Home | History | Annotate | Line # | Download | only in kern
uipc_syscalls.c revision 1.86
      1 /*	$NetBSD: uipc_syscalls.c,v 1.86 2003/11/29 10:02:42 matt 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 1.86 2003/11/29 10:02:42 matt 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));
     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 	MCLAIM(nam, ((struct socket *)fp->f_data)->so_mowner);
    132 	error = sobind((struct socket *)fp->f_data, nam, p);
    133 	m_freem(nam);
    134 	FILE_UNUSE(fp, p);
    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, p);
    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, p);
    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, p);
    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 
    276 	p = l->l_proc;
    277 	/* getsock() will use the descriptor for us */
    278 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    279 		return (error);
    280 	so = (struct socket *)fp->f_data;
    281 	if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING)) {
    282 		error = EALREADY;
    283 		goto out;
    284 	}
    285 	error = sockargs(&nam, SCARG(uap, name), SCARG(uap, namelen),
    286 	    MT_SONAME);
    287 	if (error)
    288 		goto out;
    289 	MCLAIM(nam, so->so_mowner);
    290 	error = soconnect(so, nam);
    291 	if (error)
    292 		goto bad;
    293 	if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING)) {
    294 		m_freem(nam);
    295 		error = EINPROGRESS;
    296 		goto out;
    297 	}
    298 	s = splsoftnet();
    299 	while ((so->so_state & SS_ISCONNECTING) && so->so_error == 0) {
    300 		error = tsleep((caddr_t)&so->so_timeo, PSOCK | PCATCH,
    301 			       netcon, 0);
    302 		if (error)
    303 			break;
    304 	}
    305 	if (error == 0) {
    306 		error = so->so_error;
    307 		so->so_error = 0;
    308 	}
    309 	splx(s);
    310  bad:
    311 	so->so_state &= ~SS_ISCONNECTING;
    312 	m_freem(nam);
    313 	if (error == ERESTART)
    314 		error = EINTR;
    315  out:
    316 	FILE_UNUSE(fp, p);
    317 	return (error);
    318 }
    319 
    320 int
    321 sys_socketpair(struct lwp *l, void *v, register_t *retval)
    322 {
    323 	struct sys_socketpair_args /* {
    324 		syscallarg(int)		domain;
    325 		syscallarg(int)		type;
    326 		syscallarg(int)		protocol;
    327 		syscallarg(int *)	rsv;
    328 	} */ *uap = v;
    329 	struct proc *p;
    330 	struct filedesc	*fdp;
    331 	struct file	*fp1, *fp2;
    332 	struct socket	*so1, *so2;
    333 	int		fd, error, sv[2];
    334 
    335 	p = l->l_proc;
    336 	fdp = p->p_fd;
    337 	error = socreate(SCARG(uap, domain), &so1, SCARG(uap, type),
    338 			 SCARG(uap, protocol));
    339 	if (error)
    340 		return (error);
    341 	error = socreate(SCARG(uap, domain), &so2, SCARG(uap, type),
    342 			 SCARG(uap, protocol));
    343 	if (error)
    344 		goto free1;
    345 	/* falloc() will use the descriptor for us */
    346 	if ((error = falloc(p, &fp1, &fd)) != 0)
    347 		goto free2;
    348 	sv[0] = fd;
    349 	fp1->f_flag = FREAD|FWRITE;
    350 	fp1->f_type = DTYPE_SOCKET;
    351 	fp1->f_ops = &socketops;
    352 	fp1->f_data = (caddr_t)so1;
    353 	if ((error = falloc(p, &fp2, &fd)) != 0)
    354 		goto free3;
    355 	fp2->f_flag = FREAD|FWRITE;
    356 	fp2->f_type = DTYPE_SOCKET;
    357 	fp2->f_ops = &socketops;
    358 	fp2->f_data = (caddr_t)so2;
    359 	sv[1] = fd;
    360 	if ((error = soconnect2(so1, so2)) != 0)
    361 		goto free4;
    362 	if (SCARG(uap, type) == SOCK_DGRAM) {
    363 		/*
    364 		 * Datagram socket connection is asymmetric.
