the descriptor for us */
    787 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    788 		return (error);
    789 	error = soshutdown((struct socket *)fp->f_data, SCARG(uap, how));
    790 	FILE_UNUSE(fp, l);
    791 	return (error);
    792 }
    793 
    794 /* ARGSUSED */
    795 int
    796 sys_setsockopt(struct lwp *l, void *v, register_t *retval)
    797 {
    798 	struct sys_setsockopt_args /* {
    799 		syscallarg(int)			s;
    800 		syscallarg(int)			level;
    801 		syscallarg(int)			name;
    802 		syscallarg(const void *)	val;
    803 		syscallarg(unsigned int)	valsize;
    804 	} */ *uap = v;
    805 	struct proc	*p;
    806 	struct file	*fp;
    807 	struct mbuf	*m;
    808 	struct socket	*so;
    809 	int		error;
    810 	unsigned int	len;
    811 
    812 	p = l->l_proc;
    813 	m = NULL;
    814 	/* getsock() will use the descriptor for us */
    815 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    816 		return (error);
    817 	so = (struct socket *)fp->f_data;
    818 	len = SCARG(uap, valsize);
    819 	if (len > MCLBYTES) {
    820 		error = EINVAL;
    821 		goto out;
    822 	}
    823 	if (SCARG(uap, val)) {
    824 		m = m_get(M_WAIT, MT_SOOPTS);
    825 		MCLAIM(m, so->so_mowner);
    826 		if (len > MLEN)
    827 			m_clget(m, M_WAIT);
    828 		error = copyin(SCARG(uap, val), mtod(m, caddr_t), len);
    829 		if (error) {
    830 			(void) m_free(m);
    831 			goto out;
    832 		}
    833 		m->m_len = SCARG(uap, valsize);
    834 	}
    835 	error = sosetopt(so, SCARG(uap, level), SCARG(uap, name), m);
    836  out:
    837 	FILE_UNUSE(fp, l);
    838 	return (error);
    839 }
    840 
    841 /* ARGSUSED */
    842 int
    843 sys_getsockopt(struct lwp *l, void *v, register_t *retval)
    844 {
    845 	struct sys_getsockopt_args /* {
    846 		syscallarg(int)			s;
    847 		syscallarg(int)			level;
    848 		syscallarg(int)			name;
    849 		syscallarg(void *)		val;
    850 		syscallarg(unsigned int *)	avalsize;
    851 	} */ *uap = v;
    852 	struct proc	*p;
    853 	struct file	*fp;
    854 	struct mbuf	*m;
    855 	unsigned int	op, i, valsize;
    856 	int		error;
    857 
    858 	p = l->l_proc;
    859 	m = NULL;
    860 	/* getsock() will use the descriptor for us */
    861 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    862 		return (error);
    863 	if (SCARG(uap, val)) {
    864 		error = copyin((caddr_t)SCARG(uap, avalsize),
    865 			       (caddr_t)&valsize, sizeof(valsize));
    866 		if (error)
    867 			goto out;
    868 	} else
    869 		valsize = 0;
    870 	if ((error = sogetopt((struct socket *)fp->f_data, SCARG(uap, level),
    871 	    SCARG(uap, name), &m)) == 0 && SCARG(uap, val) && valsize &&
    872 	    m != NULL) {
    873 		op = 0;
    874 		while (m && !error && op < valsize) {
    875 			i = min(m->m_len, (valsize - op));
    876 			error = copyout(mtod(m, caddr_t), SCARG(uap, val), i);
    877 			op += i;
    878 			SCARG(uap, val) = ((u_int8_t *)SCARG(uap, val)) + i;
    879 			m = m_free(m);
    880 		}
    881 		valsize = op;
    882 		if (error == 0)
    883 			error = copyout(&valsize,
    884 					SCARG(uap, avalsize), sizeof(valsize));
    885 	}
    886 	if (m != NULL)
    887 		(void) m_freem(m);
    888  out:
    889 	FILE_UNUSE(fp, l);
    890 	return (error);
