Home | History | Annotate | Line # | Download | only in kern
uipc_syscalls.c revision 1.41.6.1
      1 /*	$NetBSD: uipc_syscalls.c,v 1.41.6.1 1999/06/28 06:36:53 itojun Exp $	*/
      2 
      3 /*
      4  * Copyright (c) 1982, 1986, 1989, 1990, 1993
      5  *	The Regents of the University of California.  All rights reserved.
      6  *
      7  * Redistribution and use in source and binary forms, with or without
      8  * modification, are permitted provided that the following conditions
      9  * are met:
     10  * 1. Redistributions of source code must retain the above copyright
     11  *    notice, this list of conditions and the following disclaimer.
     12  * 2. Redistributions in binary form must reproduce the above copyright
     13  *    notice, this list of conditions and the following disclaimer in the
     14  *    documentation and/or other materials provided with the distribution.
     15  * 3. All advertising materials mentioning features or use of this software
     16  *    must display the following acknowledgement:
     17  *	This product includes software developed by the University of
     18  *	California, Berkeley and its contributors.
     19  * 4. Neither the name of the University nor the names of its contributors
     20  *    may be used to endorse or promote products derived from this software
     21  *    without specific prior written permission.
     22  *
     23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     33  * SUCH DAMAGE.
     34  *
     35  *	@(#)uipc_syscalls.c	8.6 (Berkeley) 2/14/95
     36  */
     37 
     38 #include "opt_ktrace.h"
     39 #include "opt_compat_freebsd.h"
     40 #include "opt_compat_linux.h"
     41 #include "opt_compat_sunos.h"
     42 #include "opt_compat_hpux.h"
     43 #include "opt_compat_ultrix.h"
     44 #include "opt_compat_43.h"
     45 #if defined(COMPAT_43) || defined(COMPAT_SUNOS) || defined(COMPAT_LINUX) || \
     46     defined(COMPAT_HPUX) || defined(COMPAT_FREEBSD) || defined(COMPAT_ULTRIX)
     47 #define COMPAT_OLDSOCK /* used by <sys/socket.h> */
     48 #endif
     49 
     50 #include <sys/param.h>
     51 #include <sys/systm.h>
     52 #include <sys/filedesc.h>
     53 #include <sys/proc.h>
     54 #include <sys/file.h>
     55 #include <sys/buf.h>
     56 #include <sys/malloc.h>
     57 #include <sys/mbuf.h>
     58 #include <sys/protosw.h>
     59 #include <sys/socket.h>
     60 #include <sys/socketvar.h>
     61 #include <sys/signalvar.h>
     62 #include <sys/un.h>
     63 #ifdef KTRACE
     64 #include <sys/ktrace.h>
     65 #endif
     66 
     67 #include <sys/mount.h>
     68 #include <sys/syscallargs.h>
     69 
     70 /*
     71  * System call interface to the socket abstraction.
     72  */
     73 extern	struct fileops socketops;
     74 
     75 int
     76 sys_socket(p, v, retval)
     77 	struct proc *p;
     78 	void *v;
     79 	register_t *retval;
     80 {
     81 	register struct sys_socket_args /* {
     82 		syscallarg(int) domain;
     83 		syscallarg(int) type;
     84 		syscallarg(int) protocol;
     85 	} */ *uap = v;
     86 	struct filedesc *fdp = p->p_fd;
     87 	struct socket *so;
     88 	struct file *fp;
     89 	int fd, error;
     90 
     91 	if ((error = falloc(p, &fp, &fd)) != 0)
     92 		return (error);
     93 	fp->f_flag = FREAD|FWRITE;
     94 	fp->f_type = DTYPE_SOCKET;
     95 	fp->f_ops = &socketops;
     96 	error = socreate(SCARG(uap, domain), &so, SCARG(uap, type),
     97 			 SCARG(uap, protocol));
     98 	if (error) {
     99 		fdp->fd_ofiles[fd] = 0;
    100 		ffree(fp);
    101 	} else {
    102 		fp->f_data = (caddr_t)so;
    103 		*retval = fd;
    104 	}
    105 	return (error);
    106 }
    107 
    108 /* ARGSUSED */
    109 int
    110 sys_bind(p, v, retval)
    111 	struct proc *p;
    112 	void *v;
    113 	register_t *retval;
    114 {
    115 	register struct sys_bind_args /* {
    116 		syscallarg(int) s;
    117 		syscallarg(const struct sockaddr *) name;
    118 		syscallarg(unsigned int) namelen;
    119 	} */ *uap = v;
    120 	struct file *fp;
    121 	struct mbuf *nam;
    122 	int error;
    123 
    124 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    125 		return (error);
