Home | History | Annotate | Line # | Download | only in kern
uipc_syscalls.c revision 1.38
      1 /*	$NetBSD: uipc_syscalls.c,v 1.38 1998/08/04 19:48:35 kleink 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 
     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 
     57 #include <sys/mount.h>
     58 #include <sys/syscallargs.h>
     59 
     60 /*
     61  * System call interface to the socket abstraction.
     62  */
     63 extern	struct fileops socketops;
     64 
     65 int
     66 sys_socket(p, v, retval)
     67 	struct proc *p;
     68 	void *v;
     69 	register_t *retval;
     70 {
     71 	register struct sys_socket_args /* {
     72 		syscallarg(int) domain;
     73 		syscallarg(int) type;
     74 		syscallarg(int) protocol;
     75 	} */ *uap = v;
     76 	struct filedesc *fdp = p->p_fd;
     77 	struct socket *so;
     78 	struct file *fp;
     79 	int fd, error;
     80 
     81 	if ((error = falloc(p, &fp, &fd)) != 0)
     82 		return (error);
     83 	fp->f_flag = FREAD|FWRITE;
     84 	fp->f_type = DTYPE_SOCKET;
     85 	fp->f_ops = &socketops;
     86 	error = socreate(SCARG(uap, domain), &so, SCARG(uap, type),
     87 			 SCARG(uap, protocol));
     88 	if (error) {
     89 		fdp->fd_ofiles[fd] = 0;
     90 		ffree(fp);
     91 	} else {
     92 		fp->f_data = (caddr_t)so;
     93 		*retval = fd;
     94 	}
     95 	return (error);
     96 }
     97 
     98 /* ARGSUSED */
     99 int
    100 sys_bind(p, v, retval)
    101 	struct proc *p;
    102 	void *v;
    103 	register_t *retval;
    104 {
    105 	register struct sys_bind_args /* {
    106 		syscallarg(int) s;
    107 		syscallarg(const struct sockaddr *) name;
    108 		syscallarg(int) namelen;
    109 	} */ *uap = v;
    110 	struct file *fp;
    111 	struct mbuf *nam;
    112 	int error;
    113 
    114 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    115 		return (error);
    116 	error = sockargs(&nam, SCARG(uap, name), SCARG(uap, namelen),
    117 	    MT_SONAME);
    118 	if (error)
    119 		return (error);
    120 	error = sobind((struct socket *)fp->f_data, nam);
    121 	m_freem(nam);
    122 	return (error);
    123 }
    124 
    125 /* ARGSUSED */
    126 int
    127 sys_listen(p, v, retval)
    128 	struct proc *p;
    129 	void *v;
    130 	register_t *retval;
    131 {
    132 	register struct sys_listen_args /* {
    133 		syscallarg(int) s;
    134 		syscallarg(int) backlog;
    135 	} */ *uap = v;
    136 	struct file *fp;
    137 	int error;
    138 
    139 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    140 		return (error);
    141 	return (solisten((struct socket *)fp->f_data, SCARG(uap, backlog)));
    142 }
    143 
    144 int
    145 sys_accept(p, v, retval)
    146 	struct proc *p;
    147 	void *v;
    148 	register_t *retval;
    149 {
    150 	register struct sys_accept_args /* {
    151 		syscallarg(int) s;
    152 		syscallarg(struct sockaddr *) name;
    153 		syscallarg(int *) anamelen;
    154 	} */ *uap = v;
    155 	struct file *fp;
    156 	struct mbuf *nam;
    157 	int namelen, error, s, tmpfd;
    158 	register struct socket *so;
    159 
    160 	if (SCARG(uap, name) && (error = copyin((caddr_t)SCARG(uap, anamelen),
    161 	    (caddr_t)&namelen, sizeof(namelen))))
    162 		return (error);
    163 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    164 		return (error);
    165 	s = splsoftnet();
    166 	so = (struct socket *)fp->f_data;
    167 	if ((so->so_options & SO_ACCEPTCONN) == 0) {
    168 		splx(s);
    169 		return (EINVAL);
    170 	}
    171 	if ((so->so_state & SS_NBIO) && so->so_qlen == 0) {
    172 		splx(s);
    173 		return (EWOULDBLOCK);
    174 	}
    175 	while (so->so_qlen == 0 && so->so_error == 0) {
    176 		if (so->so_state & SS_CANTRCVMORE) {
    177 			so->so_error = ECONNABORTED;
    178 			break;
    179 		}
    180 		error = tsleep((caddr_t)&so->so_timeo, PSOCK | PCATCH,
    181 			       netcon, 0);
    182 		if (error) {
    183 			splx(s);
    184 			return (error);
    185 		}
    186 	}
    187 	if (so->so_error) {
    188 		error = so->so_error;
