Home | History | Annotate | Line # | Download | only in kern
uipc_syscalls.c revision 1.57
      1 /*	$NetBSD: uipc_syscalls.c,v 1.57 2001/02/27 05:19:15 lukem 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 /*
     41  * Though COMPAT_OLDSOCK is needed only for COMPAT_43, SunOS, Linux,
     42  * HP-UX, FreeBSD, Ultrix, OSF1, we define it unconditionally so that
     43  * this would be LKM-safe.
     44  */
     45 #define COMPAT_OLDSOCK		/* used by <sys/socket.h> */
     46 
     47 #include <sys/param.h>
     48 #include <sys/systm.h>
     49 #include <sys/filedesc.h>
     50 #include <sys/proc.h>
     51 #include <sys/file.h>
     52 #include <sys/buf.h>
     53 #include <sys/malloc.h>
     54 #include <sys/mbuf.h>
     55 #include <sys/protosw.h>
     56 #include <sys/socket.h>
     57 #include <sys/socketvar.h>
     58 #include <sys/signalvar.h>
     59 #include <sys/un.h>
     60 #ifdef KTRACE
     61 #include <sys/ktrace.h>
     62 #endif
     63 
     64 #include <sys/mount.h>
     65 #include <sys/syscallargs.h>
     66 
     67 #include <uvm/uvm_extern.h>
     68 
     69 /*
     70  * System call interface to the socket abstraction.
     71  */
     72 extern	struct fileops socketops;
     73 
     74 int
     75 sys_socket(struct proc *p, void *v, register_t *retval)
     76 {
     77 	struct sys_socket_args /* {
     78 		syscallarg(int)	domain;
     79 		syscallarg(int)	type;
     80 		syscallarg(int)	protocol;
     81 	} */ *uap = v;
     82 	struct filedesc	*fdp;
     83 	struct socket	*so;
     84 	struct file	*fp;
     85 	int		fd, error;
     86 
     87 	fdp = p->p_fd;
     88 	/* falloc() will use the desciptor for us */
     89 	if ((error = falloc(p, &fp, &fd)) != 0)
     90 		return (error);
     91 	fp->f_flag = FREAD|FWRITE;
     92 	fp->f_type = DTYPE_SOCKET;
     93 	fp->f_ops = &socketops;
     94 	error = socreate(SCARG(uap, domain), &so, SCARG(uap, type),
     95 			 SCARG(uap, protocol));
     96 	if (error) {
     97 		FILE_UNUSE(fp, p);
     98 		fdremove(fdp, fd);
     99 		ffree(fp);
    100 	} else {
    101 		fp->f_data = (caddr_t)so;
    102 		FILE_UNUSE(fp, p);
    103 		*retval = fd;
    104 	}
    105 	return (error);
    106 }
    107 
    108 /* ARGSUSED */
    109 int
    110 sys_bind(struct proc *p, void *v, register_t *retval)
    111 {
    112 	struct sys_bind_args /* {
    113 		syscallarg(int)				s;
    114 		syscallarg(const struct sockaddr *)	name;
    115 		syscallarg(unsigned int)		namelen;
    116 	} */ *uap = v;
    117 	struct file	*fp;
    118 	struct mbuf	*nam;
    119 	int		error;
    120 
    121 	/* getsock() will use the descriptor for us */
    122 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    123 		return (error);
    124 	error = sockargs(&nam, SCARG(uap, name), SCARG(uap, namelen),
    125 	    MT_SONAME);
    126 	if (error) {
    127 		FILE_UNUSE(fp, p);
    128 		return (error);
    129 	}
    130 	error = sobind((struct socket *)fp->f_data, nam, p);
    131 	m_freem(nam);
    132 	FILE_UNUSE(fp, p);
    133 	return (error);
    134 }
    135 
    136 /* ARGSUSED */
    137 int
    138 sys_listen(struct proc *p, void *v, register_t *retval)
    139 {
    140 	struct sys_listen_args /* {
    141 		syscallarg(int)	s;
    142 		syscallarg(int)	backlog;