    365 		 */
    366 		 if ((error = soconnect2(so2, so1)) != 0)
    367 			goto free4;
    368 	}
    369 	error = copyout((caddr_t)sv, (caddr_t)SCARG(uap, rsv),
    370 	    2 * sizeof(int));
    371 	FILE_SET_MATURE(fp1);
    372 	FILE_SET_MATURE(fp2);
    373 	FILE_UNUSE(fp1, p);
    374 	FILE_UNUSE(fp2, p);
    375 	return (error);
    376  free4:
    377 	FILE_UNUSE(fp2, p);
    378 	ffree(fp2);
    379 	fdremove(fdp, sv[1]);
    380  free3:
    381 	FILE_UNUSE(fp1, p);
    382 	ffree(fp1);
    383 	fdremove(fdp, sv[0]);
    384  free2:
    385 	(void)soclose(so2);
    386  free1:
    387 	(void)soclose(so1);
    388 	return (error);
    389 }
    390 
    391 int
    392 sys_sendto(struct lwp *l, void *v, register_t *retval)
    393 {
    394 	struct sys_sendto_args /* {
    395 		syscallarg(int)				s;
    396 		syscallarg(const void *)		buf;
    397 		syscallarg(size_t)			len;
    398 		syscallarg(int)				flags;
    399 		syscallarg(const struct sockaddr *)	to;
    400 		syscallarg(unsigned int)		tolen;
    401 	} */ *uap = v;
    402 	struct proc	*p;
    403 	struct msghdr	msg;
    404 	struct iovec	aiov;
    405 
    406 	p = l->l_proc;
    407 	msg.msg_name = (caddr_t)SCARG(uap, to);		/* XXX kills const */
    408 	msg.msg_namelen = SCARG(uap, tolen);
    409 	msg.msg_iov = &aiov;
    410 	msg.msg_iovlen = 1;
    411 	msg.msg_control = 0;
    412 	msg.msg_flags = 0;
    413 	aiov.iov_base = (char *)SCARG(uap, buf);	/* XXX kills const */
    414 	aiov.iov_len = SCARG(uap, len);
    415 	return (sendit(p, SCARG(uap, s), &msg, SCARG(uap, flags), retval));
    416 }
    417 
    418 int
    419 sys_sendmsg(struct lwp *l, void *v, register_t *retval)
    420 {
    421 	struct sys_sendmsg_args /* {
    422 		syscallarg(int)				s;
    423 		syscallarg(const struct msghdr *)	msg;
    424 		syscallarg(int)				flags;
    425 	} */ *uap = v;
    426 	struct proc	*p;
    427 	struct msghdr	msg;
    428 	struct iovec	aiov[UIO_SMALLIOV], *iov;
    429 	int		error;
    430 
    431 	error = copyin(SCARG(uap, msg), (caddr_t)&msg, sizeof(msg));
    432 	if (error)
    433 		return (error);
    434 	if ((unsigned int)msg.msg_iovlen > UIO_SMALLIOV) {
    435 		if ((unsigned int)msg.msg_iovlen > IOV_MAX)
    436 			return (EMSGSIZE);
    437 		iov = malloc(sizeof(struct iovec) * msg.msg_iovlen,
    438 		    M_IOV, M_WAITOK);
    439 	} else
    440 		iov = aiov;
    441 	if ((unsigned int)msg.msg_iovlen > 0) {
    442 		error = copyin((caddr_t)msg.msg_iov, (caddr_t)iov,
    443 		    (size_t)(msg.msg_iovlen * sizeof(struct iovec)));
    444 		if (error)
    445 			goto done;
    446 	}
    447 	msg.msg_iov = iov;
    448 	msg.msg_flags = 0;
    449 	p = l->l_proc;
    450 	error = sendit(p, SCARG(uap, s), &msg, SCARG(uap, flags), retval);
    451 done:
    452 	if (iov != aiov)
    453 		free(iov, M_IOV);
    454 	return (error);
    455 }
    456 
    457 int
    458 sendit(struct proc *p, int s, struct msghdr *mp, int flags, register_t *retsize)
    459 {
    460 	struct file	*fp;
    461 	struct uio	auio;