    189 		so->so_error = 0;
    190 		splx(s);
    191 		return (error);
    192 	}
    193 	if ((error = falloc(p, &fp, &tmpfd)) != 0) {
    194 		splx(s);
    195 		return (error);
    196 	}
    197 	*retval = tmpfd;
    198 	{ struct socket *aso = so->so_q.tqh_first;
    199 	  if (soqremque(aso, 1) == 0)
    200 		panic("accept");
    201 	  so = aso;
    202 	}
    203 	fp->f_type = DTYPE_SOCKET;
    204 	fp->f_flag = FREAD|FWRITE;
    205 	fp->f_ops = &socketops;
    206 	fp->f_data = (caddr_t)so;
    207 	nam = m_get(M_WAIT, MT_SONAME);
    208 	(void) soaccept(so, nam);
    209 	if (SCARG(uap, name)) {
    210 		if (namelen > nam->m_len)
    211 			namelen = nam->m_len;
    212 		/* SHOULD COPY OUT A CHAIN HERE */
    213 		if ((error = copyout(mtod(nam, caddr_t),
    214 		    (caddr_t)SCARG(uap, name), (u_int)namelen)) == 0)
    215 			error = copyout((caddr_t)&namelen,
    216 			    (caddr_t)SCARG(uap, anamelen),
    217 			    sizeof(*SCARG(uap, anamelen)));
    218 	}
    219 	m_freem(nam);
    220 	splx(s);
    221 	return (error);
    222 }
    223 
    224 /* ARGSUSED */
    225 int
    226 sys_connect(p, v, retval)
    227 	struct proc *p;
    228 	void *v;
    229 	register_t *retval;
    230 {
    231 	register struct sys_connect_args /* {
    232 		syscallarg(int) s;
    233 		syscallarg(const struct sockaddr *) name;
    234 		syscallarg(int) namelen;
    235 	} */ *uap = v;
    236 	struct file *fp;
    237 	register struct socket *so;
    238 	struct mbuf *nam;
    239 	int error, s;
    240 
    241 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    242 		return (error);
    243 	so = (struct socket *)fp->f_data;
    244 	if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING))
    245 		return (EALREADY);
    246 	error = sockargs(&nam, SCARG(uap, name), SCARG(uap, namelen),
    247 	    MT_SONAME);
    248 	if (error)
    249 		return (error);
    250 	error = soconnect(so, nam);
    251 	if (error)
    252 		goto bad;
    253 	if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING)) {
    254 		m_freem(nam);
    255 		return (EINPROGRESS);
    256 	}
    257 	s = splsoftnet();
    258 	while ((so->so_state & SS_ISCONNECTING) && so->so_error == 0) {
    259 		error = tsleep((caddr_t)&so->so_timeo, PSOCK | PCATCH,
    260 			       netcon, 0);
    261 		if (error)
    262 			break;
    263 	}
    264 	if (error == 0) {
    265 		error = so->so_error;
    266 		so->so_error = 0;
    267 	}
    268 	splx(s);
    269 bad:
    270 	so->so_state &= ~SS_ISCONNECTING;
    271 	m_freem(nam);
    272 	if (error == ERESTART)
    273 		error = EINTR;
    274 	return (error);
    275 }
    276 
    277 int
    278 sys_socketpair(p, v, retval)
    279 	struct proc *p;
    280 	void *v;
    281 	register_t *retval;
    282 {
    283 	register struct sys_socketpair_args /* {
    284 		syscallarg(int) domain;
    285 		syscallarg(int) type;
    286 		syscallarg(int) protocol;
    287 		syscallarg(int *) rsv;
    288 	} */ *uap = v;
    289 	register struct filedesc *fdp = p->p_fd;
    290 	struct file *fp1, *fp2;
    291 	struct socket *so1, *so2;
    292 	int fd, error, sv[2];
    293 
    294 	error = socreate(SCARG(uap, domain), &so1, SCARG(uap, type),
    295 			 SCARG(uap, protocol));
    296 	if (error)
    297 		return (error);
    298 	error = socreate(SCARG(uap, domain), &so2, SCARG(uap, type),
    299 			 SCARG(uap, protocol));
    300 	if (error)
    301 		goto free1;
    302 	if ((error = falloc(p, &fp1, &fd)) != 0)
    303 		goto free2;
    304 	sv[0] = fd;
    305 	fp1->f_flag = FREAD|FWRITE;
    306 	fp1->f_type = DTYPE_SOCKET;
    307 	fp1->f_ops = &socketops;
    308 	fp1->f_data = (caddr_t)so1;
    309 	if ((error = falloc(p, &fp2, &fd)) != 0)
    310 		goto free3;
    311 	fp2->f_flag = FREAD|FWRITE;
    312 	fp2->f_type = DTYPE_SOCKET;
    313 	fp2->f_ops = &socketops;
    314 	fp2->f_data = (caddr_t)so2;
    315 	sv[1] = fd;
    316 	if ((error = soconnect2(so1, so2)) != 0)
    317 		goto free4;
    318 	if (SCARG(uap, type) == SOCK_DGRAM) {
    319 		/*
    320 		 * Datagram socket connection is asymmetric.
