Home | History | Annotate | Line # | Download | only in kern
uipc_syscalls.c revision 1.45.2.1
      1 /*	$NetBSD: uipc_syscalls.c,v 1.45.2.1 2000/11/20 18:09:14 bouyer 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 #include "opt_compat_osf1.h"
     46 #if defined(COMPAT_43) || defined(COMPAT_SUNOS) || defined(COMPAT_LINUX) || \
     47     defined(COMPAT_HPUX) || defined(COMPAT_FREEBSD) || \
     48     defined(COMPAT_ULTRIX) || defined(COMPAT_OSF1)
     49 #define COMPAT_OLDSOCK /* used by <sys/socket.h> */
     50 #endif
     51 
     52 #include <sys/param.h>
     53 #include <sys/systm.h>
     54 #include <sys/filedesc.h>
     55 #include <sys/proc.h>
     56 #include <sys/file.h>
     57 #include <sys/buf.h>
     58 #include <sys/malloc.h>
     59 #include <sys/mbuf.h>
     60 #include <sys/protosw.h>
     61 #include <sys/socket.h>
     62 #include <sys/socketvar.h>
     63 #include <sys/signalvar.h>
     64 #include <sys/un.h>
     65 #ifdef KTRACE
     66 #include <sys/ktrace.h>
     67 #endif
     68 
     69 #include <sys/mount.h>
     70 #include <sys/syscallargs.h>
     71 
     72 #include <uvm/uvm_extern.h>
     73 
     74 /*
     75  * System call interface to the socket abstraction.
     76  */
     77 extern	struct fileops socketops;
     78 
     79 int
     80 sys_socket(p, v, retval)
     81 	struct proc *p;
     82 	void *v;
     83 	register_t *retval;
     84 {
     85 	struct sys_socket_args /* {
     86 		syscallarg(int) domain;
     87 		syscallarg(int) type;
     88 		syscallarg(int) protocol;
     89 	} */ *uap = v;
     90 	struct filedesc *fdp = p->p_fd;
     91 	struct socket *so;
     92 	struct file *fp;
     93 	int fd, error;
     94 
     95 	/* falloc() will use the desciptor for us */
     96 	if ((error = falloc(p, &fp, &fd)) != 0)
     97 		return (error);
     98 	fp->f_flag = FREAD|FWRITE;
     99 	fp->f_type = DTYPE_SOCKET;
    100 	fp->f_ops = &socketops;
    101 	error = socreate(SCARG(uap, domain), &so, SCARG(uap, type),
    102 			 SCARG(uap, protocol));
    103 	if (error) {
    104 		FILE_UNUSE(fp, p);
    105 		fdremove(fdp, fd);
    106 		ffree(fp);
    107 	} else {
    108 		fp->f_data = (caddr_t)so;
    109 		FILE_UNUSE(fp, p);
    110 		*retval = fd;
    111 	}
    112 	return (error);
    113 }
    114 
    115 /* ARGSUSED */
    116 int
    117 sys_bind(p, v, retval)
    118 	struct proc *p;
    119 	void *v;
    120 	register_t *retval;
    121 {
    122 	struct sys_bind_args /* {
    123 		syscallarg(int) s;
    124 		syscallarg(const struct sockaddr *) name;
    125 		syscallarg(unsigned int) namelen;
    126 	} */ *uap = v;
    127 	struct file *fp;
    128 	struct mbuf *nam;
    129 	int error;
    130 
    131 	/* getsock() will use the descriptor for us */
    132 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    133 		return (error);
    134 	error = sockargs(&nam, SCARG(uap, name), SCARG(uap, namelen),
    135 	    MT_SONAME);
    136 	if (error) {
    137 		FILE_UNUSE(fp, p);
    138 		return (error);
    139 	}
    140 	error = sobind((struct socket *)fp->f_data, nam);
    141 	m_freem(nam);
    142 	FILE_UNUSE(fp, p);
    143 	return (error);
    144 }
    145 
    146 /* ARGSUSED */
    147 int
    148 sys_listen(p, v, retval)
    149 	struct proc *p;
    150 	void *v;
    151 	register_t *retval;
    152 {
    153 	struct sys_listen_args /* {
    154 		syscallarg(int) s;
    155 		syscallarg(int) backlog;
    156 	} */ *uap = v;
    157 	struct file *fp;
    158 	int error;
    159 
    160 	/* getsock() will use the descriptor for us */
    161 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    162 		return (error);
    163 	error = solisten((struct socket *)fp->f_data, SCARG(uap, backlog));
    164 	FILE_UNUSE(fp, p);
    165 	return (error);
    166 }
    167 
    168 int
    169 sys_accept(p, v, retval)
    170 	struct proc *p;
    171 	void *v;
    172 	register_t *retval;
    173 {
    174 	struct sys_accept_args /* {
    175 		syscallarg(int) s;
    176 		syscallarg(struct sockaddr *) name;
    177 		syscallarg(unsigned int *) anamelen;
    178 	} */ *uap = v;
    179 	struct filedesc *fdp = p->p_fd;
    180 	struct file *fp;
    181 	struct mbuf *nam;
    182 	unsigned int namelen;
    183 	int error, s, fd;
    184 	struct socket *so;
    185 
    186 	if (SCARG(uap, name) && (error = copyin((caddr_t)SCARG(uap, anamelen),
    187 	    (caddr_t)&namelen, sizeof(namelen))))
    188 		return (error);
    189 	if (SCARG(uap, name) != NULL &&
    190 	    uvm_useracc((caddr_t)SCARG(uap, name), sizeof(struct sockaddr),