    462 	struct iovec	*iov;
    463 	int		i, len, error;
    464 	struct mbuf	*to, *control;
    465 	struct socket	*so;
    466 #ifdef KTRACE
    467 	struct iovec	*ktriov;
    468 #endif
    469 
    470 #ifdef KTRACE
    471 	ktriov = NULL;
    472 #endif
    473 	/* getsock() will use the descriptor for us */
    474 	if ((error = getsock(p->p_fd, s, &fp)) != 0)
    475 		return (error);
    476 	so = (struct socket *)fp->f_data;
    477 	auio.uio_iov = mp->msg_iov;
    478 	auio.uio_iovcnt = mp->msg_iovlen;
    479 	auio.uio_segflg = UIO_USERSPACE;
    480 	auio.uio_rw = UIO_WRITE;
    481 	auio.uio_procp = p;
    482 	auio.uio_offset = 0;			/* XXX */
    483 	auio.uio_resid = 0;
    484 	iov = mp->msg_iov;
    485 	for (i = 0; i < mp->msg_iovlen; i++, iov++) {
    486 #if 0
    487 		/* cannot happen; iov_len is unsigned */
    488 		if (iov->iov_len < 0) {
    489 			error = EINVAL;
    490 			goto out;
    491 		}
    492 #endif
    493 		/*
    494 		 * Writes return ssize_t because -1 is returned on error.
    495 		 * Therefore, we must restrict the length to SSIZE_MAX to
    496 		 * avoid garbage return values.
    497 		 */
    498 		auio.uio_resid += iov->iov_len;
    499 		if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
    500 			error = EINVAL;
    501 			goto out;
    502 		}
    503 	}
    504 	if (mp->msg_name) {
    505 		error = sockargs(&to, mp->msg_name, mp->msg_namelen,
    506 				 MT_SONAME);
    507 		if (error)
    508 			goto out;
    509 		MCLAIM(to, so->so_mowner);
    510 	} else
    511 		to = 0;
    512 	if (mp->msg_control) {
    513 		if (mp->msg_controllen < sizeof(struct cmsghdr)) {
    514 			error = EINVAL;
    515 			goto bad;
    516 		}
    517 		error = sockargs(&control, mp->msg_control,
    518 				 mp->msg_controllen, MT_CONTROL);
    519 		if (error)
    520 			goto bad;
    521 		MCLAIM(control, so->so_mowner);
    522 	} else
    523 		control = 0;
    524 #ifdef KTRACE
    525 	if (KTRPOINT(p, KTR_GENIO)) {
    526 		int iovlen = auio.uio_iovcnt * sizeof(struct iovec);
    527 
    528 		ktriov = malloc(iovlen, M_TEMP, M_WAITOK);
    529 		memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen);
    530 	}
    531 #endif
    532 	len = auio.uio_resid;
    533 	error = (*so->so_send)(so, to, &auio, NULL, control, flags);
    534 	if (error) {
    535 		if (auio.uio_resid != len && (error == ERESTART ||
    536 		    error == EINTR || error == EWOULDBLOCK))
    537 			error = 0;
    538 		if (error == EPIPE)
    539 			psignal(p, SIGPIPE);
    540 	}
    541 	if (error == 0)
    542 		*retsize = len - auio.uio_resid;
    543 #ifdef KTRACE
    544 	if (ktriov != NULL) {
    545 		if (error == 0)
    546 			ktrgenio(p, s, UIO_WRITE, ktriov, *retsize, error);
    547 		free(ktriov, M_TEMP);
    548 	}
    549 #endif
    550  bad:
    551 	if (to)
    552 		m_freem(to);
    553  out:
    554 	FILE_UNUSE(fp, p);
    555 	return (error);
    556 }
    557 
    558 int
    559 sys_recvfrom(struct lwp *l, void *v, register_t *retval)