    891 }
    892 
    893 #ifdef PIPE_SOCKETPAIR
    894 /* ARGSUSED */
    895 int
    896 sys_pipe(struct lwp *l, void *v, register_t *retval)
    897 {
    898 	struct proc	*p;
    899 	struct filedesc	*fdp;
    900 	struct file	*rf, *wf;
    901 	struct socket	*rso, *wso;
    902 	int		fd, error;
    903 
    904 	p = l->l_proc;
    905 	fdp = p->p_fd;
    906 	if ((error = socreate(AF_LOCAL, &rso, SOCK_STREAM, 0, p)) != 0)
    907 		return (error);
    908 	if ((error = socreate(AF_LOCAL, &wso, SOCK_STREAM, 0, p)) != 0)
    909 		goto free1;
    910 	/* remember this socket pair implements a pipe */
    911 	wso->so_state |= SS_ISAPIPE;
    912 	rso->so_state |= SS_ISAPIPE;
    913 	/* falloc() will use the descriptor for us */
    914 	if ((error = falloc(p, &rf, &fd)) != 0)
    915 		goto free2;
    916 	retval[0] = fd;
    917 	rf->f_flag = FREAD;
    918 	rf->f_type = DTYPE_SOCKET;
    919 	rf->f_ops = &socketops;
    920 	rf->f_data = (caddr_t)rso;
    921 	if ((error = falloc(p, &wf, &fd)) != 0)
    922 		goto free3;
    923 	wf->f_flag = FWRITE;
    924 	wf->f_type = DTYPE_SOCKET;
    925 	wf->f_ops = &socketops;
    926 	wf->f_data = (caddr_t)wso;
    927 	retval[1] = fd;
    928 	if ((error = unp_connect2(wso, rso, PRU_CONNECT2)) != 0)
    929 		goto free4;
    930 	FILE_SET_MATURE(rf);
    931 	FILE_SET_MATURE(wf);
    932 	FILE_UNUSE(rf, l);
    933 	FILE_UNUSE(wf, l);
    934 	return (0);
    935  free4:
    936 	FILE_UNUSE(wf, l);
    937 	ffree(wf);
    938 	fdremove(fdp, retval[1]);
    939  free3:
    940 	FILE_UNUSE(rf, l);
    941 	ffree(rf);
    942 	fdremove(fdp, retval[0]);
    943  free2:
    944 	(void)soclose(wso);
    945  free1:
    946 	(void)soclose(rso);
    947 	return (error);
    948 }
    949 #endif /* PIPE_SOCKETPAIR */
    950 
    951 /*
    952  * Get socket name.
    953  */
    954 /* ARGSUSED */
    955 int
    956 sys_getsockname(struct lwp *l, void *v, register_t *retval)
    957 {
    958 	struct sys_getsockname_args /* {
    959 		syscallarg(int)			fdes;
    960 		syscallarg(struct sockaddr *)	asa;
    961 		syscallarg(unsigned int *)	alen;
    962 	} */ *uap = v;
    963 	struct proc	*p;
    964 	struct file	*fp;
    965 	struct socket	*so;
    966 	struct mbuf	*m;
    967 	unsigned int	len;
    968 	int		error;
    969 
    970 	p = l->l_proc;
    971 	/* getsock() will use the descriptor for us */
    972 	if ((error = getsock(p->p_fd, SCARG(uap, fdes), &fp)) != 0)
    973 		return (error);
    974 	error = copyin((caddr_t)SCARG(uap, alen), (caddr_t)&len, sizeof(len));
    975 	if (error)
    976 		goto out;
    977 	so = (struct socket *)fp->f_data;
    978 	m = m_getclr(M_WAIT, MT_SONAME);
    979 	MCLAIM(m, so->so_mowner);
    980 	error = (*so->so_proto->pr_usrreq)(so, PRU_SOCKADDR, (struct mbuf *)0,
    981 	    m, (struct mbuf *)0, (struct lwp *)0);
    982 	if (error)
    983 		goto bad;
    984 	if (len > m->m_len)
    985 		len = m->m_len;
    986 	error = copyout(mtod(m, caddr_t), (caddr_t)SCARG(uap, asa), len);
    987 	if (error == 0)
    988 		error = copyout((caddr_t)&len, (caddr_t)SCARG(uap, alen),
    989 		    sizeof(len));
    990  bad:
    991 	m_freem(m);
    992  out:
    993 	FILE_UNUSE(fp, l);
    994 	return (error);
    995 }
    996 
    997 /*
    998  * Get name of peer for connected socket.