    191 	     B_WRITE) == FALSE)
    192 		return (EFAULT);
    193 
    194 	/* getsock() will use the descriptor for us */
    195 	if ((error = getsock(fdp, SCARG(uap, s), &fp)) != 0)
    196 		return (error);
    197 	s = splsoftnet();
    198 	so = (struct socket *)fp->f_data;
    199 	FILE_UNUSE(fp, p);
    200 	if (!(so->so_proto->pr_flags & PR_LISTEN)) {
    201 		splx(s);
    202 		return (EOPNOTSUPP);
    203 	}
    204 	if ((so->so_options & SO_ACCEPTCONN) == 0) {
    205 		splx(s);
    206 		return (EINVAL);
    207 	}
    208 	if ((so->so_state & SS_NBIO) && so->so_qlen == 0) {
    209 		splx(s);
    210 		return (EWOULDBLOCK);
    211 	}
    212 	while (so->so_qlen == 0 && so->so_error == 0) {
    213 		if (so->so_state & SS_CANTRCVMORE) {
    214 			so->so_error = ECONNABORTED;
    215 			break;
    216 		}
    217 		error = tsleep((caddr_t)&so->so_timeo, PSOCK | PCATCH,
    218 			       netcon, 0);
    219 		if (error) {
    220 			splx(s);
    221 			return (error);
    222 		}
    223 	}
    224 	if (so->so_error) {
    225 		error = so->so_error;
    226 		so->so_error = 0;
    227 		splx(s);
    228 		return (error);
    229 	}
    230 	/* falloc() will use the descriptor for us */
    231 	if ((error = falloc(p, &fp, &fd)) != 0) {
    232 		splx(s);
    233 		return (error);
    234 	}
    235 	*retval = fd;
    236 	{ struct socket *aso = so->so_q.tqh_first;
    237 	  if (soqremque(aso, 1) == 0)
    238 		panic("accept");
    239 	  so = aso;
    240 	}
    241 	fp->f_type = DTYPE_SOCKET;
    242 	fp->f_flag = FREAD|FWRITE;
    243 	fp->f_ops = &socketops;
    244 	fp->f_data = (caddr_t)so;
    245 	FILE_UNUSE(fp, p);
    246 	nam = m_get(M_WAIT, MT_SONAME);
    247 	if ((error = soaccept(so, nam)) == 0 && SCARG(uap, name)) {
    248 		if (namelen > nam->m_len)
    249 			namelen = nam->m_len;
    250 		/* SHOULD COPY OUT A CHAIN HERE */
    251 		if ((error = copyout(mtod(nam, caddr_t),
    252 		    (caddr_t)SCARG(uap, name), namelen)) == 0)
    253 			error = copyout((caddr_t)&namelen,
    254 			    (caddr_t)SCARG(uap, anamelen),
    255 			    sizeof(*SCARG(uap, anamelen)));
    256 	}
    257 	/* if an error occured, free the file descriptor */
    258 	if (error) {
    259 		fdremove(fdp, fd);
    260 		ffree(fp);
    261 	}
    262 	m_freem(nam);
    263 	splx(s);
    264 	return (error);
    265 }
    266 
    267 /* ARGSUSED */
    268 int
    269 sys_connect(p, v, retval)
    270 	struct proc *p;
    271 	void *v;
    272 	register_t *retval;
    273 {
    274 	struct sys_connect_args /* {
    275 		syscallarg(int) s;
    276 		syscallarg(const struct sockaddr *) name;
    277 		syscallarg(unsigned int) namelen;
    278 	} */ *uap = v;
    279 	struct file *fp;
    280 	struct socket *so;
    281 	struct mbuf *nam;
    282 	int error, s;
    283 
    284 	/* getsock() will use the descriptor for us */
    285 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    286 		return (error);
    287 	so = (struct socket *)fp->f_data;
    288 	FILE_UNUSE(fp, p);
    289 	if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING))
    290 		return (EALREADY);
    291 	error = sockargs(&nam, SCARG(uap, name), SCARG(uap, namelen),
    292 	    MT_SONAME);
    293 	if (error)
    294 		return (error);
    295 	error = soconnect(so, nam);
    296 	if (error)
    297 		goto bad;
    298 	if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING)) {
    299 		m_freem(nam);
    300 		return (EINPROGRESS);
    301 	}
    302 	s = splsoftnet();
    303 	while ((so->so_state & SS_ISCONNECTING) && so->so_error == 0) {
    304 		error = tsleep((caddr_t)&so->so_timeo, PSOCK | PCATCH,
    305 			       netcon, 0);
    306 		if (error)
    307 			break;
    308 	}
    309 	if (error == 0) {
    310 		error = so->so_error;
    311 		so->so_error = 0;
    312 	}
    313 	splx(s);
    314 bad:
    315 	so->so_state &= ~SS_ISCONNECTING;
    316 	m_freem(nam);
    317 	if (error == ERESTART)
    318 		error = EINTR;
    319 	return (error);
    320 }
    321 
    322 int
    323 sys_socketpair(p, v, retval)
    324 	struct proc *p;
    325 	void *v;
    326 	register_t *retval;
    327 {
    328 	struct sys_socketpair_args /* {
    329 		syscallarg(int) domain;
    330 		syscallarg(int) type;
    331 		syscallarg(int) protocol;
    332 		syscallarg(int *) rsv;
    333 	} */ *uap = v;
    334 	struct filedesc *fdp = p->p_fd;
    335 	struct file *fp1, *fp2;
    336 	struct socket *so1, *so2;
    337 	int fd, error, sv[2];
    338 
    339 	error = socreate(SCARG(uap, domain), &so1, SCARG(uap, type),
    340 			 SCARG(uap, protocol));