    126 	error = sockargs(&nam, SCARG(uap, name), SCARG(uap, namelen),
    127 	    MT_SONAME);
    128 	if (error)
    129 		return (error);
    130 	error = sobind((struct socket *)fp->f_data, nam);
    131 	m_freem(nam);
    132 	return (error);
    133 }
    134 
    135 /* ARGSUSED */
    136 int
    137 sys_listen(p, v, retval)
    138 	struct proc *p;
    139 	void *v;
    140 	register_t *retval;
    141 {
    142 	register struct sys_listen_args /* {
    143 		syscallarg(int) s;
    144 		syscallarg(int) backlog;
    145 	} */ *uap = v;
    146 	struct file *fp;
    147 	int error;
    148 
    149 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    150 		return (error);
    151 	return (solisten((struct socket *)fp->f_data, SCARG(uap, backlog)));
    152 }
    153 
    154 int
    155 sys_accept(p, v, retval)
    156 	struct proc *p;
    157 	void *v;
    158 	register_t *retval;
    159 {
    160 	register struct sys_accept_args /* {
    161 		syscallarg(int) s;
    162 		syscallarg(struct sockaddr *) name;
    163 		syscallarg(unsigned int *) anamelen;
    164 	} */ *uap = v;
    165 	struct file *fp;
    166 	struct mbuf *nam;
    167 	unsigned int namelen;
    168 	int error, s, tmpfd;
    169 	register struct socket *so;
    170 
    171 	if (SCARG(uap, name) && (error = copyin((caddr_t)SCARG(uap, anamelen),
    172 	    (caddr_t)&namelen, sizeof(namelen))))
    173 		return (error);
    174 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    175 		return (error);
    176 	s = splsoftnet();
    177 	so = (struct socket *)fp->f_data;
    178 	if ((so->so_options & SO_ACCEPTCONN) == 0) {
    179 		splx(s);
    180 		return (EINVAL);
    181 	}
    182 	if ((so->so_state & SS_NBIO) && so->so_qlen == 0) {
    183 		splx(s);
    184 		return (EWOULDBLOCK);
    185 	}
    186 	while (so->so_qlen == 0 && so->so_error == 0) {
    187 		if (so->so_state & SS_CANTRCVMORE) {
    188 			so->so_error = ECONNABORTED;
    189 			break;
    190 		}
    191 		error = tsleep((caddr_t)&so->so_timeo, PSOCK | PCATCH,
    192 			       netcon, 0);
    193 		if (error) {
    194 			splx(s);
    195 			return (error);
    196 		}
    197 	}
    198 	if (so->so_error) {
    199 		error = so->so_error;
    200 		so->so_error = 0;
    201 		splx(s);
    202 		return (error);
    203 	}
    204 	if ((error = falloc(p, &fp, &tmpfd)) != 0) {
    205 		splx(s);
    206 		return (error);
    207 	}
    208 	*retval = tmpfd;
    209 	{ struct socket *aso = so->so_q.tqh_first;
    210 	  if (soqremque(aso, 1) == 0)
    211 		panic("accept");
    212 	  so = aso;
    213 	}
    214 	fp->f_type = DTYPE_SOCKET;
    215 	fp->f_flag = FREAD|FWRITE;
    216 	fp->f_ops = &socketops;
    217 	fp->f_data = (caddr_t)so;
    218 	nam = m_get(M_WAIT, MT_SONAME);
    219 	(void) soaccept(so, nam);
    220 	if (SCARG(uap, name)) {
    221 		if (namelen > nam->m_len)
    222 			namelen = nam->m_len;
    223 		/* SHOULD COPY OUT A CHAIN HERE */
    224 		if ((error = copyout(mtod(nam, caddr_t),
    225 		    (caddr_t)SCARG(uap, name), namelen)) == 0)
    226 			error = copyout((caddr_t)&namelen,
    227 			    (caddr_t)SCARG(uap, anamelen),
    228 			    sizeof(*SCARG(uap, anamelen)));
    229 	}
    230 	m_freem(nam);
    231 	splx(s);
    232 	return (error);
    233 }
    234 
    235 /* ARGSUSED */
    236 int
    237 sys_connect(p, v, retval)
    238 	struct proc *p;
    239 	void *v;
    240 	register_t *retval;
    241 {
    242 	register struct sys_connect_args /* {
    243 		syscallarg(int) s;
    244 		syscallarg(const struct sockaddr *) name;
    245 		syscallarg(unsigned int) namelen;
    246 	} */ *uap = v;
    247 	struct file *fp;
    248 	register struct socket *so;
    249 	struct mbuf *nam;
    250 	int error, s;
    251 
    252 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    253 		return (error);
    254 	so = (struct socket *)fp->f_data;
    255 	if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING))
    256 		return (EALREADY);
    257 	error = sockargs(&nam, SCARG(uap, name), SCARG(uap, namelen),
    258 	    MT_SONAME);
    259 	if (error)
    260 		return (error);
    261 	error = soconnect(so, nam);