    143 	} */ *uap = v;
    144 	struct file	*fp;
    145 	int		error;
    146 
    147 	/* getsock() will use the descriptor for us */
    148 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    149 		return (error);
    150 	error = solisten((struct socket *)fp->f_data, SCARG(uap, backlog));
    151 	FILE_UNUSE(fp, p);
    152 	return (error);
    153 }
    154 
    155 int
    156 sys_accept(struct proc *p, void *v, register_t *retval)
    157 {
    158 	struct sys_accept_args /* {
    159 		syscallarg(int)			s;
    160 		syscallarg(struct sockaddr *)	name;
    161 		syscallarg(unsigned int *)	anamelen;
    162 	} */ *uap = v;
    163 	struct filedesc	*fdp;
    164 	struct file	*fp;
    165 	struct mbuf	*nam;
    166 	unsigned int	namelen;
    167 	int		error, s, fd;
    168 	struct socket	*so;
    169 
    170 	fdp = p->p_fd;
    171 	if (SCARG(uap, name) && (error = copyin((caddr_t)SCARG(uap, anamelen),
    172 	    (caddr_t)&namelen, sizeof(namelen))))
    173 		return (error);
    174 	if (SCARG(uap, name) != NULL &&
    175 	    uvm_useracc((caddr_t)SCARG(uap, name), sizeof(struct sockaddr),
    176 	     B_WRITE) == FALSE)
    177 		return (EFAULT);
    178 
    179 	/* getsock() will use the descriptor for us */
    180 	if ((error = getsock(fdp, SCARG(uap, s), &fp)) != 0)
    181 		return (error);
    182 	s = splsoftnet();
    183 	so = (struct socket *)fp->f_data;
    184 	FILE_UNUSE(fp, p);
    185 	if (!(so->so_proto->pr_flags & PR_LISTEN)) {
    186 		splx(s);
    187 		return (EOPNOTSUPP);
    188 	}
    189 	if ((so->so_options & SO_ACCEPTCONN) == 0) {
    190 		splx(s);
    191 		return (EINVAL);
    192 	}
    193 	if ((so->so_state & SS_NBIO) && so->so_qlen == 0) {
    194 		splx(s);
    195 		return (EWOULDBLOCK);
    196 	}
    197 	while (so->so_qlen == 0 && so->so_error == 0) {
    198 		if (so->so_state & SS_CANTRCVMORE) {
    199 			so->so_error = ECONNABORTED;
    200 			break;
    201 		}
    202 		error = tsleep((caddr_t)&so->so_timeo, PSOCK | PCATCH,
    203 			       netcon, 0);
    204 		if (error) {
    205 			splx(s);
    206 			return (error);
    207 		}
    208 	}
    209 	if (so->so_error) {
    210 		error = so->so_error;
    211 		so->so_error = 0;
    212 		splx(s);
    213 		return (error);
    214 	}
    215 	/* falloc() will use the descriptor for us */
    216 	if ((error = falloc(p, &fp, &fd)) != 0) {
    217 		splx(s);
    218 		return (error);
    219 	}
    220 	*retval = fd;
    221 	{ struct socket *aso = so->so_q.tqh_first;
    222 	  if (soqremque(aso, 1) == 0)
    223 		panic("accept");
    224 	  so = aso;
    225 	}
    226 	fp->f_type = DTYPE_SOCKET;
    227 	fp->f_flag = FREAD|FWRITE;
    228 	fp->f_ops = &socketops;
    229 	fp->f_data = (caddr_t)so;
    230 	FILE_UNUSE(fp, p);
    231 	nam = m_get(M_WAIT, MT_SONAME);
    232 	if ((error = soaccept(so, nam)) == 0 && SCARG(uap, name)) {
    233 		if (namelen > nam->m_len)
    234 			namelen = nam->m_len;
    235 		/* SHOULD COPY OUT A CHAIN HERE */
    236 		if ((error = copyout(mtod(nam, caddr_t),
    237 		    (caddr_t)SCARG(uap, name), namelen)) == 0)
    238 			error = copyout((caddr_t)&namelen,
    239 			    (caddr_t)SCARG(uap, anamelen),
    240 			    sizeof(*SCARG(uap, anamelen)));
    241 	}