    341 	if (error)
    342 		return (error);
    343 	error = socreate(SCARG(uap, domain), &so2, SCARG(uap, type),
    344 			 SCARG(uap, protocol));
    345 	if (error)
    346 		goto free1;
    347 	/* falloc() will use the descriptor for us */
    348 	if ((error = falloc(p, &fp1, &fd)) != 0)
    349 		goto free2;
    350 	sv[0] = fd;
    351 	fp1->f_flag = FREAD|FWRITE;
    352 	fp1->f_type = DTYPE_SOCKET;
    353 	fp1->f_ops = &socketops;
    354 	fp1->f_data = (caddr_t)so1;
    355 	if ((error = falloc(p, &fp2, &fd)) != 0)
    356 		goto free3;
    357 	fp2->f_flag = FREAD|FWRITE;
    358 	fp2->f_type = DTYPE_SOCKET;
    359 	fp2->f_ops = &socketops;
    360 	fp2->f_data = (caddr_t)so2;
    361 	sv[1] = fd;
    362 	if ((error = soconnect2(so1, so2)) != 0)
    363 		goto free4;
    364 	if (SCARG(uap, type) == SOCK_DGRAM) {
    365 		/*
    366 		 * Datagram socket connection is asymmetric.
    367 		 */
    368 		 if ((error = soconnect2(so2, so1)) != 0)
    369 			goto free4;
    370 	}
    371 	error = copyout((caddr_t)sv, (caddr_t)SCARG(uap, rsv),
    372 	    2 * sizeof(int));
    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(p, v, retval)
    393 	struct proc *p;
    394 	void *v;
    395 	register_t *retval;
    396 {
    397 	struct sys_sendto_args /* {
    398 		syscallarg(int) s;
    399 		syscallarg(const void *) buf;
    400 		syscallarg(size_t) len;
    401 		syscallarg(int) flags;
    402 		syscallarg(const struct sockaddr *) to;
    403 		syscallarg(unsigned int) tolen;
    404 	} */ *uap = v;
    405 	struct msghdr msg;
    406 	struct iovec aiov;
    407 
    408 	msg.msg_name = (caddr_t)SCARG(uap, to);		/* XXX kills const */
    409 	msg.msg_namelen = SCARG(uap, tolen);
    410 	msg.msg_iov = &aiov;
    411 	msg.msg_iovlen = 1;
    412 	msg.msg_control = 0;
    413 #ifdef COMPAT_OLDSOCK
    414 	msg.msg_flags = 0;
    415 #endif
    416 	aiov.iov_base = (char *)SCARG(uap, buf);	/* XXX kills const */
    417 	aiov.iov_len = SCARG(uap, len);
    418 	return (sendit(p, SCARG(uap, s), &msg, SCARG(uap, flags), retval));
    419 }
    420 
    421 int
    422 sys_sendmsg(p, v, retval)
    423 	struct proc *p;
    424 	void *v;
    425 	register_t *retval;
    426 {
    427 	struct sys_sendmsg_args /* {
    428 		syscallarg(int) s;
    429 		syscallarg(const struct msghdr *) msg;
    430 		syscallarg(int) flags;
    431 	} */ *uap = v;
    432 	struct msghdr msg;
    433 	struct iovec aiov[UIO_SMALLIOV], *iov;
    434 	int error;
    435 
    436 	error = copyin(SCARG(uap, msg), (caddr_t)&msg, sizeof(msg));
    437 	if (error)
    438 		return (error);
    439 	if ((unsigned int)msg.msg_iovlen > UIO_SMALLIOV) {
    440 		if ((unsigned int)msg.msg_iovlen > IOV_MAX)
    441 			return (EMSGSIZE);
    442 		iov = malloc(sizeof(struct iovec) * msg.msg_iovlen,
    443 		    M_IOV, M_WAITOK);
    444 	} else
    445 		iov = aiov;
    446 	if ((unsigned int)msg.msg_iovlen > 0) {
    447 		error = copyin((caddr_t)msg.msg_iov, (caddr_t)iov,
    448 		    (size_t)(msg.msg_iovlen * sizeof(struct iovec)));
    449 		if (error)
    450 			goto done;
    451 	}
    452 	msg.msg_iov = iov;
    453 #ifdef COMPAT_OLDSOCK
    454 	msg.msg_flags = 0;
    455 #endif
    456 	error = sendit(p, SCARG(uap, s), &msg, SCARG(uap, flags), retval);
    457 done:
    458 	if (iov != aiov)
    459 		free(iov, M_IOV);
    460 	return (error);
    461 }
    462 
    463 int
    464 sendit(p, s, mp, flags, retsize)
    465 	struct proc *p;
    466 	int s;
    467 	struct msghdr *mp;
    468 	int flags;
    469 	register_t *retsize;
    470 {
    471 	struct file *fp;
    472 	struct uio auio;
    473 	struct iovec *iov;
    474 	int i;
    475 	struct mbuf *to, *control;
    476 	int len, error;
    477 	struct socket *so;
    478 #ifdef KTRACE
    479 	struct iovec *ktriov = NULL;
    480 #endif
    481 
    482 	/* getsock() will use the descriptor for us */
    483 	if ((error = getsock(p->p_fd, s, &fp)) != 0)
    484 		return (error);
    485 	auio.uio_iov = mp->msg_iov;
    486 	auio.uio_iovcnt = mp->msg_iovlen;
    487 	auio.uio_segflg = UIO_USERSPACE;
    488 	auio.uio_rw = UIO_WRITE;
    489 	auio.uio_procp = p;
    490 	auio.uio_offset = 0;			/* XXX */
    491 	auio.uio_resid = 0;
    492 	iov = mp->msg_iov;
    493 	for (i = 0; i < mp->msg_iovlen; i++, iov++) {
    494 #if 0
    495 		/* cannot happen; iov_len is unsigned */
    496 		if (iov->iov_len < 0) {
    497 			error = EINVAL;
    498 			goto out;
    499 		}
    500 #endif
    501 		/*
    502 		 * Writes return ssize_t because -1 is returned on error.