    321 		 */
    322 		 if ((error = soconnect2(so2, so1)) != 0)
    323 			goto free4;
    324 	}
    325 	error = copyout((caddr_t)sv, (caddr_t)SCARG(uap, rsv),
    326 	    2 * sizeof(int));
    327 	return (error);
    328 free4:
    329 	ffree(fp2);
    330 	fdp->fd_ofiles[sv[1]] = 0;
    331 free3:
    332 	ffree(fp1);
    333 	fdp->fd_ofiles[sv[0]] = 0;
    334 free2:
    335 	(void)soclose(so2);
    336 free1:
    337 	(void)soclose(so1);
    338 	return (error);
    339 }
    340 
    341 int
    342 sys_sendto(p, v, retval)
    343 	struct proc *p;
    344 	void *v;
    345 	register_t *retval;
    346 {
    347 	register struct sys_sendto_args /* {
    348 		syscallarg(int) s;
    349 		syscallarg(const void *) buf;
    350 		syscallarg(size_t) len;
    351 		syscallarg(int) flags;
    352 		syscallarg(const struct sockaddr *) to;
    353 		syscallarg(int) tolen;
    354 	} */ *uap = v;
    355 	struct msghdr msg;
    356 	struct iovec aiov;
    357 
    358 	msg.msg_name = (caddr_t)SCARG(uap, to);		/* XXX kills const */
    359 	msg.msg_namelen = SCARG(uap, tolen);
    360 	msg.msg_iov = &aiov;
    361 	msg.msg_iovlen = 1;
    362 	msg.msg_control = 0;
    363 #ifdef COMPAT_OLDSOCK
    364 	msg.msg_flags = 0;
    365 #endif
    366 	aiov.iov_base = (char *)SCARG(uap, buf);	/* XXX kills const */
    367 	aiov.iov_len = SCARG(uap, len);
    368 	return (sendit(p, SCARG(uap, s), &msg, SCARG(uap, flags), retval));
    369 }
    370 
    371 int
    372 sys_sendmsg(p, v, retval)
    373 	struct proc *p;
    374 	void *v;
    375 	register_t *retval;
    376 {
    377 	register struct sys_sendmsg_args /* {
    378 		syscallarg(int) s;
    379 		syscallarg(const struct msghdr *) msg;
    380 		syscallarg(int) flags;
    381 	} */ *uap = v;
    382 	struct msghdr msg;
    383 	struct iovec aiov[UIO_SMALLIOV], *iov;
    384 	int error;
    385 
    386 	error = copyin(SCARG(uap, msg), (caddr_t)&msg, sizeof(msg));
    387 	if (error)
    388 		return (error);
    389 	if ((u_int)msg.msg_iovlen > UIO_SMALLIOV) {
    390 		if ((u_int)msg.msg_iovlen > IOV_MAX)
    391 			return (EMSGSIZE);
    392 		MALLOC(iov, struct iovec *,
    393 		       sizeof(struct iovec) * (u_int)msg.msg_iovlen, M_IOV,
    394 		       M_WAITOK);
    395 	} else if ((u_int)msg.msg_iovlen > 0)
    396 		iov = aiov;
    397 	else
    398 		return (EMSGSIZE);
    399 	error = copyin((caddr_t)msg.msg_iov, (caddr_t)iov,
    400 	    (size_t)(msg.msg_iovlen * sizeof(struct iovec)));
    401 	if (error)
    402 		goto done;
    403 	msg.msg_iov = iov;
    404 #ifdef COMPAT_OLDSOCK
    405 	msg.msg_flags = 0;
    406 #endif
    407 	error = sendit(p, SCARG(uap, s), &msg, SCARG(uap, flags), retval);
    408 done:
    409 	if (iov != aiov)
    410 		FREE(iov, M_IOV);
    411 	return (error);
    412 }
    413 
    414 int
    415 sendit(p, s, mp, flags, retsize)
    416 	register struct proc *p;
    417 	int s;
    418 	register struct msghdr *mp;
    419 	int flags;
    420 	register_t *retsize;
    421 {
    422 	struct file *fp;
    423 	struct uio auio;
    424 	register struct iovec *iov;
    425 	register int i;
    426 	struct mbuf *to, *control;
    427 	int len, error;
    428 	struct socket *so;
    429 #ifdef KTRACE
    430 	struct iovec *ktriov = NULL;
    431 #endif
    432 
    433 	if ((error = getsock(p->p_fd, s, &fp)) != 0)
    434 		return (error);
    435 	auio.uio_iov = mp->msg_iov;
    436 	auio.uio_iovcnt = mp->msg_iovlen;
    437 	auio.uio_segflg = UIO_USERSPACE;
    438 	auio.uio_rw = UIO_WRITE;
    439 	auio.uio_procp = p;
    440 	auio.uio_offset = 0;			/* XXX */
    441 	auio.uio_resid = 0;
    442 	iov = mp->msg_iov;
    443 	for (i = 0; i < mp->msg_iovlen; i++, iov++) {
    444 #if 0
    445 		/* cannot happen; iov_len is unsigned */
    446 		if (iov->iov_len < 0)
    447 			return (EINVAL);
    448 #endif
    449 		/*
    450 		 * Writes return ssize_t because -1 is returned on error.