    242 	/* if an error occured, free the file descriptor */
    243 	if (error) {
    244 		fdremove(fdp, fd);
    245 		ffree(fp);
    246 	}
    247 	m_freem(nam);
    248 	splx(s);
    249 	return (error);
    250 }
    251 
    252 /* ARGSUSED */
    253 int
    254 sys_connect(struct proc *p, void *v, register_t *retval)
    255 {
    256 	struct sys_connect_args /* {
    257 		syscallarg(int)				s;
    258 		syscallarg(const struct sockaddr *)	name;
    259 		syscallarg(unsigned int)		namelen;
    260 	} */ *uap = v;
    261 	struct file	*fp;
    262 	struct socket	*so;
    263 	struct mbuf	*nam;
    264 	int		error, s;
    265 
    266 	/* getsock() will use the descriptor for us */
    267 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    268 		return (error);
    269 	so = (struct socket *)fp->f_data;
    270 	FILE_UNUSE(fp, p);
    271 	if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING))
    272 		return (EALREADY);
    273 	error = sockargs(&nam, SCARG(uap, name), SCARG(uap, namelen),
    274 	    MT_SONAME);
    275 	if (error)
    276 		return (error);
    277 	error = soconnect(so, nam);
    278 	if (error)
    279 		goto bad;
    280 	if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING)) {
    281 		m_freem(nam);
    282 		return (EINPROGRESS);
    283 	}
    284 	s = splsoftnet();
    285 	while ((so->so_state & SS_ISCONNECTING) && so->so_error == 0) {
    286 		error = tsleep((caddr_t)&so->so_timeo, PSOCK | PCATCH,
    287 			       netcon, 0);
    288 		if (error)
    289 			break;
    290 	}
    291 	if (error == 0) {
    292 		error = so->so_error;
    293 		so->so_error = 0;
    294 	}
    295 	splx(s);
    296  bad:
    297 	so->so_state &= ~SS_ISCONNECTING;
    298 	m_freem(nam);
    299 	if (error == ERESTART)
    300 		error = EINTR;
    301 	return (error);
    302 }
    303 
    304 int
    305 sys_socketpair(struct proc *p, void *v, register_t *retval)
    306 {
    307 	struct sys_socketpair_args /* {
    308 		syscallarg(int)		domain;
    309 		syscallarg(int)		type;
    310 		syscallarg(int)		protocol;
    311 		syscallarg(int *)	rsv;
    312 	} */ *uap = v;
    313 	struct filedesc	*fdp;
    314 	struct file	*fp1, *fp2;
    315 	struct socket	*so1, *so2;
    316 	int		fd, error, sv[2];
    317 
    318 	fdp = p->p_fd;
    319 	error = socreate(SCARG(uap, domain), &so1, SCARG(uap, type),
    320 			 SCARG(uap, protocol));
    321 	if (error)
    322 		return (error);
    323 	error = socreate(SCARG(uap, domain), &so2, SCARG(uap, type),
    324 			 SCARG(uap, protocol));
    325 	if (error)
    326 		goto free1;
    327 	/* falloc() will use the descriptor for us */
    328 	if ((error = falloc(p, &fp1, &fd)) != 0)
    329 		goto free2;
    330 	sv[0] = fd;
    331 	fp1->f_flag = FREAD|FWRITE;
    332 	fp1->f_type = DTYPE_SOCKET;
    333 	fp1->f_ops = &socketops;
    334 	fp1->f_data = (caddr_t)so1;
    335 	if ((error = falloc(p, &fp2, &fd)) != 0)
    336 		goto free3;
    337 	fp2->f_flag = FREAD|FWRITE;
    338 	fp2->f_type = DTYPE_SOCKET;
    339 	fp2->f_ops = &socketops;
    340 	fp2->f_data = (caddr_t)so2;
    341 	sv[1] = fd;
    342 	if ((error = soconnect2(so1, so2)) != 0)
    343 		goto free4;
    344 	if (SCARG(uap, type) == SOCK_DGRAM) {
    345 		/*
    346 		 * Datagram socket connection is asymmetric.