    560 {
    561 	struct sys_recvfrom_args /* {
    562 		syscallarg(int)			s;
    563 		syscallarg(void *)		buf;
    564 		syscallarg(size_t)		len;
    565 		syscallarg(int)			flags;
    566 		syscallarg(struct sockaddr *)	from;
    567 		syscallarg(unsigned int *)	fromlenaddr;
    568 	} */ *uap = v;
    569 	struct proc	*p;
    570 	struct msghdr	msg;
    571 	struct iovec	aiov;
    572 	int		error;
    573 
    574 	if (SCARG(uap, fromlenaddr)) {
    575 		error = copyin((caddr_t)SCARG(uap, fromlenaddr),
    576 			       (caddr_t)&msg.msg_namelen,
    577 			       sizeof(msg.msg_namelen));
    578 		if (error)
    579 			return (error);
    580 	} else
    581 		msg.msg_namelen = 0;
    582 	msg.msg_name = (caddr_t)SCARG(uap, from);
    583 	msg.msg_iov = &aiov;
    584 	msg.msg_iovlen = 1;
    585 	aiov.iov_base = SCARG(uap, buf);
    586 	aiov.iov_len = SCARG(uap, len);
    587 	msg.msg_control = 0;
    588 	msg.msg_flags = SCARG(uap, flags);
    589 	p = l->l_proc;
    590 	return (recvit(p, SCARG(uap, s), &msg,
    591 		       (caddr_t)SCARG(uap, fromlenaddr), retval));
    592 }
    593 
    594 int
    595 sys_recvmsg(struct lwp *l, void *v, register_t *retval)
    596 {
    597 	struct sys_recvmsg_args /* {
    598 		syscallarg(int)			s;
    599 		syscallarg(struct msghdr *)	msg;
    600 		syscallarg(int)			flags;
    601 	} */ *uap = v;
    602 	struct proc	*p;
    603 	struct msghdr	msg;
    604 	struct iovec	aiov[UIO_SMALLIOV], *uiov, *iov;
    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 	p = l->l_proc;
    628 	if ((error = recvit(p, SCARG(uap, s), &msg, (caddr_t)0, retval)) == 0) {
    629 		msg.msg_iov = uiov;
    630 		error = copyout((caddr_t)&msg, (caddr_t)SCARG(uap, msg),
    631 		    sizeof(msg));
    632 	}
    633 done:
    634 	if (iov != aiov)
    635 		free(iov, M_IOV);
    636 	return (error);
    637 }
    638 
    639 int
    640 recvit(struct proc *p, int s, struct msghdr *mp, caddr_t namelenp,
    641 	register_t *retsize)
    642 {
    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 	from = 0;
    654 	control = 0;
    655 #ifdef KTRACE
    656 	ktriov = NULL;
    657 #endif
    658 
    659 	/* getsock() will use the descriptor for us */
    660 	if ((error = getsock(p->p_fd, s, &fp)) != 0)
    661 		return (error);
    662 	so = (struct socket *)fp->f_data;
    663 	auio.uio_iov = mp->msg_iov;
    664 	auio.uio_iovcnt = mp->msg_iovlen;
    665 	auio.uio_segflg = UIO_USERSPACE;
    666 	auio.uio_rw = UIO_READ;
    667 	auio.uio_procp = p;
    668 	auio.uio_offset = 0;			/* XXX */
    669 	auio.uio_resid = 0;
    670 	iov = mp->msg_iov;
    671 	for (i = 0; i < mp->msg_iovlen; i++, iov++) {
    672 #if 0
    673 		/* cannot happen iov_len is unsigned */
    674 		if (iov->iov_len < 0) {
    675 			error = EINVAL;
    676 			goto out1;
    677 		}
    678 #endif
    679 		/*
    680 		 * Reads return ssize_t because -1 is returned on error.
    681 		 * Therefore we must restrict the length to SSIZE_MAX to
    682 		 * avoid garbage return values.