    503 		 * Therefore, we must restrict the length to SSIZE_MAX to
    504 		 * avoid garbage return values.
    505 		 */
    506 		auio.uio_resid += iov->iov_len;
    507 		if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
    508 			error = EINVAL;
    509 			goto out;
    510 		}
    511 	}
    512 	if (mp->msg_name) {
    513 		error = sockargs(&to, mp->msg_name, mp->msg_namelen,
    514 				 MT_SONAME);
    515 		if (error)
    516 			goto out;
    517 	} else
    518 		to = 0;
    519 	if (mp->msg_control) {
    520 		if (mp->msg_controllen < sizeof(struct cmsghdr)
    521 #ifdef COMPAT_OLDSOCK
    522 		    && mp->msg_flags != MSG_COMPAT
    523 #endif
    524 		) {
    525 			error = EINVAL;
    526 			goto bad;
    527 		}
    528 		error = sockargs(&control, mp->msg_control,
    529 				 mp->msg_controllen, MT_CONTROL);
    530 		if (error)
    531 			goto bad;
    532 #ifdef COMPAT_OLDSOCK
    533 		if (mp->msg_flags == MSG_COMPAT) {
    534 			struct cmsghdr *cm;
    535 
    536 			M_PREPEND(control, sizeof(*cm), M_WAIT);
    537 			if (control == 0) {
    538 				error = ENOBUFS;
    539 				goto bad;
    540 			} else {
    541 				cm = mtod(control, struct cmsghdr *);
    542 				cm->cmsg_len = control->m_len;
    543 				cm->cmsg_level = SOL_SOCKET;
    544 				cm->cmsg_type = SCM_RIGHTS;
    545 			}
    546 		}
    547 #endif
    548 	} else
    549 		control = 0;
    550 #ifdef KTRACE
    551 	if (KTRPOINT(p, KTR_GENIO)) {
    552 		int iovlen = auio.uio_iovcnt * sizeof(struct iovec);
    553 
    554 		ktriov = malloc(iovlen, M_TEMP, M_WAITOK);
    555 		memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen);
    556 	}
    557 #endif
    558 	len = auio.uio_resid;
    559 	so = (struct socket *)fp->f_data;
    560 	error = (*so->so_send)(so, to, &auio, NULL, control, flags);
    561 	if (error) {
    562 		if (auio.uio_resid != len && (error == ERESTART ||
    563 		    error == EINTR || error == EWOULDBLOCK))
    564 			error = 0;
    565 		if (error == EPIPE)
    566 			psignal(p, SIGPIPE);
    567 	}
    568 	if (error == 0)
    569 		*retsize = len - auio.uio_resid;
    570 #ifdef KTRACE
    571 	if (ktriov != NULL) {
    572 		if (error == 0)
    573 			ktrgenio(p, s, UIO_WRITE, ktriov, *retsize, error);
    574 		free(ktriov, M_TEMP);
    575 	}
    576 #endif
    577  bad:
    578 	if (to)
    579 		m_freem(to);
    580  out:
    581 	FILE_UNUSE(fp, p);
    582 	return (error);
    583 }
    584 
    585 int
    586 sys_recvfrom(p, v, retval)
    587 	struct proc *p;
    588 	void *v;
    589 	register_t *retval;
    590 {
    591 	struct sys_recvfrom_args /* {
    592 		syscallarg(int) s;
    593 		syscallarg(void *) buf;
    594 		syscallarg(size_t) len;
    595 		syscallarg(int) flags;
    596 		syscallarg(struct sockaddr *) from;
    597 		syscallarg(unsigned int *) fromlenaddr;
    598 	} */ *uap = v;
    599 	struct msghdr msg;
    600 	struct iovec aiov;
    601 	int error;
    602 
    603 	if (SCARG(uap, fromlenaddr)) {
    604 		error = copyin((caddr_t)SCARG(uap, fromlenaddr),
    605 			       (caddr_t)&msg.msg_namelen,
    606 			       sizeof(msg.msg_namelen));
    607 		if (error)
    608 			return (error);
    609 	} else
    610 		msg.msg_namelen = 0;