    262 	if (error)
    263 		goto bad;
    264 	if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING)) {
    265 		m_freem(nam);
    266 		return (EINPROGRESS);
    267 	}
    268 	s = splsoftnet();
    269 	while ((so->so_state & SS_ISCONNECTING) && so->so_error == 0) {
    270 		error = tsleep((caddr_t)&so->so_timeo, PSOCK | PCATCH,
    271 			       netcon, 0);
    272 		if (error)
    273 			break;
    274 	}
    275 	if (error == 0) {
    276 		error = so->so_error;
    277 		so->so_error = 0;
    278 	}
    279 	splx(s);
    280 bad:
    281 	so->so_state &= ~SS_ISCONNECTING;
    282 	m_freem(nam);
    283 	if (error == ERESTART)
    284 		error = EINTR;
    285 	return (error);
    286 }
    287 
    288 int
    289 sys_socketpair(p, v, retval)
    290 	struct proc *p;
    291 	void *v;
    292 	register_t *retval;
    293 {
    294 	register struct sys_socketpair_args /* {
    295 		syscallarg(int) domain;
    296 		syscallarg(int) type;
    297 		syscallarg(int) protocol;
    298 		syscallarg(int *) rsv;
    299 	} */ *uap = v;
    300 	register struct filedesc *fdp = p->p_fd;
    301 	struct file *fp1, *fp2;
    302 	struct socket *so1, *so2;
    303 	int fd, error, sv[2];
    304 
    305 	error = socreate(SCARG(uap, domain), &so1, SCARG(uap, type),
    306 			 SCARG(uap, protocol));
    307 	if (error)
    308 		return (error);
    309 	error = socreate(SCARG(uap, domain), &so2, SCARG(uap, type),
    310 			 SCARG(uap, protocol));
    311 	if (error)
    312 		goto free1;
    313 	if ((error = falloc(p, &fp1, &fd)) != 0)
    314 		goto free2;
    315 	sv[0] = fd;
    316 	fp1->f_flag = FREAD|FWRITE;
    317 	fp1->f_type = DTYPE_SOCKET;
    318 	fp1->f_ops = &socketops;
    319 	fp1->f_data = (caddr_t)so1;
    320 	if ((error = falloc(p, &fp2, &fd)) != 0)
    321 		goto free3;
    322 	fp2->f_flag = FREAD|FWRITE;
    323 	fp2->f_type = DTYPE_SOCKET;
    324 	fp2->f_ops = &socketops;
    325 	fp2->f_data = (caddr_t)so2;
    326 	sv[1] = fd;
    327 	if ((error = soconnect2(so1, so2)) != 0)
    328 		goto free4;
    329 	if (SCARG(uap, type) == SOCK_DGRAM) {
    330 		/*
    331 		 * Datagram socket connection is asymmetric.
    332 		 */
    333 		 if ((error = soconnect2(so2, so1)) != 0)
    334 			goto free4;
    335 	}
    336 	error = copyout((caddr_t)sv, (caddr_t)SCARG(uap, rsv),
    337 	    2 * sizeof(int));
    338 	return (error);
    339 free4:
    340 	ffree(fp2);
    341 	fdp->fd_ofiles[sv[1]] = 0;
    342 free3:
    343 	ffree(fp1);
    344 	fdp->fd_ofiles[sv[0]] = 0;
    345 free2:
    346 	(void)soclose(so2);
    347 free1:
    348 	(void)soclose(so1);
    349 	return (error);
    350 }
    351 
    352 int
    353 sys_sendto(p, v, retval)
    354 	struct proc *p;
    355 	void *v;
    356 	register_t *retval;
    357 {
    358 	register struct sys_sendto_args /* {
    359 		syscallarg(int) s;
    360 		syscallarg(const void *) buf;
    361 		syscallarg(size_t) len;
    362 		syscallarg(int) flags;
    363 		syscallarg(const struct sockaddr *) to;
    364 		syscallarg(unsigned int) tolen;
    365 	} */ *uap = v;
    366 	struct msghdr msg;
    367 	struct iovec aiov;
    368 
    369 	msg.msg_name = (caddr_t)SCARG(uap, to);		/* XXX kills const */
    370 	msg.msg_namelen = SCARG(uap, tolen);
    371 	msg.msg_iov = &aiov;
    372 	msg.msg_iovlen = 1;
    373 	msg.msg_control = 0;
    374 #ifdef COMPAT_OLDSOCK
    375 	msg.msg_flags = 0;
    376 #endif
    377 	aiov.iov_base = (char *)SCARG(uap, buf);	/* XXX kills const */
    378 	aiov.iov_len = SCARG(uap, len);
    379 	return (sendit(p, SCARG(uap, s), &msg, SCARG(uap, flags), retval));
    380 }
    381 
    382 int
    383 sys_sendmsg(p, v, retval)
    384 	struct proc *p;
    385 	void *v;
    386 	register_t *retval;
    387 {
    388 	register struct sys_sendmsg_args /* {
    389 		syscallarg(int) s;
    390 		syscallarg(const struct msghdr *) msg;
    391 		syscallarg(int) flags;
    392 	} */ *uap = v;
    393 	struct msghdr msg;
    394 	struct iovec aiov[UIO_SMALLIOV], *iov;
    395 	int error;
    396 
    397 	error = copyin(SCARG(uap, msg), (caddr_t)&msg, sizeof(msg));
    398 	if (error)
    399 		return (error);
    400 	if ((unsigned int)msg.msg_iovlen > UIO_SMALLIOV) {