    451 		 * Therefore, we must restrict the length to SSIZE_MAX to
    452 		 * avoid garbage return values.
    453 		 */
    454 		auio.uio_resid += iov->iov_len;
    455 		if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX)
    456 			return (EINVAL);
    457 	}
    458 	if (mp->msg_name) {
    459 		error = sockargs(&to, mp->msg_name, mp->msg_namelen,
    460 				 MT_SONAME);
    461 		if (error)
    462 			return (error);
    463 	} else
    464 		to = 0;
    465 	if (mp->msg_control) {
    466 		if (mp->msg_controllen < sizeof(struct cmsghdr)
    467 #ifdef COMPAT_OLDSOCK
    468 		    && mp->msg_flags != MSG_COMPAT
    469 #endif
    470 		) {
    471 			error = EINVAL;
    472 			goto bad;
    473 		}
    474 		error = sockargs(&control, mp->msg_control,
    475 				 mp->msg_controllen, MT_CONTROL);
    476 		if (error)
    477 			goto bad;
    478 #ifdef COMPAT_OLDSOCK
    479 		if (mp->msg_flags == MSG_COMPAT) {
    480 			register struct cmsghdr *cm;
    481 
    482 			M_PREPEND(control, sizeof(*cm), M_WAIT);
    483 			if (control == 0) {
    484 				error = ENOBUFS;
    485 				goto bad;
    486 			} else {
    487 				cm = mtod(control, struct cmsghdr *);
    488 				cm->cmsg_len = control->m_len;
    489 				cm->cmsg_level = SOL_SOCKET;
    490 				cm->cmsg_type = SCM_RIGHTS;
    491 			}
    492 		}
    493 #endif
    494 	} else
    495 		control = 0;
    496 #ifdef KTRACE
    497 	if (KTRPOINT(p, KTR_GENIO)) {
    498 		int iovlen = auio.uio_iovcnt * sizeof(struct iovec);
    499 
    500 		MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK);
    501 		memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen);
    502 	}
    503 #endif
    504 	len = auio.uio_resid;
    505 	so = (struct socket *)fp->f_data;
    506 	error = (*so->so_send)(so, to, &auio, NULL, control, flags);
    507 	if (error) {
    508 		if (auio.uio_resid != len && (error == ERESTART ||
    509 		    error == EINTR || error == EWOULDBLOCK))
    510 			error = 0;
    511 		if (error == EPIPE)
    512 			psignal(p, SIGPIPE);
    513 	}
    514 	if (error == 0)
    515 		*retsize = len - auio.uio_resid;
    516 #ifdef KTRACE
    517 	if (ktriov != NULL) {
    518 		if (error == 0)
    519 			ktrgenio(p->p_tracep, s, UIO_WRITE,
    520 				ktriov, *retsize, error);
    521 		FREE(ktriov, M_TEMP);
    522 	}
    523 #endif
    524 bad:
    525 	if (to)
    526 		m_freem(to);
    527 	return (error);
    528 }
    529 
    530 int
    531 sys_recvfrom(p, v, retval)
    532 	struct proc *p;
    533 	void *v;
    534 	register_t *retval;
    535 {
    536 	register struct sys_recvfrom_args /* {
    537 		syscallarg(int) s;
    538 		syscallarg(void *) buf;
    539 		syscallarg(size_t) len;
    540 		syscallarg(int) flags;
    541 		syscallarg(struct sockaddr *) from;
    542 		syscallarg(int *) fromlenaddr;
    543 	} */ *uap = v;
    544 	struct msghdr msg;
    545 	struct iovec aiov;
    546 	int error;
    547 
    548 	if (SCARG(uap, fromlenaddr)) {
    549 		error = copyin((caddr_t)SCARG(uap, fromlenaddr),
    550 			       (caddr_t)&msg.msg_namelen,
    551 			       sizeof(msg.msg_namelen));
    552 		if (error)
    553 			return (error);
    554 	} else
    555 		msg.msg_namelen = 0;