    611 	msg.msg_name = (caddr_t)SCARG(uap, from);
    612 	msg.msg_iov = &aiov;
    613 	msg.msg_iovlen = 1;
    614 	aiov.iov_base = SCARG(uap, buf);
    615 	aiov.iov_len = SCARG(uap, len);
    616 	msg.msg_control = 0;
    617 	msg.msg_flags = SCARG(uap, flags);
    618 	return (recvit(p, SCARG(uap, s), &msg,
    619 		       (caddr_t)SCARG(uap, fromlenaddr), retval));
    620 }
    621 
    622 int
    623 sys_recvmsg(p, v, retval)
    624 	struct proc *p;
    625 	void *v;
    626 	register_t *retval;
    627 {
    628 	struct sys_recvmsg_args /* {
    629 		syscallarg(int) s;
    630 		syscallarg(struct msghdr *) msg;
    631 		syscallarg(int) flags;
    632 	} */ *uap = v;
    633 	struct msghdr msg;
    634 	struct iovec aiov[UIO_SMALLIOV], *uiov, *iov;
    635 	int error;
    636 
    637 	error = copyin((caddr_t)SCARG(uap, msg), (caddr_t)&msg,
    638 		       sizeof(msg));
    639 	if (error)
    640 		return (error);
    641 	if ((unsigned int)msg.msg_iovlen > UIO_SMALLIOV) {
    642 		if ((unsigned int)msg.msg_iovlen > IOV_MAX)
    643 			return (EMSGSIZE);
    644 		iov = malloc(sizeof(struct iovec) * msg.msg_iovlen,
    645 		    M_IOV, M_WAITOK);
    646 	} else
    647 		iov = aiov;
    648 	if ((unsigned int)msg.msg_iovlen > 0) {
    649 		error = copyin((caddr_t)msg.msg_iov, (caddr_t)iov,
    650 		    (size_t)(msg.msg_iovlen * sizeof(struct iovec)));
    651 		if (error)
    652 			goto done;
    653 	}
    654 	uiov = msg.msg_iov;
    655 	msg.msg_iov = iov;
    656 #ifdef COMPAT_OLDSOCK
    657 	msg.msg_flags = SCARG(uap, flags) &~ MSG_COMPAT;
    658 #else
    659 	msg.msg_flags = SCARG(uap, flags);
    660 #endif
    661 	if ((error = recvit(p, SCARG(uap, s), &msg, (caddr_t)0, retval)) == 0) {
    662 		msg.msg_iov = uiov;
    663 		error = copyout((caddr_t)&msg, (caddr_t)SCARG(uap, msg),
    664 		    sizeof(msg));
    665 	}
    666 done:
    667 	if (iov != aiov)
    668 		free(iov, M_IOV);
    669 	return (error);
    670 }
    671 
    672 int
    673 recvit(p, s, mp, namelenp, retsize)
    674 	struct proc *p;
    675 	int s;
    676 	struct msghdr *mp;
    677 	caddr_t namelenp;
    678 	register_t *retsize;
    679 {
    680 	struct file *fp;
    681 	struct uio auio;
    682 	struct iovec *iov;
    683 	int i;
    684 	int len, error;
    685 	struct mbuf *from = 0, *control = 0;
    686 	struct socket *so;
    687 #ifdef KTRACE
    688 	struct iovec *ktriov = NULL;
    689 #endif
    690 
    691 	/* getsock() will use the descriptor for us */
    692 	if ((error = getsock(p->p_fd, s, &fp)) != 0)
    693 		return (error);
    694 	auio.uio_iov = mp->msg_iov;
    695 	auio.uio_iovcnt = mp->msg_iovlen;
    696 	auio.uio_segflg = UIO_USERSPACE;
    697 	auio.uio_rw = UIO_READ;
    698 	auio.uio_procp = p;
    699 	auio.uio_offset = 0;			/* XXX */
    700 	auio.uio_resid = 0;
    701 	iov = mp->msg_iov;
    702 	for (i = 0; i < mp->msg_iovlen; i++, iov++) {
    703 #if 0
    704 		/* cannot happen iov_len is unsigned */
    705 		if (iov->iov_len < 0) {
    706 			error = EINVAL;
    707 			goto out1;
    708 		}
    709 #endif
    710 		/*
    711 		 * Reads return ssize_t because -1 is returned on error.
    712 		 * Therefore we must restrict the length to SSIZE_MAX to
    713 		 * avoid garbage return values.