    683 		 */
    684 		auio.uio_resid += iov->iov_len;
    685 		if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
    686 			error = EINVAL;
    687 			goto out1;
    688 		}
    689 	}
    690 #ifdef KTRACE
    691 	if (KTRPOINT(p, KTR_GENIO)) {
    692 		int iovlen = auio.uio_iovcnt * sizeof(struct iovec);
    693 
    694 		ktriov = malloc(iovlen, M_TEMP, M_WAITOK);
    695 		memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen);
    696 	}
    697 #endif
    698 	len = auio.uio_resid;
    699 	error = (*so->so_receive)(so, &from, &auio, NULL,
    700 			  mp->msg_control ? &control : NULL, &mp->msg_flags);
    701 	if (error) {
    702 		if (auio.uio_resid != len && (error == ERESTART ||
    703 		    error == EINTR || error == EWOULDBLOCK))
    704 			error = 0;
    705 	}
    706 #ifdef KTRACE
    707 	if (ktriov != NULL) {
    708 		if (error == 0)
    709 			ktrgenio(p, s, UIO_READ, ktriov,
    710 			    len - auio.uio_resid, error);
    711 		free(ktriov, M_TEMP);
    712 	}
    713 #endif
    714 	if (error)
    715 		goto out;
    716 	*retsize = len - auio.uio_resid;
    717 	if (mp->msg_name) {
    718 		len = mp->msg_namelen;
    719 		if (len <= 0 || from == 0)
    720 			len = 0;
    721 		else {
    722 			if (len > from->m_len)
    723 				len = from->m_len;
    724 			/* else if len < from->m_len ??? */
    725 			error = copyout(mtod(from, caddr_t),
    726 					(caddr_t)mp->msg_name, (unsigned)len);
    727 			if (error)
    728 				goto out;
    729 		}
    730 		mp->msg_namelen = len;
    731 		if (namelenp &&
    732 		    (error = copyout((caddr_t)&len, namelenp, sizeof(int))))
    733 			goto out;
    734 	}
    735 	if (mp->msg_control) {
    736 		len = mp->msg_controllen;
    737 		if (len <= 0 || control == 0)
    738 			len = 0;
    739 		else {
    740 			struct mbuf *m = control;
    741 			caddr_t p = (caddr_t)mp->msg_control;
    742 
    743 			do {
    744 				i = m->m_len;
    745 				if (len < i) {
    746 					mp->msg_flags |= MSG_CTRUNC;
    747 					i = len;
    748 				}
    749 				error = copyout(mtod(m, caddr_t), p,
    750 				    (unsigned)i);
    751 				if (m->m_next)
    752 					i = ALIGN(i);
    753 				p += i;
    754 				len -= i;
    755 				if (error != 0 || len <= 0)
    756 					break;
    757 			} while ((m = m->m_next) != NULL);
    758 			len = p - (caddr_t)mp->msg_control;
    759 		}
    760 		mp->msg_controllen = len;
    761 	}
    762  out:
    763 	if (from)
    764 		m_freem(from);
    765 	if (control)
    766 		m_freem(control);
    767  out1:
    768 	FILE_UNUSE(fp, p);
    769 	return (error);
    770 }
    771 
    772 /* ARGSUSED */
    773 int
    774 sys_shutdown(struct lwp *l, void *v, register_t *retval)
    775 {
    776 	struct sys_shutdown_args /* {
    777 		syscallarg(int)	s;
    778 		syscallarg(int)	how;
    779 	} */ *uap = v;
    780 	struct proc	*p;
    781 	struct file	*fp;
    782 	int		error;
    783 
    784 	p = l->l_proc;
    785 	/* getsock() will use the descriptor for us */
    786 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    787 		return (error);
    788 	error = soshutdown((struct socket *)fp->f_data, SCARG(uap, how));
    789 	FILE_UNUSE(fp, p);