    556 	msg.msg_name = (caddr_t)SCARG(uap, from);
    557 	msg.msg_iov = &aiov;
    558 	msg.msg_iovlen = 1;
    559 	aiov.iov_base = SCARG(uap, buf);
    560 	aiov.iov_len = SCARG(uap, len);
    561 	msg.msg_control = 0;
    562 	msg.msg_flags = SCARG(uap, flags);
    563 	return (recvit(p, SCARG(uap, s), &msg,
    564 		       (caddr_t)SCARG(uap, fromlenaddr), retval));
    565 }
    566 
    567 int
    568 sys_recvmsg(p, v, retval)
    569 	struct proc *p;
    570 	void *v;
    571 	register_t *retval;
    572 {
    573 	register struct sys_recvmsg_args /* {
    574 		syscallarg(int) s;
    575 		syscallarg(struct msghdr *) msg;
    576 		syscallarg(int) flags;
    577 	} */ *uap = v;
    578 	struct msghdr msg;
    579 	struct iovec aiov[UIO_SMALLIOV], *uiov, *iov;
    580 	register int error;
    581 
    582 	error = copyin((caddr_t)SCARG(uap, msg), (caddr_t)&msg,
    583 		       sizeof(msg));
    584 	if (error)
    585 		return (error);
    586 	if ((u_int)msg.msg_iovlen > UIO_SMALLIOV) {
    587 		if ((u_int)msg.msg_iovlen > IOV_MAX)
    588 			return (EMSGSIZE);
    589 		MALLOC(iov, struct iovec *,
    590 		       sizeof(struct iovec) * (u_int)msg.msg_iovlen, M_IOV,
    591 		       M_WAITOK);
    592 	} else if ((u_int)msg.msg_iovlen > 0)
    593 		iov = aiov;
    594 	else
    595 		return (EMSGSIZE);
    596 #ifdef COMPAT_OLDSOCK
    597 	msg.msg_flags = SCARG(uap, flags) &~ MSG_COMPAT;
    598 #else
    599 	msg.msg_flags = SCARG(uap, flags);
    600 #endif
    601 	uiov = msg.msg_iov;
    602 	msg.msg_iov = iov;
    603 	error = copyin((caddr_t)uiov, (caddr_t)iov,
    604 	    (size_t)(msg.msg_iovlen * sizeof(struct iovec)));
    605 	if (error)
    606 		goto done;
    607 	if ((error = recvit(p, SCARG(uap, s), &msg, (caddr_t)0, retval)) == 0) {
    608 		msg.msg_iov = uiov;
    609 		error = copyout((caddr_t)&msg, (caddr_t)SCARG(uap, msg),
    610 		    sizeof(msg));
    611 	}
    612 done:
    613 	if (iov != aiov)
    614 		FREE(iov, M_IOV);
    615 	return (error);
    616 }
    617 
    618 int
    619 recvit(p, s, mp, namelenp, retsize)
    620 	register struct proc *p;
    621 	int s;
    622 	register struct msghdr *mp;
    623 	caddr_t namelenp;
    624 	register_t *retsize;
    625 {
    626 	struct file *fp;
    627 	struct uio auio;
    628 	register struct iovec *iov;
    629 	register int i;
    630 	int len, error;
    631 	struct mbuf *from = 0, *control = 0;
    632 	struct socket *so;
    633 #ifdef KTRACE
    634 	struct iovec *ktriov = NULL;
    635 #endif
    636 
    637 	if ((error = getsock(p->p_fd, s, &fp)) != 0)
    638 		return (error);
    639 	auio.uio_iov = mp->msg_iov;
    640 	auio.uio_iovcnt = mp->msg_iovlen;
    641 	auio.uio_segflg = UIO_USERSPACE;
    642 	auio.uio_rw = UIO_READ;
    643 	auio.uio_procp = p;
    644 	auio.uio_offset = 0;			/* XXX */
    645 	auio.uio_resid = 0;
    646 	iov = mp->msg_iov;
    647 	for (i = 0; i < mp->msg_iovlen; i++, iov++) {
    648 #if 0
    649 		/* cannot happen iov_len is unsigned */
    650 		if (iov->iov_len < 0)
    651 			return (EINVAL);
    652 #endif
    653 		/*
    654 		 * Reads return ssize_t because -1 is returned on error.
    655 		 * Therefore we must restrict the length to SSIZE_MAX to
    656 		 * avoid garbage return values.