    714 		 */
    715 		auio.uio_resid += iov->iov_len;
    716 		if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
    717 			error = EINVAL;
    718 			goto out1;
    719 		}
    720 	}
    721 #ifdef KTRACE
    722 	if (KTRPOINT(p, KTR_GENIO)) {
    723 		int iovlen = auio.uio_iovcnt * sizeof(struct iovec);
    724 
    725 		ktriov = malloc(iovlen, M_TEMP, M_WAITOK);
    726 		memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen);
    727 	}
    728 #endif
    729 	len = auio.uio_resid;
    730 	so = (struct socket *)fp->f_data;
    731 	error = (*so->so_receive)(so, &from, &auio, NULL,
    732 			  mp->msg_control ? &control : NULL, &mp->msg_flags);
    733 	if (error) {
    734 		if (auio.uio_resid != len && (error == ERESTART ||
    735 		    error == EINTR || error == EWOULDBLOCK))
    736 			error = 0;
    737 	}
    738 #ifdef KTRACE
    739 	if (ktriov != NULL) {
    740 		if (error == 0)
    741 			ktrgenio(p, s, UIO_READ, ktriov,
    742 			    len - auio.uio_resid, error);
    743 		free(ktriov, M_TEMP);
    744 	}
    745 #endif
    746 	if (error)
    747 		goto out;
    748 	*retsize = len - auio.uio_resid;
    749 	if (mp->msg_name) {
    750 		len = mp->msg_namelen;
    751 		if (len <= 0 || from == 0)
    752 			len = 0;
    753 		else {
    754 #ifdef COMPAT_OLDSOCK
    755 			if (mp->msg_flags & MSG_COMPAT)
    756 				mtod(from, struct osockaddr *)->sa_family =
    757 				    mtod(from, struct sockaddr *)->sa_family;
    758 #endif
    759 			if (len > from->m_len)
    760 				len = from->m_len;
    761 			/* else if len < from->m_len ??? */
    762 			error = copyout(mtod(from, caddr_t),
    763 					(caddr_t)mp->msg_name, (unsigned)len);
    764 			if (error)
    765 				goto out;
    766 		}
    767 		mp->msg_namelen = len;
    768 		if (namelenp &&
    769 		    (error = copyout((caddr_t)&len, namelenp, sizeof(int)))) {
    770 #ifdef COMPAT_OLDSOCK
    771 			if (mp->msg_flags & MSG_COMPAT)
    772 				error = 0;	/* old recvfrom didn't check */
    773 			else
    774 #endif
    775 			goto out;
    776 		}
    777 	}
    778 	if (mp->msg_control) {
    779 #ifdef COMPAT_OLDSOCK
    780 		/*
    781 		 * We assume that old recvmsg calls won't receive access
    782 		 * rights and other control info, esp. as control info
    783 		 * is always optional and those options didn't exist in 4.3.
    784 		 * If we receive rights, trim the cmsghdr; anything else
    785 		 * is tossed.
    786 		 */
    787 		if (control && mp->msg_flags & MSG_COMPAT) {
    788 			if (mtod(control, struct cmsghdr *)->cmsg_level !=
    789 			    SOL_SOCKET ||
    790 			    mtod(control, struct cmsghdr *)->cmsg_type !=
    791 			    SCM_RIGHTS) {
    792 				mp->msg_controllen = 0;
    793 				goto out;
    794 			}
    795 			control->m_len -= sizeof(struct cmsghdr);
    796 			control->m_data += sizeof(struct cmsghdr);
    797 		}
    798 #endif
    799 		len = mp->msg_controllen;
    800 		if (len <= 0 || control == 0)
    801 			len = 0;
    802 		else {
    803 			struct mbuf *m = control;
    804 			caddr_t p = (caddr_t)mp->msg_control;
    805 
    806 			do {
    807 				i = m->m_len;
    808 				if (len < i) {
    809 					mp->msg_flags |= MSG_CTRUNC;
    810 					i = len;
    811 				}
    812 				error = copyout(mtod(m, caddr_t), p,
    813 				    (unsigned)i);
    814 				if (m->m_next)
    815 					i = ALIGN(i);
    816 				p += i;
    817 				len -= i;
    818 				if (error != 0 || len <= 0)
    819 					break;
    820 			} while ((m = m->m_next) != NULL);
    821 			len = p - (caddr_t)mp->msg_control;
    822 		}
    823 		mp->msg_controllen = len;
    824 	}
    825  out:
    826 	if (from)
    827 		m_freem(from);
    828 	if (control)
    829 		m_freem(control);
    830  out1:
    831 	FILE_UNUSE(fp, p);
    832 	return (error);
    833 }
    834 
    835 /* ARGSUSED */
    836 int
    837 sys_shutdown(p, v, retval)
    838 	struct proc *p;
    839 	void *v;
    840 	register_t *retval;
    841 {
    842 	struct sys_shutdown_args /* {
    843 		syscallarg(int) s;
    844 		syscallarg(int) how;
    845 	} */ *uap = v;
    846 	struct file *fp;
    847 	int error;
    848 
    849 	/* getsock() will use the descriptor for us */
    850 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    851 		return (error);
    852 	error = soshutdown((struct socket *)fp->f_data, SCARG(uap, how));
    853 	FILE_UNUSE(fp, p);
    854 	return (error);
    855 }
    856 
    857 /* ARGSUSED */
    858 int
    859 sys_setsockopt(p, v, retval)
    860 	struct proc *p;
    861 	void *v;
    862 	register_t *retval;
    863 {
    864 	struct sys_setsockopt_args /* {
    865 		syscallarg(int) s;
    866 		syscallarg(int) level;
    867 		syscallarg(int) name;
    868 		syscallarg(const void *) val;
    869 		syscallarg(unsigned int) valsize;
    870 	} */ *uap = v;
    871 	struct file *fp;
    872 	struct mbuf *m = NULL;
    873 	int error;
    874 
    875 	/* getsock() will use the descriptor for us */
    876 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    877 		return (error);
    878 	if (SCARG(uap, valsize) > MLEN) {
    879 		error = EINVAL;