    999  */
   1000 /* ARGSUSED */
   1001 int
   1002 sys_getpeername(struct lwp *l, void *v, register_t *retval)
   1003 {
   1004 	struct sys_getpeername_args /* {
   1005 		syscallarg(int)			fdes;
   1006 		syscallarg(struct sockaddr *)	asa;
   1007 		syscallarg(unsigned int *)	alen;
   1008 	} */ *uap = v;
   1009 	struct proc	*p;
   1010 	struct file	*fp;
   1011 	struct socket	*so;
   1012 	struct mbuf	*m;
   1013 	unsigned int	len;
   1014 	int		error;
   1015 
   1016 	p = l->l_proc;
   1017 	/* getsock() will use the descriptor for us */
   1018 	if ((error = getsock(p->p_fd, SCARG(uap, fdes), &fp)) != 0)
   1019 		return (error);
   1020 	so = (struct socket *)fp->f_data;
   1021 	if ((so->so_state & (SS_ISCONNECTED|SS_ISCONFIRMING)) == 0) {
   1022 		error = ENOTCONN;
   1023 		goto out;
   1024 	}
   1025 	error = copyin((caddr_t)SCARG(uap, alen), (caddr_t)&len, sizeof(len));
   1026 	if (error)
   1027 		goto out;
   1028 	m = m_getclr(M_WAIT, MT_SONAME);
   1029 	MCLAIM(m, so->so_mowner);
   1030 	error = (*so->so_proto->pr_usrreq)(so, PRU_PEERADDR, (struct mbuf *)0,
   1031 	    m, (struct mbuf *)0, (struct lwp *)0);
   1032 	if (error)
   1033 		goto bad;
   1034 	if (len > m->m_len)
   1035 		len = m->m_len;
   1036 	error = copyout(mtod(m, caddr_t), (caddr_t)SCARG(uap, asa), len);
   1037 	if (error)
   1038 		goto bad;
   1039 	error = copyout((caddr_t)&len, (caddr_t)SCARG(uap, alen), sizeof(len));
   1040  bad:
   1041 	m_freem(m);
   1042  out:
   1043 	FILE_UNUSE(fp, l);
   1044 	return (error);
   1045 }
   1046 
   1047 /*
   1048  * XXX In a perfect world, we wouldn't pass around socket control
   1049  * XXX arguments in mbufs, and this could go away.
   1050  */
   1051 int
   1052 sockargs(struct mbuf **mp, const void *buf, size_t buflen, int type)
   1053 {
   1054 	struct sockaddr	*sa;
   1055 	struct mbuf	*m;
   1056 	int		error;
   1057 
   1058 	/*
   1059 	 * We can't allow socket names > UCHAR_MAX in length, since that
   1060 	 * will overflow sa_len.  Control data more than a page size in
   1061 	 * length is just too much.
   1062 	 */
   1063 	if (buflen > (type == MT_SONAME ? UCHAR_MAX : PAGE_SIZE))
   1064 		return (EINVAL);
   1065 
   1066 	/* Allocate an mbuf to hold the arguments. */
   1067 	m = m_get(M_WAIT, type);
   1068 	/* can't claim.  don't who to assign it to. */
   1069 	if (buflen > MLEN) {
   1070 		/*
   1071 		 * Won't fit into a regular mbuf, so we allocate just
   1072 		 * enough external storage to hold the argument.
   1073 		 */
   1074 		MEXTMALLOC(m, buflen, M_WAITOK);
   1075 	}
   1076 	m->m_len = buflen;
   1077 	error = copyin(buf, mtod(m, caddr_t), buflen);
   1078 	if (error) {
   1079 		(void) m_free(m);
   1080 		return (error);
   1081 	}
   1082 	*mp = m;
   1083 	if (type == MT_SONAME) {
   1084 		sa = mtod(m, struct sockaddr *);
   1085 #if BYTE_ORDER != BIG_ENDIAN
   1086 		/*
   1087 		 * 4.3BSD compat thing - need to stay, since bind(2),
   1088 		 * connect(2), sendto(2) were not versioned for COMPAT_43.
   1089 		 */
   1090 		if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
   1091 			sa->sa_family = sa->sa_len;
   1092 #endif
   1093 		sa->sa_len = buflen;
   1094 	}
   1095 	return (0);
   1096 }
   1097 
   1098 int
   1099 getsock(struct filedesc *fdp, int fdes, struct file **fpp)
   1100 {
   1101 	struct file	*fp;
   1102 
   1103 	if ((fp = fd_getfile(fdp, fdes)) == NULL)
   1104 		return (EBADF);
   1105 
   1106 	FILE_USE(fp);
   1107 
   1108 	if (fp->f_type != DTYPE_SOCKET) {
   1109 		FILE_UNUSE(fp, NULL);
   1110 		return (ENOTSOCK);
   1111 	}
   1112 	*fpp = fp;
   1113 	return (0);
   1114 }
   1115