    401 		if ((unsigned int)msg.msg_iovlen > IOV_MAX)
    402 			return (EMSGSIZE);
    403 		MALLOC(iov, struct iovec *,
    404 		    sizeof(struct iovec) * msg.msg_iovlen, M_IOV, M_WAITOK);
    405 	} else
    406 		iov = aiov;
    407 	if ((unsigned int)msg.msg_iovlen > 0) {
    408 		error = copyin((caddr_t)msg.msg_iov, (caddr_t)iov,
    409 		    (size_t)(msg.msg_iovlen * sizeof(struct iovec)));
    410 		if (error)
    411 			goto done;
    412 	}
    413 	msg.msg_iov = iov;
    414 #ifdef COMPAT_OLDSOCK
    415 	msg.msg_flags = 0;
    416 #endif
    417 	error = sendit(p, SCARG(uap, s), &msg, SCARG(uap, flags), retval);
    418 done:
    419 	if (iov != aiov)
    420 		FREE(iov, M_IOV);
    421 	return (error);
    422 }
    423 
    424 int
    425 sendit(p, s, mp, flags, retsize)
    426 	register struct proc *p;
    427 	int s;
    428 	register struct msghdr *mp;
    429 	int flags;
    430 	register_t *retsize;
    431 {
    432 	struct file *fp;
    433 	struct uio auio;
    434 	register struct iovec *iov;
    435 	register int i;
    436 	struct mbuf *to, *control;
    437 	int len, error;
    438 	struct socket *so;
    439 #ifdef KTRACE
    440 	struct iovec *ktriov = NULL;
    441 #endif
    442 
    443 	if ((error = getsock(p->p_fd, s, &fp)) != 0)
    444 		return (error);
    445 	auio.uio_iov = mp->msg_iov;
    446 	auio.uio_iovcnt = mp->msg_iovlen;
    447 	auio.uio_segflg = UIO_USERSPACE;
    448 	auio.uio_rw = UIO_WRITE;
    449 	auio.uio_procp = p;
    450 	auio.uio_offset = 0;			/* XXX */
    451 	auio.uio_resid = 0;
    452 	iov = mp->msg_iov;
    453 	for (i = 0; i < mp->msg_iovlen; i++, iov++) {
    454 #if 0
    455 		/* cannot happen; iov_len is unsigned */
    456 		if (iov->iov_len < 0)
    457 			return (EINVAL);
    458 #endif
    459 		/*
    460 		 * Writes return ssize_t because -1 is returned on error.
    461 		 * Therefore, we must restrict the length to SSIZE_MAX to
    462 		 * avoid garbage return values.
    463 		 */
    464 		auio.uio_resid += iov->iov_len;
    465 		if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX)
    466 			return (EINVAL);
    467 	}
    468 	if (mp->msg_name) {
    469 		error = sockargs(&to, mp->msg_name, mp->msg_namelen,
    470 				 MT_SONAME);
    471 		if (error)
    472 			return (error);
    473 	} else
    474 		to = 0;
    475 	if (mp->msg_control) {
    476 		if (mp->msg_controllen < sizeof(struct cmsghdr)
    477 #ifdef COMPAT_OLDSOCK
    478 		    && mp->msg_flags != MSG_COMPAT
    479 #endif
    480 		) {
    481 			error = EINVAL;
    482 			goto bad;
    483 		}
    484 		error = sockargs(&control, mp->msg_control,
    485 				 mp->msg_controllen, MT_CONTROL);
    486 		if (error)
    487 			goto bad;
    488 #ifdef COMPAT_OLDSOCK
    489 		if (mp->msg_flags == MSG_COMPAT) {
    490 			register struct cmsghdr *cm;
    491 
    492 			M_PREPEND(control, sizeof(*cm), M_WAIT);
    493 			if (control == 0) {
    494 				error = ENOBUFS;
    495 				goto bad;
    496 			} else {
    497 				cm = mtod(control, struct cmsghdr *);
    498 				cm->cmsg_len = control->m_len;
    499 				cm->cmsg_level = SOL_SOCKET;
    500 				cm->cmsg_type = SCM_RIGHTS;
    501 			}
    502 		}
    503 #endif
    504 	} else
    505 		control = 0;
    506 #ifdef KTRACE
    507 	if (KTRPOINT(p, KTR_GENIO)) {
    508 		int iovlen = auio.uio_iovcnt * sizeof(struct iovec);
    509 
    510 		MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK);
    511 		memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen);
    512 	}
    513 #endif
    514 	len = auio.uio_resid;
    515 	so = (struct socket *)fp->f_data;
    516 	error = (*so->so_send)(so, to, &auio, NULL, control, flags);
    517 	if (error) {
    518 		if (auio.uio_resid != len && (error == ERESTART ||
    519 		    error == EINTR || error == EWOULDBLOCK))
    520 			error = 0;
    521 		if (error == EPIPE)
    522 			psignal(p, SIGPIPE);
    523 	}
    524 	if (error == 0)
    525 		*retsize = len - auio.uio_resid;
    526 #ifdef KTRACE
    527 	if (ktriov != NULL) {
    528 		if (error == 0)
    529 			ktrgenio(p->p_tracep, s, UIO_WRITE,
    530 				ktriov, *retsize, error);