    880 		goto out;
    881 	}
    882 	if (SCARG(uap, val)) {
    883 		m = m_get(M_WAIT, MT_SOOPTS);
    884 		error = copyin(SCARG(uap, val), mtod(m, caddr_t),
    885 			       SCARG(uap, valsize));
    886 		if (error) {
    887 			(void) m_free(m);
    888 			goto out;
    889 		}
    890 		m->m_len = SCARG(uap, valsize);
    891 	}
    892 	error = sosetopt((struct socket *)fp->f_data, SCARG(uap, level),
    893 			 SCARG(uap, name), m);
    894  out:
    895 	FILE_UNUSE(fp, p);
    896 	return (error);
    897 }
    898 
    899 /* ARGSUSED */
    900 int
    901 sys_getsockopt(p, v, retval)
    902 	struct proc *p;
    903 	void *v;
    904 	register_t *retval;
    905 {
    906 	struct sys_getsockopt_args /* {
    907 		syscallarg(int) s;
    908 		syscallarg(int) level;
    909 		syscallarg(int) name;
    910 		syscallarg(void *) val;
    911 		syscallarg(unsigned int *) avalsize;
    912 	} */ *uap = v;
    913 	struct file *fp;
    914 	struct mbuf *m = NULL, *m0;
    915 	unsigned int op, i, valsize;
    916 	int error;
    917 
    918 	/* getsock() will use the descriptor for us */
    919 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    920 		return (error);
    921 	if (SCARG(uap, val)) {
    922 		error = copyin((caddr_t)SCARG(uap, avalsize),
    923 			       (caddr_t)&valsize, sizeof(valsize));
    924 		if (error)
    925 			goto out;
    926 	} else
    927 		valsize = 0;
    928 	if ((error = sogetopt((struct socket *)fp->f_data, SCARG(uap, level),
    929 	    SCARG(uap, name), &m)) == 0 && SCARG(uap, val) && valsize &&
    930 	    m != NULL) {
    931 		op = 0;
    932 		while (m && !error && op < valsize) {
    933 			i = min(m->m_len, (valsize - op));
    934 			error = copyout(mtod(m, caddr_t), SCARG(uap, val), i);
    935 			op += i;
    936 			SCARG(uap, val) = ((u_int8_t *)SCARG(uap, val)) + i;
    937 			m0 = m;
    938 			MFREE(m0, m);
    939 		}
    940 		valsize = op;
    941 		if (error == 0)
    942 			error = copyout(&valsize,
    943 					SCARG(uap, avalsize), sizeof(valsize));
    944 	}
    945 	if (m != NULL)
    946 		(void) m_free(m);
    947  out:
    948 	FILE_UNUSE(fp, p);
    949 	return (error);
    950 }
    951 
    952 /* ARGSUSED */
    953 int
    954 sys_pipe(p, v, retval)
    955 	struct proc *p;
    956 	void *v;
    957 	register_t *retval;
    958 {
    959 	struct filedesc *fdp = p->p_fd;
    960 	struct file *rf, *wf;
    961 	struct socket *rso, *wso;
    962 	int fd, error;
    963 
    964 	if ((error = socreate(AF_LOCAL, &rso, SOCK_STREAM, 0)) != 0)
    965 		return (error);
    966 	if ((error = socreate(AF_LOCAL, &wso, SOCK_STREAM, 0)) != 0)
    967 		goto free1;
    968 	/* falloc() will use the descriptor for us */
    969 	if ((error = falloc(p, &rf, &fd)) != 0)
    970 		goto free2;
    971 	retval[0] = fd;
    972 	rf->f_flag = FREAD;
    973 	rf->f_type = DTYPE_SOCKET;
    974 	rf->f_ops = &socketops;
    975 	rf->f_data = (caddr_t)rso;
    976 	if ((error = falloc(p, &wf, &fd)) != 0)
    977 		goto free3;
    978 	wf->f_flag = FWRITE;
    979 	wf->f_type = DTYPE_SOCKET;
    980 	wf->f_ops = &socketops;
    981 	wf->f_data = (caddr_t)wso;
    982 	retval[1] = fd;
    983 	if ((error = unp_connect2(wso, rso)) != 0)
    984 		goto free4;
    985 	FILE_UNUSE(rf, p);
    986 	FILE_UNUSE(wf, p);
    987 	return (0);
    988 free4:
    989 	FILE_UNUSE(wf, p);
    990 	ffree(wf);
    991 	fdremove(fdp, retval[1]);
    992 free3:
    993 	FILE_UNUSE(rf, p);
    994 	ffree(rf);
    995 	fdremove(fdp, retval[0]);
    996 free2:
    997 	(void)soclose(wso);
    998 free1:
    999 	(void)soclose(rso);
   1000 	return (error);
   1001 }
   1002 
   1003 /*
   1004  * Get socket name.
   1005  */
   1006 /* ARGSUSED */
   1007 int
   1008 sys_getsockname(p, v, retval)
   1009 	struct proc *p;
   1010 	void *v;
   1011 	register_t *retval;
   1012 {
   1013 	struct sys_getsockname_args /* {
   1014 		syscallarg(int) fdes;
   1015 		syscallarg(struct sockaddr *) asa;
   1016 		syscallarg(unsigned int *) alen;
   1017 	} */ *uap = v;
   1018 	struct file *fp;
   1019 	struct socket *so;
   1020 	struct mbuf *m;
   1021 	unsigned int len;
   1022 	int error;
   1023 
   1024 	/* getsock() will use the descriptor for us */
   1025 	if ((error = getsock(p->p_fd, SCARG(uap, fdes), &fp)) != 0)
   1026 		return (error);
   1027 	error = copyin((caddr_t)SCARG(uap, alen), (caddr_t)&len, sizeof(len));
   1028 	if (error)
   1029 		goto out;
   1030 	so = (struct socket *)fp->f_data;
   1031 	m = m_getclr(M_WAIT, MT_SONAME);
   1032 	error = (*so->so_proto->pr_usrreq)(so, PRU_SOCKADDR, (struct mbuf *)0,
   1033 	    m, (struct mbuf *)0, (struct proc *)0);
   1034 	if (error)
   1035 		goto bad;
   1036 	if (len > m->m_len)
   1037 		len = m->m_len;
   1038 	error = copyout(mtod(m, caddr_t), (caddr_t)SCARG(uap, asa), len);
   1039 	if (error == 0)
   1040 		error = copyout((caddr_t)&len, (caddr_t)SCARG(uap, alen),
   1041 		    sizeof(len));
   1042  bad:
   1043 	m_freem(m);
   1044  out:
   1045 	FILE_UNUSE(fp, p);
   1046 	return (error);
   1047 }
   1048 
   1049 /*
   1050  * Get name of peer for connected socket.