    657 		 */
    658 		auio.uio_resid += iov->iov_len;
    659 		if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX)
    660 			return (EINVAL);
    661 	}
    662 #ifdef KTRACE
    663 	if (KTRPOINT(p, KTR_GENIO)) {
    664 		int iovlen = auio.uio_iovcnt * sizeof(struct iovec);
    665 
    666 		MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK);
    667 		memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen);
    668 	}
    669 #endif
    670 	len = auio.uio_resid;
    671 	so = (struct socket *)fp->f_data;
    672 	error = (*so->so_receive)(so, &from, &auio, NULL,
    673 			  mp->msg_control ? &control : NULL, &mp->msg_flags);
    674 	if (error) {
    675 		if (auio.uio_resid != len && (error == ERESTART ||
    676 		    error == EINTR || error == EWOULDBLOCK))
    677 			error = 0;
    678 	}
    679 #ifdef KTRACE
    680 	if (ktriov != NULL) {
    681 		if (error == 0)
    682 			ktrgenio(p->p_tracep, s, UIO_READ,
    683 				ktriov, len - auio.uio_resid, error);
    684 		FREE(ktriov, M_TEMP);
    685 	}
    686 #endif
    687 	if (error)
    688 		goto out;
    689 	*retsize = len - auio.uio_resid;
    690 	if (mp->msg_name) {
    691 		len = mp->msg_namelen;
    692 		if (len <= 0 || from == 0)
    693 			len = 0;
    694 		else {
    695 #ifdef COMPAT_OLDSOCK
    696 			if (mp->msg_flags & MSG_COMPAT)
    697 				mtod(from, struct osockaddr *)->sa_family =
    698 				    mtod(from, struct sockaddr *)->sa_family;
    699 #endif
    700 			if (len > from->m_len)
    701 				len = from->m_len;
    702 			/* else if len < from->m_len ??? */
    703 			error = copyout(mtod(from, caddr_t),
    704 					(caddr_t)mp->msg_name, (unsigned)len);
    705 			if (error)
    706 				goto out;
    707 		}
    708 		mp->msg_namelen = len;
    709 		if (namelenp &&
    710 		    (error = copyout((caddr_t)&len, namelenp, sizeof(int)))) {
    711 #ifdef COMPAT_OLDSOCK
    712 			if (mp->msg_flags & MSG_COMPAT)
    713 				error = 0;	/* old recvfrom didn't check */
    714 			else
    715 #endif
    716 			goto out;
    717 		}
    718 	}
    719 	if (mp->msg_control) {
    720 #ifdef COMPAT_OLDSOCK
    721 		/*
    722 		 * We assume that old recvmsg calls won't receive access
    723 		 * rights and other control info, esp. as control info
    724 		 * is always optional and those options didn't exist in 4.3.
    725 		 * If we receive rights, trim the cmsghdr; anything else
    726 		 * is tossed.
    727 		 */
    728 		if (control && mp->msg_flags & MSG_COMPAT) {
    729 			if (mtod(control, struct cmsghdr *)->cmsg_level !=
    730 			    SOL_SOCKET ||
    731 			    mtod(control, struct cmsghdr *)->cmsg_type !=
    732 			    SCM_RIGHTS) {
    733 				mp->msg_controllen = 0;
    734 				goto out;
    735 			}
    736 			control->m_len -= sizeof(struct cmsghdr);
    737 			control->m_data += sizeof(struct cmsghdr);
    738 		}
    739 #endif
    740 		len = mp->msg_controllen;
    741 		if (len <= 0 || control == 0)
    742 			len = 0;
    743 		else {
    744 			struct mbuf *m = control;
    745 			caddr_t p = (caddr_t)mp->msg_control;
    746 
    747 			do {
    748 				i = m->m_len;
    749 				if (len < i) {
    750 					mp->msg_flags |= MSG_CTRUNC;
    751 					i = len;
    752 				}
    753 				error = copyout(mtod(m, caddr_t), p,
    754 				    (unsigned)i);
    755 				if (m->m_next)
    756 					i = ALIGN(i);
    757 				p += i;
    758 				len -= i;
    759 				if (error != 0 || len <= 0)
    760 					break;
    761 			} while ((m = m->m_next) != NULL);
    762 			len = p - (caddr_t)mp->msg_control;
    763 		}
    764 		mp->msg_controllen = len;
    765 	}
    766 out:
    767 	if (from)
    768 		m_freem(from);
    769 	if (control)
    770 		m_freem(control);
    771 	return (error);
    772 }
    773 
    774 /* ARGSUSED */
    775 int
    776 sys_shutdown(p, v, retval)
    777 	struct proc *p;
    778 	void *v;
    779 	register_t *retval;
    780 {
    781 	register struct sys_shutdown_args /* {
    782 		syscallarg(int) s;
    783 		syscallarg(int) how;
    784 	} */ *uap = v;
    785 	struct file *fp;
    786 	int error;
    787 
    788 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    789 		return (error);
    790 	return (soshutdown((struct socket *)fp->f_data, SCARG(uap, how)));
    791 }
    792 
    793 /* ARGSUSED */
    794 int
    795 sys_setsockopt(p, v, retval)
    796 	struct proc *p;
    797 	void *v;
    798 	register_t *retval;
    799 {
    800 	register struct sys_setsockopt_args /* {
    801 		syscallarg(int) s;
    802 		syscallarg(int) level;
    803 		syscallarg(int) name;
    804 		syscallarg(const void *) val;
    805 		syscallarg(int) valsize;
    806 	} */ *uap = v;
    807 	struct file *fp;