    790 	return (error);
    791 }
    792 
    793 /* ARGSUSED */
    794 int
    795 sys_setsockopt(struct lwp *l, void *v, register_t *retval)
    796 {
    797 	struct sys_setsockopt_args /* {
    798 		syscallarg(int)			s;
    799 		syscallarg(int)			level;
    800 		syscallarg(int)			name;
    801 		syscallarg(const void *)	val;
    802 		syscallarg(unsigned int)	valsize;
    803 	} */ *uap = v;
    804 	struct proc	*p;
    805 	struct file	*fp;
    806 	struct mbuf	*m;
    807 	struct socket	*so;
    808 	int		error;
    809 	unsigned int	len;
    810 
    811 	p = l->l_proc;
    812 	m = NULL;
    813 	/* getsock() will use the descriptor for us */
    814 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    815 		return (error);
    816 	so = (struct socket *)fp->f_data;
    817 	len = SCARG(uap, valsize);
    818 	if (len > MCLBYTES) {
    819 		error = EINVAL;
    820 		goto out;
    821 	}
    822 	if (SCARG(uap, val)) {
    823 		m = m_get(M_WAIT, MT_SOOPTS);
    824 		MCLAIM(m, so->so_mowner);
    825 		if (len > MLEN)
    826 			m_clget(m, M_WAIT);
    827 		error = copyin(SCARG(uap, val), mtod(m, caddr_t), len);
    828 		if (error) {
    829 			(void) m_free(m);
    830 			goto out;
    831 		}
    832 		m->m_len = SCARG(uap, valsize);
    833 	}
    834 	error = sosetopt(so, SCARG(uap, level), SCARG(uap, name), m);
    835  out:
    836 	FILE_UNUSE(fp, p);
    837 	return (error);
    838 }
    839 
    840 /* ARGSUSED */
    841 int
    842 sys_getsockopt(struct lwp *l, void *v, register_t *retval)
    843 {
    844 	struct sys_getsockopt_args /* {
    845 		syscallarg(int)			s;
    846 		syscallarg(int)			level;
    847 		syscallarg(int)			name;
    848 		syscallarg(void *)		val;
    849 		syscallarg(unsigned int *)	avalsize;
    850 	} */ *uap = v;
    851 	struct proc	*p;
    852 	struct file	*fp;
    853 	struct mbuf	*m;
    854 	unsigned int	op, i, valsize;
    855 	int		error;
    856 
    857 	p = l->l_proc;
    858 	m = NULL;
    859 	/* getsock() will use the descriptor for us */
    860 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    861 		return (error);
    862 	if (SCARG(uap, val)) {
    863 		error = copyin((caddr_t)SCARG(uap, avalsize),
    864 			       (caddr_t)&valsize, sizeof(valsize));
    865 		if (error)
    866 			goto out;
    867 	} else
    868 		valsize = 0;
    869 	if ((error = sogetopt((struct socket *)fp->f_data, SCARG(uap, level),
    870 	    SCARG(uap, name), &m)) == 0 && SCARG(uap, val) && valsize &&
    871 	    m != NULL) {
    872 		op = 0;
    873 		while (m && !error && op < valsize) {
    874 			i = min(m->m_len, (valsize - op));
    875 			error = copyout(mtod(m, caddr_t), SCARG(uap, val), i);
    876 			op += i;
    877 			SCARG(uap, val) = ((u_int8_t *)SCARG(uap, val)) + i;
    878 			m = m_free(m);
    879 		}
    880 		valsize = op;
    881 		if (error == 0)
    882 			error = copyout(&valsize,
    883 					SCARG(uap, avalsize), sizeof(valsize));
    884 	}
    885 	if (m != NULL)
    886 		(void) m_freem(m);
    887  out:
    888 	FILE_UNUSE(fp, p);
    889 	return (error);
    890 }
    891 
    892 #ifdef PIPE_SOCKETPAIR
    893 /* ARGSUSED */
    894 int