    531 		FREE(ktriov, M_TEMP);
    532 	}
    533 #endif
    534 bad:
    535 	if (to)
    536 		m_freem(to);
    537 	return (error);
    538 }
    539 
    540 int
    541 sys_recvfrom(p, v, retval)
    542 	struct proc *p;
    543 	void *v;
    544 	register_t *retval;
    545 {
    546 	register struct sys_recvfrom_args /* {
    547 		syscallarg(int) s;
    548 		syscallarg(void *) buf;
    549 		syscallarg(size_t) len;
    550 		syscallarg(int) flags;
    551 		syscallarg(struct sockaddr *) from;
    552 		syscallarg(unsigned int *) fromlenaddr;
    553 	} */ *uap = v;
    554 	struct msghdr msg;
    555 	struct iovec aiov;
    556 	int error;
    557 
    558 	if (SCARG(uap, fromlenaddr)) {
    559 		error = copyin((caddr_t)SCARG(uap, fromlenaddr),
    560 			       (caddr_t)&msg.msg_namelen,
    561 			       sizeof(msg.msg_namelen));
    562 		if (error)
    563 			return (error);
    564 	} else
    565 		msg.msg_namelen = 0;
    566 	msg.msg_name = (caddr_t)SCARG(uap, from);
    567 	msg.msg_iov = &aiov;
    568 	msg.msg_iovlen = 1;
    569 	aiov.iov_base = SCARG(uap, buf);
    570 	aiov.iov_len = SCARG(uap, len);
    571 	msg.msg_control = 0;
    572 	msg.msg_flags = SCARG(uap, flags);
    573 	return (recvit(p, SCARG(uap, s), &msg,
    574 		       (caddr_t)SCARG(uap, fromlenaddr), retval));
    575 }
    576 
    577 int
    578 sys_recvmsg(p, v, retval)
    579 	struct proc *p;
    580 	void *v;
    581 	register_t *retval;
    582 {
    583 	register struct sys_recvmsg_args /* {
    584 		syscallarg(int) s;
    585 		syscallarg(struct msghdr *) msg;
    586 		syscallarg(int) flags;
    587 	} */ *uap = v;
    588 	struct msghdr msg;
    589 	struct iovec aiov[UIO_SMALLIOV], *uiov, *iov;
    590 	register int error;
    591 
    592 	error = copyin((caddr_t)SCARG(uap, msg), (caddr_t)&msg,
    593 		       sizeof(msg));
    594 	if (error)
    595 		return (error);
    596 	if ((unsigned int)msg.msg_iovlen > UIO_SMALLIOV) {
    597 		if ((unsigned int)msg.msg_iovlen > IOV_MAX)
    598 			return (EMSGSIZE);
    599 		MALLOC(iov, struct iovec *,
    600 		    sizeof(struct iovec) * msg.msg_iovlen, M_IOV, M_WAITOK);
    601 	} else
    602 		iov = aiov;
    603 	if ((unsigned int)msg.msg_iovlen > 0) {
    604 		error = copyin((caddr_t)msg.msg_iov, (caddr_t)iov,
    605 		    (size_t)(msg.msg_iovlen * sizeof(struct iovec)));
    606 		if (error)
    607 			goto done;
    608 	}
    609 	uiov = msg.msg_iov;
    610 	msg.msg_iov = iov;
    611 #ifdef COMPAT_OLDSOCK
    612 	msg.msg_flags = SCARG(uap, flags) &~ MSG_COMPAT;
    613 #else
    614 	msg.msg_flags = SCARG(uap, flags);
    615 #endif
    616 	if ((error = recvit(p, SCARG(uap, s), &msg, (caddr_t)0, retval)) == 0) {
    617 		msg.msg_iov = uiov;
    618 		error = copyout((caddr_t)&msg, (caddr_t)SCARG(uap, msg),
    619 		    sizeof(msg));
    620 	}
    621 done:
    622 	if (iov != aiov)
    623 		FREE(iov, M_IOV);
    624 	return (error);
    625 }
    626 
    627 int
    628 recvit(p, s, mp, namelenp, retsize)
    629 	register struct proc *p;
    630 	int s;
    631 	register struct msghdr *mp;
    632 	caddr_t namelenp;
    633 	register_t *retsize;
    634 {
    635 	struct file *fp;
    636 	struct uio auio;
    637 	register struct iovec *iov;
    638 	register int i;
    639 	int len, error;
    640 	struct mbuf *from = 0, *control = 0;
    641 	struct socket *so;
    642 #ifdef KTRACE
    643 	struct iovec *ktriov = NULL;
    644 #endif
    645 
    646 	if ((error = getsock(p->p_fd, s, &fp)) != 0)
    647 		return (error);
    648 	auio.uio_iov = mp->msg_iov;
    649 	auio.uio_iovcnt = mp->msg_iovlen;
    650 	auio.uio_segflg = UIO_USERSPACE;
    651 	auio.uio_rw = UIO_READ;
    652 	auio.uio_procp = p;
    653 	auio.uio_offset = 0;			/* XXX */
    654 	auio.uio_resid = 0;
    655 	iov = mp->msg_iov;
    656 	for (i = 0; i < mp->msg_iovlen; i++, iov++) {
    657 #if 0
    658 		/* cannot happen iov_len is unsigned */
    659 		if (iov->iov_len < 0)
    660 			return (EINVAL);
    661 #endif
    662 		/*
    663 		 * Reads return ssize_t because -1 is returned on error.
    664 		 * Therefore we must restrict the length to SSIZE_MAX to
    665 		 * avoid garbage return values.