    808 	struct mbuf *m = NULL;
    809 	int error;
    810 
    811 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    812 		return (error);
    813 	if (SCARG(uap, valsize) > MLEN)
    814 		return (EINVAL);
    815 	if (SCARG(uap, val)) {
    816 		m = m_get(M_WAIT, MT_SOOPTS);
    817 		error = copyin(SCARG(uap, val), mtod(m, caddr_t),
    818 			       (u_int)SCARG(uap, valsize));
    819 		if (error) {
    820 			(void) m_free(m);
    821 			return (error);
    822 		}
    823 		m->m_len = SCARG(uap, valsize);
    824 	}
    825 	return (sosetopt((struct socket *)fp->f_data, SCARG(uap, level),
    826 			 SCARG(uap, name), m));
    827 }
    828 
    829 /* ARGSUSED */
    830 int
    831 sys_getsockopt(p, v, retval)
    832 	struct proc *p;
    833 	void *v;
    834 	register_t *retval;
    835 {
    836 	register struct sys_getsockopt_args /* {
    837 		syscallarg(int) s;
    838 		syscallarg(int) level;
    839 		syscallarg(int) name;
    840 		syscallarg(void *) val;
    841 		syscallarg(int *) avalsize;
    842 	} */ *uap = v;
    843 	struct file *fp;
    844 	struct mbuf *m = NULL;
    845 	int valsize, error;
    846 
    847 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    848 		return (error);
    849 	if (SCARG(uap, val)) {
    850 		error = copyin((caddr_t)SCARG(uap, avalsize),
    851 			       (caddr_t)&valsize, sizeof(valsize));
    852 		if (error)
    853 			return (error);
    854 	} else
    855 		valsize = 0;
    856 	if ((error = sogetopt((struct socket *)fp->f_data, SCARG(uap, level),
    857 	    SCARG(uap, name), &m)) == 0 && SCARG(uap, val) && valsize &&
    858 	    m != NULL) {
    859 		if (valsize > m->m_len)
    860 			valsize = m->m_len;
    861 		error = copyout(mtod(m, caddr_t), SCARG(uap, val),
    862 		    (u_int)valsize);
    863 		if (error == 0)
    864 			error = copyout((caddr_t)&valsize,
    865 			    (caddr_t)SCARG(uap, avalsize), sizeof(valsize));
    866 	}
    867 	if (m != NULL)
    868 		(void) m_free(m);
    869 	return (error);
    870 }
    871 
    872 /* ARGSUSED */
    873 int
    874 sys_pipe(p, v, retval)
    875 	struct proc *p;
    876 	void *v;
    877 	register_t *retval;
    878 {
    879 	register struct filedesc *fdp = p->p_fd;
    880 	struct file *rf, *wf;
    881 	struct socket *rso, *wso;
    882 	int fd, error;
    883 
    884 	if ((error = socreate(AF_LOCAL, &rso, SOCK_STREAM, 0)) != 0)
    885 		return (error);
    886 	if ((error = socreate(AF_LOCAL, &wso, SOCK_STREAM, 0)) != 0)
    887 		goto free1;
    888 	if ((error = falloc(p, &rf, &fd)) != 0)
    889 		goto free2;
    890 	retval[0] = fd;
    891 	rf->f_flag = FREAD;
    892 	rf->f_type = DTYPE_SOCKET;
    893 	rf->f_ops = &socketops;
    894 	rf->f_data = (caddr_t)rso;
    895 	if ((error = falloc(p, &wf, &fd)) != 0)
    896 		goto free3;
    897 	wf->f_flag = FWRITE;
    898 	wf->f_type = DTYPE_SOCKET;
    899 	wf->f_ops = &socketops;
    900 	wf->f_data = (caddr_t)wso;
    901 	retval[1] = fd;
    902 	if ((error = unp_connect2(wso, rso)) != 0)
    903 		goto free4;
    904 	return (0);
    905 free4:
    906 	ffree(wf);
    907 	fdp->fd_ofiles[retval[1]] = 0;
    908 free3:
    909 	ffree(rf);
    910 	fdp->fd_ofiles[retval[0]] = 0;
    911 free2:
    912 	(void)soclose(wso);
    913 free1:
    914 	(void)soclose(rso);
    915 	return (error);
    916 }
    917 
    918 /*
    919  * Get socket name.
    920  */
    921 /* ARGSUSED */
    922 int
    923 sys_getsockname(p, v, retval)
    924 	struct proc *p;
    925 	void *v;
    926 	register_t *retval;
    927 {
    928 	register struct sys_getsockname_args /* {
    929 		syscallarg(int) fdes;
    930 		syscallarg(struct sockaddr *) asa;
    931 		syscallarg(int *) alen;
    932 	} */ *uap = v;
    933 	struct file *fp;
    934 	register struct socket *so;
    935 	struct mbuf *m;
    936 	int len, error;
    937 
    938 	if ((error = getsock(p->p_fd, SCARG(uap, fdes), &fp)) != 0)
    939 		return (error);
    940 	error = copyin((caddr_t)SCARG(uap, alen), (caddr_t)&len, sizeof(len));
    941 	if (error)
    942 		return (error);
    943 	so = (struct socket *)fp->f_data;
    944 	m = m_getclr(M_WAIT, MT_SONAME);
    945 	error = (*so->so_proto->pr_usrreq)(so, PRU_SOCKADDR, (struct mbuf *)0,
    946 	    m, (struct mbuf *)0, (struct proc *)0);
    947 	if (error)
    948 		goto bad;
    949 	if (len > m->m_len)
    950 		len = m->m_len;
    951 	error = copyout(mtod(m, caddr_t), (caddr_t)SCARG(uap, asa), (u_int)len);
    952 	if (error == 0)
    953 		error = copyout((caddr_t)&len, (caddr_t)SCARG(uap, alen),
    954 		    sizeof(len));
    955 bad:
    956 	m_freem(m);
    957 	return (error);
    958 }
    959 
    960 /*
    961  * Get name of peer for connected socket.