   1051  */
   1052 /* ARGSUSED */
   1053 int
   1054 sys_getpeername(p, v, retval)
   1055 	struct proc *p;
   1056 	void *v;
   1057 	register_t *retval;
   1058 {
   1059 	struct sys_getpeername_args /* {
   1060 		syscallarg(int) fdes;
   1061 		syscallarg(struct sockaddr *) asa;
   1062 		syscallarg(unsigned int *) alen;
   1063 	} */ *uap = v;
   1064 	struct file *fp;
   1065 	struct socket *so;
   1066 	struct mbuf *m;
   1067 	unsigned int len;
   1068 	int error;
   1069 
   1070 	/* getsock() will use the descriptor for us */
   1071 	if ((error = getsock(p->p_fd, SCARG(uap, fdes), &fp)) != 0)
   1072 		return (error);
   1073 	so = (struct socket *)fp->f_data;
   1074 	if ((so->so_state & (SS_ISCONNECTED|SS_ISCONFIRMING)) == 0) {
   1075 		error = ENOTCONN;
   1076 		goto out;
   1077 	}
   1078 	error = copyin((caddr_t)SCARG(uap, alen), (caddr_t)&len, sizeof(len));
   1079 	if (error)
   1080 		goto out;
   1081 	m = m_getclr(M_WAIT, MT_SONAME);
   1082 	error = (*so->so_proto->pr_usrreq)(so, PRU_PEERADDR, (struct mbuf *)0,
   1083 	    m, (struct mbuf *)0, (struct proc *)0);
   1084 	if (error)
   1085 		goto bad;
   1086 	if (len > m->m_len)
   1087 		len = m->m_len;
   1088 	error = copyout(mtod(m, caddr_t), (caddr_t)SCARG(uap, asa), len);
   1089 	if (error)
   1090 		goto bad;
   1091 	error = copyout((caddr_t)&len, (caddr_t)SCARG(uap, alen), sizeof(len));
   1092  bad:
   1093 	m_freem(m);
   1094  out:
   1095 	FILE_UNUSE(fp, p);
   1096 	return (error);
   1097 }
   1098 
   1099 /*
   1100  * XXX In a perfect world, we wouldn't pass around socket control
   1101  * XXX arguments in mbufs, and this could go away.
   1102  */
   1103 int
   1104 sockargs(mp, buf, buflen, type)
   1105 	struct mbuf **mp;
   1106 	const void *buf;
   1107 	int buflen, type;
   1108 {
   1109 	struct sockaddr *sa;
   1110 	struct mbuf *m;
   1111 	int error;
   1112 
   1113 	/*
   1114 	 * We can't allow socket names > UCHAR_MAX in length, since that
   1115 	 * will overflow sa_len.
   1116 	 */
   1117 	if (type == MT_SONAME && (u_int)buflen > UCHAR_MAX)
   1118 		return (EINVAL);
   1119 
   1120 	/* Allocate an mbuf to hold the arguments. */
   1121 	m = m_get(M_WAIT, type);
   1122 	if ((u_int)buflen > MLEN) {
   1123 		/*
   1124 		 * Won't fit into a regular mbuf, so we allocate just
   1125 		 * enough external storage to hold the argument.
   1126 		 */
   1127 		MEXTMALLOC(m, buflen, M_WAITOK);
   1128 	}
   1129 	m->m_len = buflen;
   1130 	error = copyin(buf, mtod(m, caddr_t), (u_int)buflen);
   1131 	if (error) {
   1132 		(void) m_free(m);
   1133 		return (error);
   1134 	}
   1135 	*mp = m;
   1136 	if (type == MT_SONAME) {
   1137 		sa = mtod(m, struct sockaddr *);
   1138 
   1139 #if defined(COMPAT_OLDSOCK) && BYTE_ORDER != BIG_ENDIAN
   1140 		if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
   1141 			sa->sa_family = sa->sa_len;
   1142 #endif
   1143 		sa->sa_len = buflen;
   1144 	}
   1145 	return (0);
   1146 }
   1147 
   1148 int
   1149 getsock(fdp, fdes, fpp)
   1150 	struct filedesc *fdp;
   1151 	int fdes;
   1152 	struct file **fpp;
   1153 {
   1154 	struct file *fp;
   1155 
   1156 	if ((unsigned)fdes >= fdp->fd_nfiles ||
   1157 	    (fp = fdp->fd_ofiles[fdes]) == NULL ||
   1158 	    (fp->f_iflags & FIF_WANTCLOSE) != 0)
   1159 		return (EBADF);
   1160 
   1161 	FILE_USE(fp);
   1162 
   1163 	if (fp->f_type != DTYPE_SOCKET) {
   1164 		FILE_UNUSE(fp, NULL);
   1165 		return (ENOTSOCK);
   1166 	}
   1167 	*fpp = fp;
   1168 	return (0);
   1169 }
   1170