    666 		 */
    667 		auio.uio_resid += iov->iov_len;
    668 		if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX)
    669 			return (EINVAL);
    670 	}
    671 #ifdef KTRACE
    672 	if (KTRPOINT(p, KTR_GENIO)) {
    673 		int iovlen = auio.uio_iovcnt * sizeof(struct iovec);
    674 
    675 		MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK);
    676 		memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen);
    677 	}
    678 #endif
    679 	len = auio.uio_resid;
    680 	so = (struct socket *)fp->f_data;
    681 	error = (*so->so_receive)(so, &from, &auio, NULL,
    682 			  mp->msg_control ? &control : NULL, &mp->msg_flags);
    683 	if (error) {
    684 		if (auio.uio_resid != len && (error == ERESTART ||
    685 		    error == EINTR || error == EWOULDBLOCK))
    686 			error = 0;
    687 	}
    688 #ifdef KTRACE
    689 	if (ktriov != NULL) {
    690 		if (error == 0)
    691 			ktrgenio(p->p_tracep, s, UIO_READ,
    692 				ktriov, len - auio.uio_resid, error);
    693 		FREE(ktriov, M_TEMP);
    694 	}
    695 #endif
    696 	if (error)
    697 		goto out;
    698 	*retsize = len - auio.uio_resid;
    699 	if (mp->msg_name) {
    700 		len = mp->msg_namelen;
    701 		if (len <= 0 || from == 0)
    702 			len = 0;
    703 		else {
    704 #ifdef COMPAT_OLDSOCK
    705 			if (mp->msg_flags & MSG_COMPAT)
    706 				mtod(from, struct osockaddr *)->sa_family =
    707 				    mtod(from, struct sockaddr *)->sa_family;
    708 #endif
    709 			if (len > from->m_len)
    710 				len = from->m_len;
    711 			/* else if len < from->m_len ??? */
    712 			error = copyout(mtod(from, caddr_t),
    713 					(caddr_t)mp->msg_name, (unsigned)len);
    714 			if (error)
    715 				goto out;
    716 		}
    717 		mp->msg_namelen = len;
    718 		if (namelenp &&
    719 		    (error = copyout((caddr_t)&len, namelenp, sizeof(int)))) {
    720 #ifdef COMPAT_OLDSOCK
    721 			if (mp->msg_flags & MSG_COMPAT)
    722 				error = 0;	/* old recvfrom didn't check */
    723 			else
    724 #endif
    725 			goto out;
    726 		}
    727 	}
    728 	if (mp->msg_control) {
    729 #ifdef COMPAT_OLDSOCK
    730 		/*
    731 		 * We assume that old recvmsg calls won't receive access
    732 		 * rights and other control info, esp. as control info
    733 		 * is always optional and those options didn't exist in 4.3.
    734 		 * If we receive rights, trim the cmsghdr; anything else
    735 		 * is tossed.
    736 		 */
    737 		if (control && mp->msg_flags & MSG_COMPAT) {
    738 			if (mtod(control, struct cmsghdr *)->cmsg_level !=
    739 			    SOL_SOCKET ||
    740 			    mtod(control, struct cmsghdr *)->cmsg_type !=
    741 			    SCM_RIGHTS) {
    742 				mp->msg_controllen = 0;
    743 				goto out;
    744 			}
    745 			control->m_len -= sizeof(struct cmsghdr);
    746 			control->m_data += sizeof(struct cmsghdr);
    747 		}
    748 #endif
    749 		len = mp->msg_controllen;
    750 		if (len <= 0 || control == 0)
    751 			len = 0;
    752 		else {
    753 			struct mbuf *m = control;
    754 			caddr_t p = (caddr_t)mp->msg_control;
    755 
    756 			do {
    757 				i = m->m_len;
    758 				if (len < i) {
    759 					mp->msg_flags |= MSG_CTRUNC;
    760 					i = len;
    761 				}
    762 				error = copyout(mtod(m, caddr_t), p,
    763 				    (unsigned)i);
    764 				if (m->m_next)
    765 					i = ALIGN(i);
    766 				p += i;
    767 				len -= i;
    768 				if (error != 0 || len <= 0)
    769 					break;
    770 			} while ((m = m->m_next) != NULL);
    771 			len = p - (caddr_t)mp->msg_control;
    772 		}
    773 		mp->msg_controllen = len;
    774 	}
    775 out:
    776 	if (from)
    777 		m_freem(from);
    778 	if (control)
    779 		m_freem(control);
    780 	return (error);
    781 }
    782 
    783 /* ARGSUSED */
    784 int
    785 sys_shutdown(p, v, retval)
    786 	struct proc *p;
    787 	void *v;
    788 	register_t *retval;
    789 {
    790 	register struct sys_shutdown_args /* {
    791 		syscallarg(int) s;
    792 		syscallarg(int) how;
    793 	} */ *uap = v;
    794 	struct file *fp;
    795 	int error;
    796 
    797 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    798 		return (error);
    799 	return (soshutdown((struct socket *)fp->f_data, SCARG(uap, how)));