    962  */
    963 /* ARGSUSED */
    964 int
    965 sys_getpeername(p, v, retval)
    966 	struct proc *p;
    967 	void *v;
    968 	register_t *retval;
    969 {
    970 	register struct sys_getpeername_args /* {
    971 		syscallarg(int) fdes;
    972 		syscallarg(struct sockaddr *) asa;
    973 		syscallarg(int *) alen;
    974 	} */ *uap = v;
    975 	struct file *fp;
    976 	register struct socket *so;
    977 	struct mbuf *m;
    978 	int len, error;
    979 
    980 	if ((error = getsock(p->p_fd, SCARG(uap, fdes), &fp)) != 0)
    981 		return (error);
    982 	so = (struct socket *)fp->f_data;
    983 	if ((so->so_state & (SS_ISCONNECTED|SS_ISCONFIRMING)) == 0)
    984 		return (ENOTCONN);
    985 	error = copyin((caddr_t)SCARG(uap, alen), (caddr_t)&len, sizeof(len));
    986 	if (error)
    987 		return (error);
    988 	m = m_getclr(M_WAIT, MT_SONAME);
    989 	error = (*so->so_proto->pr_usrreq)(so, PRU_PEERADDR, (struct mbuf *)0,
    990 	    m, (struct mbuf *)0, (struct proc *)0);
    991 	if (error)
    992 		goto bad;
    993 	if (len > m->m_len)
    994 		len = m->m_len;
    995 	error = copyout(mtod(m, caddr_t), (caddr_t)SCARG(uap, asa), (u_int)len);
    996 	if (error)
    997 		goto bad;
    998 	error = copyout((caddr_t)&len, (caddr_t)SCARG(uap, alen), sizeof(len));
    999 bad:
   1000 	m_freem(m);
   1001 	return (error);
   1002 }
   1003 
   1004 /*
   1005  * XXX In a perfect world, we wouldn't pass around socket control
   1006  * XXX arguments in mbufs, and this could go away.
   1007  */
   1008 int
   1009 sockargs(mp, buf, buflen, type)
   1010 	struct mbuf **mp;
   1011 	const void *buf;
   1012 	int buflen, type;
   1013 {
   1014 	register struct sockaddr *sa;
   1015 	register struct mbuf *m;
   1016 	int error;
   1017 
   1018 	/*
   1019 	 * We can't allow socket names > UCHAR_MAX in length, since that
   1020 	 * will overflow sa_len.
   1021 	 */
   1022 	if (type == MT_SONAME && (u_int)buflen > UCHAR_MAX)
   1023 		return (EINVAL);
   1024 
   1025 	/* Allocate an mbuf to hold the arguments. */
   1026 	m = m_get(M_WAIT, type);
   1027 	if ((u_int)buflen > MLEN) {
   1028 		/*
   1029 		 * Won't fit into a regular mbuf, so we allocate just
   1030 		 * enough external storage to hold the argument.
   1031 		 */
   1032 		MEXTMALLOC(m, buflen, M_WAITOK);
   1033 	}
   1034 	m->m_len = buflen;
   1035 	error = copyin(buf, mtod(m, caddr_t), (u_int)buflen);
   1036 	if (error) {
   1037 		(void) m_free(m);
   1038 		return (error);
   1039 	}
   1040 	*mp = m;
   1041 	if (type == MT_SONAME) {
   1042 		sa = mtod(m, struct sockaddr *);
   1043 
   1044 #if defined(COMPAT_OLDSOCK) && BYTE_ORDER != BIG_ENDIAN
   1045 		if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
   1046 			sa->sa_family = sa->sa_len;
   1047 #endif
   1048 		sa->sa_len = buflen;
   1049 	}
   1050 	return (0);
   1051 }
   1052 
   1053 int
   1054 getsock(fdp, fdes, fpp)
   1055 	struct filedesc *fdp;
   1056 	int fdes;
   1057 	struct file **fpp;
   1058 {
   1059 	register struct file *fp;
   1060 
   1061 	if ((unsigned)fdes >= fdp->fd_nfiles ||
   1062 	    (fp = fdp->fd_ofiles[fdes]) == NULL)
   1063 		return (EBADF);
   1064 	if (fp->f_type != DTYPE_SOCKET)
   1065 		return (ENOTSOCK);
   1066 	*fpp = fp;
   1067 	return (0);
   1068 }
   1069