    895 sys_pipe(struct lwp *l, void *v, register_t *retval)
    896 {
    897 	struct proc	*p;
    898 	struct filedesc	*fdp;
    899 	struct file	*rf, *wf;
    900 	struct socket	*rso, *wso;
    901 	int		fd, error;
    902 
    903 	p = l->l_proc;
    904 	fdp = p->p_fd;
    905 	if ((error = socreate(AF_LOCAL, &rso, SOCK_STREAM, 0)) != 0)
    906 		return (error);
    907 	if ((error = socreate(AF_LOCAL, &wso, SOCK_STREAM, 0)) != 0)
    908 		goto free1;
    909 	/* remember this socket pair implements a pipe */
    910 	wso->so_state |= SS_ISAPIPE;
    911 	rso->so_state |= SS_ISAPIPE;
    912 	/* falloc() will use the descriptor for us */
    913 	if ((error = falloc(p, &rf, &fd)) != 0)
    914 		goto free2;
    915 	retval[0] = fd;
    916 	rf->f_flag = FREAD;
    917 	rf->f_type = DTYPE_SOCKET;
    918 	rf->f_ops = &socketops;
    919 	rf->f_data = (caddr_t)rso;
    920 	if ((error = falloc(p, &wf, &fd)) != 0)
    921 		goto free3;
    922 	wf->f_flag = FWRITE;
    923 	wf->f_type = DTYPE_SOCKET;
    924 	wf->f_ops = &socketops;
    925 	wf->f_data = (caddr_t)wso;
    926 	retval[1] = fd;
    927 	if ((error = unp_connect2(wso, rso, PRU_CONNECT2)) != 0)
    928 		goto free4;
    929 	FILE_SET_MATURE(rf);
    930 	FILE_SET_MATURE(wf);
    931 	FILE_UNUSE(rf, p);
    932 	FILE_UNUSE(wf, p);
    933 	return (0);
    934  free4:
    935 	FILE_UNUSE(wf, p);
    936 	ffree(wf);
    937 	fdremove(fdp, retval[1]);
    938  free3:
    939 	FILE_UNUSE(rf, p);
    940 	ffree(rf);
    941 	fdremove(fdp, retval[0]);
    942  free2:
    943 	(void)soclose(wso);
    944  free1:
    945 	(void)soclose(rso);
    946 	return (error);
    947 }
    948 #endif /* PIPE_SOCKETPAIR */
    949 
    950 /*
    951  * Get socket name.
    952  */
    953 /* ARGSUSED */
    954 int
    955 sys_getsockname(struct lwp *l, void *v, register_t *retval)
    956 {
    957 	struct sys_getsockname_args /* {
    958 		syscallarg(int)			fdes;
    959 		syscallarg(struct sockaddr *)	asa;
    960 		syscallarg(unsigned int *)	alen;
    961 	} */ *uap = v;
    962 	struct proc	*p;
    963 	struct file	*fp;
    964 	struct socket	*so;
    965 	struct mbuf	*m;
    966 	unsigned int	len;
    967 	int		error;
    968 
    969 	p = l->l_proc;
    970 	/* getsock() will use the descriptor for us */
    971 	if ((error = getsock(p->p_fd, SCARG(uap, fdes), &fp)) != 0)
    972 		return (error);
    973 	error = copyin((caddr_t)SCARG(uap, alen), (caddr_t)&len, sizeof(len));
    974 	if (error)
    975 		goto out;
    976 	so = (struct socket *)fp->f_data;
    977 	m = m_getclr(M_WAIT, MT_SONAME);
    978 	MCLAIM(m, so->so_mowner);
    979 	error = (*so->so_proto->pr_usrreq)(so, PRU_SOCKADDR, (struct mbuf *)0,
    980 	    m, (struct mbuf *)0, (struct proc *)0);
    981 	if (error)
    982 		goto bad;
    983 	if (len > m->m_len)
    984 		len = m->m_len;
    985 	error = copyout(mtod(m, caddr_t), (caddr_t)SCARG(uap, asa), len);
    986 	if (error == 0)
    987 		error = copyout((caddr_t)&len, (caddr_t)SCARG(uap, alen),
    988 		    sizeof(len));
    989  bad:
    990 	m_freem(m);
    991  out:
    992 	FILE_UNUSE(fp, p);
    993 	return (error);
    994 }
    995 
    996 /*
    997  * Get name of peer for connected socket.