    800 }
    801 
    802 /* ARGSUSED */
    803 int
    804 sys_setsockopt(p, v, retval)
    805 	struct proc *p;
    806 	void *v;
    807 	register_t *retval;
    808 {
    809 	register struct sys_setsockopt_args /* {
    810 		syscallarg(int) s;
    811 		syscallarg(int) level;
    812 		syscallarg(int) name;
    813 		syscallarg(const void *) val;
    814 		syscallarg(unsigned int) valsize;
    815 	} */ *uap = v;
    816 	struct file *fp;
    817 	struct mbuf *m = NULL;
    818 	int error;
    819 
    820 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    821 		return (error);
    822 	if (SCARG(uap, valsize) > MCLBYTES)
    823 		return (EINVAL);
    824 	if (SCARG(uap, val)) {
    825 		m = m_get(M_WAIT, MT_SOOPTS);
    826 		if (SCARG(uap, valsize) > MLEN) {
    827 			MCLGET(m, M_DONTWAIT);
    828 			if ((m->m_flags & M_EXT) == 0) {
    829 				m_freem(m);
    830 				return (ENOBUFS);
    831 			}
    832 		}
    833 		error = copyin(SCARG(uap, val), mtod(m, caddr_t),
    834 			       SCARG(uap, valsize));
    835 		if (error) {
    836 			(void) m_free(m);
    837 			return (error);
    838 		}
    839 		m->m_len = SCARG(uap, valsize);
    840 	}
    841 	return (sosetopt((struct socket *)fp->f_data, SCARG(uap, level),
    842 			 SCARG(uap, name), m));
    843 }
    844 
    845 /* ARGSUSED */
    846 int
    847 sys_getsockopt(p, v, retval)
    848 	struct proc *p;
    849 	void *v;
    850 	register_t *retval;
    851 {
    852 	register struct sys_getsockopt_args /* {
    853 		syscallarg(int) s;
    854 		syscallarg(int) level;
    855 		syscallarg(int) name;
    856 		syscallarg(void *) val;
    857 		syscallarg(unsigned int *) avalsize;
    858 	} */ *uap = v;
    859 	struct file *fp;
    860 	struct mbuf *m = NULL, *m0;
    861 	unsigned int op, i, valsize;
    862 	int error;
    863 
    864 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    865 		return (error);
    866 	if (SCARG(uap, val)) {
    867 		error = copyin((caddr_t)SCARG(uap, avalsize),
    868 			       (caddr_t)&valsize, sizeof(valsize));
    869 		if (error)
    870 			return (error);
    871 	} else
    872 		valsize = 0;
    873 	if ((error = sogetopt((struct socket *)fp->f_data, SCARG(uap, level),
    874 	    SCARG(uap, name), &m)) == 0 && SCARG(uap, val) && valsize &&
    875 	    m != NULL) {
    876 		op = 0;
    877 		while (m && !error && op < valsize) {
    878 			i = min(m->m_len, (valsize - op));
    879 			error = copyout(mtod(m, caddr_t), SCARG(uap, val), i);
    880 			op += i;
    881 			SCARG(uap, val) = ((u_int8_t *)SCARG(uap, val)) + i;
    882 			m0 = m;
    883 			MFREE(m0, m);
    884 		}
    885 		valsize = op;
    886 		if (error == 0)
    887 			error = copyout(&valsize,
    888 					SCARG(uap, avalsize), sizeof(valsize));
    889 	}
    890 	if (m != NULL)
    891 		(void) m_free(m);
    892 	return (error);
    893 }
    894 
    895 /* ARGSUSED */
    896 int
    897 sys_pipe(p, v, retval)
    898 	struct proc *p;
    899 	void *v;
    900 	register_t *retval;
    901 {
    902 	register struct filedesc *fdp = p->p_fd;
    903 	struct file *rf, *wf;
    904 	struct socket *rso, *wso;
    905 	int fd, error;
    906 
    907 	if ((error = socreate(AF_LOCAL, &rso, SOCK_STREAM, 0)) != 0)
    908 		return (error);
    909 	if ((error = socreate(AF_LOCAL, &wso, SOCK_STREAM, 0)) != 0)
    910 		goto free1;
    911 	if ((error = falloc(p, &rf, &fd)) != 0)
    912 		goto free2;
    913 	retval[0] = fd;
    914 	rf->f_flag = FREAD;
    915 	rf->f_type = DTYPE_SOCKET;
    916 	rf->f_ops = &socketops;
    917 	rf->f_data = (caddr_t)rso;
    918 	if ((error = falloc(p, &wf, &fd)) != 0)
    919 		goto free3;
    920 	wf->f_flag = FWRITE;
    921 	wf->f_type = DTYPE_SOCKET;
    922 	wf->f_ops = &socketops;
    923 	wf->f_data = (caddr_t)wso;
    924 	retval[1] = fd;
    925 	if ((error = unp_connect2(wso, rso)) != 0)
    926 		goto free4;
    927 	return (0);
    928 free4:
    929 	ffree(wf);
    930 	fdp->fd_ofiles[retval[1]] = 0;
    931 free3:
    932 	ffree(rf);
    933 	fdp->fd_ofiles[retval[0]] = 0;
    934 free2:
    935 	(void)soclose(wso);
    936 free1:
    937 	(void)soclose(rso);
    938 	return (error);
    939 }
    940 
    941 /*
    942  * Get socket name.
    943  */
    944 /* ARGSUSED */
    945 int
    946 sys_getsockname(p, v, retval)
    947 	struct proc *p;
    948 	void *v;
    949 	register_t *retval;
    950 {
    951 	register struct sys_getsockname_args /* {
    952 		syscallarg(int) fdes;
    953 		syscallarg(struct sockaddr *) asa;
    954 		syscallarg(unsigned int *) alen;
    955 	} */ *uap = v;
    956 	struct file *fp;
    957 	register struct socket *so;
    958 	struct mbuf *m;
    959 	unsigned int len;
    960 	int error;
    961 
    962 	if ((error = getsock(p->p_fd, SCARG(uap, fdes), &fp)) != 0)
    963 		return (error);
    964 	error = copyin((caddr_t)SCARG(uap, alen), (caddr_t)&len, sizeof(len));
    965 	if (error)
    966 		return (error);
    967 	so = (struct socket *)fp->f_data;
    968 	m = m_getclr(M_WAIT, MT_SONAME);
    969 	error = (*so->so_proto->pr_usrreq)(so, PRU_SOCKADDR, (struct mbuf *)0,
    970 	    m, (struct mbuf *)0, (struct proc *)0);
    971 	if (error)
    972 		goto bad;
    973 	if (len > m->m_len)
    974 		len = m->m_len;
    975 	error = copyout(mtod(m, caddr_t), (caddr_t)SCARG(uap, asa), len);
    976 	if (error == 0)
    977 		error = copyout((caddr_t)&len, (caddr_t)SCARG(uap, alen),
    978 		    sizeof(len));
    979 bad:
    980 	m_freem(m);
    981 	return (error);
    982 }
    983 
    984 /*
    985  * Get name of peer for connected socket.
    986  */
    987 /* ARGSUSED */
    988 int
    989 sys_getpeername(p, v, retval)
    990 	struct proc *p;
    991 	void *v;
    992 	register_t *retval;
    993 {
    994 	register struct sys_getpeername_args /* {
    995 		syscallarg(int) fdes;
    996 		syscallarg(struct sockaddr *) asa;
    997 		syscallarg(unsigned int *) alen;
    998 	} */ *uap = v;
    999 	struct file *fp;
   1000 	register struct socket *so;
   1001 	struct mbuf *m;
   1002 	unsigned int len;
   1003 	int error;
   1004 
   1005 	if ((error = getsock(p->p_fd, SCARG(uap, fdes), &fp)) != 0)
   1006 		return (error);
   1007 	so = (struct socket *)fp->f_data;
   1008 	if ((so->so_state & (SS_ISCONNECTED|SS_ISCONFIRMING)) == 0)
   1009 		return (ENOTCONN);
   1010 	error = copyin((caddr_t)SCARG(uap, alen), (caddr_t)&len, sizeof(len));
   1011 	if (error)
   1012 		return (error);
   1013 	m = m_getclr(M_WAIT, MT_SONAME);
   1014 	error = (*so->so_proto->pr_usrreq)(so, PRU_PEERADDR, (struct mbuf *)0,
   1015 	    m, (struct mbuf *)0, (struct proc *)0);
   1016 	if (error)
   1017 		goto bad;
   1018 	if (len > m->m_len)
   1019 		len = m->m_len;
   1020 	error = copyout(mtod(m, caddr_t), (caddr_t)SCARG(uap, asa), len);
   1021 	if (error)
   1022 		goto bad;
   1023 	error = copyout((caddr_t)&len, (caddr_t)SCARG(uap, alen), sizeof(len));
   1024 bad:
   1025 	m_freem(m);
   1026 	return (error);
   1027 }
   1028 
   1029 /*
   1030  * XXX In a perfect world, we wouldn't pass around socket control
   1031  * XXX arguments in mbufs, and this could go away.
   1032  */
   1033 int
   1034 sockargs(mp, buf, buflen, type)
   1035 	struct mbuf **mp;
   1036 	const void *buf;
   1037 	int buflen, type;
   1038 {
   1039 	register struct sockaddr *sa;
   1040 	register struct mbuf *m;
   1041 	int error;
   1042 
   1043 	/*
   1044 	 * We can't allow socket names > UCHAR_MAX in length, since that
   1045 	 * will overflow sa_len.
   1046 	 */
   1047 	if (type == MT_SONAME && (u_int)buflen > UCHAR_MAX)
   1048 		return (EINVAL);
   1049 
   1050 	/* Allocate an mbuf to hold the arguments. */
   1051 	m = m_get(M_WAIT, type);
   1052 	if ((u_int)buflen > MLEN) {
   1053 		/*
   1054 		 * Won't fit into a regular mbuf, so we allocate just
   1055 		 * enough external storage to hold the argument.
   1056 		 */
   1057 		MEXTMALLOC(m, buflen, M_WAITOK);
   1058 	}
   1059 	m->m_len = buflen;
   1060 	error = copyin(buf, mtod(m, caddr_t), (u_int)buflen);
   1061 	if (error) {
   1062 		(void) m_free(m);
   1063 		return (error);
   1064 	}
   1065 	*mp = m;
   1066 	if (type == MT_SONAME) {
   1067 		sa = mtod(m, struct sockaddr *);
   1068 
   1069 #if defined(COMPAT_OLDSOCK) && BYTE_ORDER != BIG_ENDIAN
   1070 		if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
   1071 			sa->sa_family = sa->sa_len;
   1072 #endif
   1073 		sa->sa_len = buflen;
   1074 	}
   1075 	return (0);
   1076 }
   1077 
   1078 int
   1079 getsock(fdp, fdes, fpp)
   1080 	struct filedesc *fdp;
   1081 	int fdes;
   1082 	struct file **fpp;
   1083 {
   1084 	register struct file *fp;
   1085 
   1086 	if ((unsigned)fdes >= fdp->fd_nfiles ||
   1087 	    (fp = fdp->fd_ofiles[fdes]) == NULL)
   1088 		return (EBADF);
   1089 	if (fp->f_type != DTYPE_SOCKET)
   1090 		return (ENOTSOCK);
   1091 	*fpp = fp;
   1092 	return (0);
   1093 }
   1094