    998  */
    999 /* ARGSUSED */
   1000 int
   1001 sys_getpeername(struct lwp *l, void *v, register_t *retval)
   1002 {
   1003 	struct sys_getpeername_args /* {
   1004 		syscallarg(int)			fdes;
   1005 		syscallarg(struct sockaddr *)	asa;
   1006 		syscallarg(unsigned int *)	alen;
   1007 	} */ *uap = v;
   1008 	struct proc	*p;
   1009 	struct file	*fp;
   1010 	struct socket	*so;
   1011 	struct mbuf	*m;
   1012 	unsigned int	len;
   1013 	int		error;
   1014 
   1015 	p = l->l_proc;
   1016 	/* getsock() will use the descriptor for us */
   1017 	if ((error = getsock(p->p_fd, SCARG(uap, fdes), &fp)) != 0)
   1018 		return (error);
   1019 	so = (struct socket *)fp->f_data;
   1020 	if ((so->so_state & (SS_ISCONNECTED|SS_ISCONFIRMING)) == 0) {
   1021 		error = ENOTCONN;
   1022 		goto out;
   1023 	}
   1024 	error = copyin((caddr_t)SCARG(uap, alen), (caddr_t)&len, sizeof(len));
   1025 	if (error)
   1026 		goto out;
   1027 	m = m_getclr(M_WAIT, MT_SONAME);
   1028 	MCLAIM(m, so->so_mowner);
   1029 	error = (*so->so_proto->pr_usrreq)(so, PRU_PEERADDR, (struct mbuf *)0,
   1030 	    m, (struct mbuf *)0, (struct proc *)0);
   1031 	if (error)
   1032 		goto bad;
   1033 	if (len > m->m_len)
   1034 		len = m->m_len;
   1035 	error = copyout(mtod(m, caddr_t), (caddr_t)SCARG(uap, asa), len);
   1036 	if (error)
   1037 		goto bad;
   1038 	error = copyout((caddr_t)&len, (caddr_t)SCARG(uap, alen), sizeof(len));
   1039  bad:
   1040 	m_freem(m);
   1041  out:
   1042 	FILE_UNUSE(fp, p);
   1043 	return (error);
   1044 }
   1045 
   1046 /*
   1047  * XXX In a perfect world, we wouldn't pass around socket control
   1048  * XXX arguments in mbufs, and this could go away.
   1049  */
   1050 int
   1051 sockargs(struct mbuf **mp, const void *buf, size_t buflen, int type)
   1052 {
   1053 	struct sockaddr	*sa;
   1054 	struct mbuf	*m;
   1055 	int		error;
   1056 
   1057 	/*
   1058 	 * We can't allow socket names > UCHAR_MAX in length, since that
   1059 	 * will overflow sa_len.  Control data more than a page size in
   1060 	 * length is just too much.
   1061 	 */
   1062 	if (buflen > (type == MT_SONAME ? UCHAR_MAX : PAGE_SIZE))
   1063 		return (EINVAL);
   1064 
   1065 	/* Allocate an mbuf to hold the arguments. */
   1066 	m = m_get(M_WAIT, type);
   1067 	/* can't claim.  don't who to assign it to. */
   1068 	if (buflen > MLEN) {
   1069 		/*
   1070 		 * Won't fit into a regular mbuf, so we allocate just
   1071 		 * enough external storage to hold the argument.
   1072 		 */
   1073 		MEXTMALLOC(m, buflen, M_WAITOK);
   1074 	}
   1075 	m->m_len = buflen;
   1076 	error = copyin(buf, mtod(m, caddr_t), buflen);
   1077 	if (error) {
   1078 		(void) m_free(m);
   1079 		return (error);
   1080 	}
   1081 	*mp = m;
   1082 	if (type == MT_SONAME) {
   1083 		sa = mtod(m, struct sockaddr *);
   1084 #if BYTE_ORDER != BIG_ENDIAN
   1085 		/*
   1086 		 * 4.3BSD compat thing - need to stay, since bind(2),
   1087 		 * connect(2), sendto(2) were not versioned for COMPAT_43.
   1088 		 */
   1089 		if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
   1090 			sa->sa_family = sa->sa_len;
   1091 #endif
   1092 		sa->sa_len = buflen;
   1093 	}
   1094 	return (0);
   1095 }
   1096 
   1097 int
   1098 getsock(struct filedesc *fdp, int fdes, struct file **fpp)
   1099 {
   1100 	struct file	*fp;
   1101 
   1102 	if ((fp = fd_getfile(fdp, fdes)) == NULL)
   1103 		return (EBADF);
   1104 
   1105 	FILE_USE(fp);
   1106 
   1107 	if (fp->f_type != DTYPE_SOCKET) {
   1108 		FILE_UNUSE(fp, NULL);
   1109 		return (ENOTSOCK);
   1110 	}
   1111 	*fpp = fp;
   1112 	return (0);
   1113 }
   1114