    347 		 */
    348 		 if ((error = soconnect2(so2, so1)) != 0)
    349 			goto free4;
    350 	}
    351 	error = copyout((caddr_t)sv, (caddr_t)SCARG(uap, rsv),
    352 	    2 * sizeof(int));
    353 	FILE_UNUSE(fp1, p);
    354 	FILE_UNUSE(fp2, p);
    355 	return (error);
    356  free4:
    357 	FILE_UNUSE(fp2, p);
    358 	ffree(fp2);
    359 	fdremove(fdp, sv[1]);
    360  free3:
    361 	FILE_UNUSE(fp1, p);
    362 	ffree(fp1);
    363 	fdremove(fdp, sv[0]);
    364  free2:
    365 	(void)soclose(so2);
    366  free1:
    367 	(void)soclose(so1);
    368 	return (error);
    369 }
    370 
    371 int
    372 sys_sendto(struct proc *p, void *v, register_t *retval)
    373 {
    374 	struct sys_sendto_args /* {
    375 		syscallarg(int)				s;
    376 		syscallarg(const void *)		buf;
    377 		syscallarg(size_t)			len;
    378 		syscallarg(int)				flags;
    379 		syscallarg(const struct sockaddr *)	to;
    380 		syscallarg(unsigned int)		tolen;
    381 	} */ *uap = v;
    382 	struct msghdr	msg;
    383 	struct iovec	aiov;
    384 
    385 	msg.msg_name = (caddr_t)SCARG(uap, to);		/* XXX kills const */
    386 	msg.msg_namelen = SCARG(uap, tolen);
    387 	msg.msg_iov = &aiov;
    388 	msg.msg_iovlen = 1;
    389 	msg.msg_control = 0;
    390 #ifdef COMPAT_OLDSOCK
    391 	msg.msg_flags = 0;
    392 #endif
    393 	aiov.iov_base = (char *)SCARG(uap, buf);	/* XXX kills const */
    394 	aiov.iov_len = SCARG(uap, len);
    395 	return (sendit(p, SCARG(uap, s), &msg, SCARG(uap, flags), retval));
    396 }
    397 
    398 int
    399 sys_sendmsg(struct proc *p, void *v, register_t *retval)
    400 {
    401 	struct sys_sendmsg_args /* {
    402 		syscallarg(int)				s;
    403 		syscallarg(const struct msghdr *)	msg;
    404 		syscallarg(int)				flags;
    405 	} */ *uap = v;
    406 	struct msghdr	msg;
    407 	struct iovec	aiov[UIO_SMALLIOV], *iov;
    408 	int		error;
    409 
    410 	error = copyin(SCARG(uap, msg), (caddr_t)&msg, sizeof(msg));
    411 	if (error)
    412 		return (error);
    413 	if ((unsigned int)msg.msg_iovlen > UIO_SMALLIOV) {
    414 		if ((unsigned int)msg.msg_iovlen > IOV_MAX)
    415 			return (EMSGSIZE);
    416 		iov = malloc(sizeof(struct iovec) * msg.msg_iovlen,
    417 		    M_IOV, M_WAITOK);
    418 	} else
    419 		iov = aiov;
    420 	if ((unsigned int)msg.msg_iovlen > 0) {
    421 		error = copyin((caddr_t)msg.msg_iov, (caddr_t)iov,
    422 		    (size_t)(msg.msg_iovlen * sizeof(struct iovec)));
    423 		if (error)
    424 			goto done;
    425 	}
    426 	msg.msg_iov = iov;
    427 #ifdef COMPAT_OLDSOCK
    428 	msg.msg_flags = 0;
    429 #endif
    430 	error = sendit(p, SCARG(uap, s), &msg, SCARG(uap, flags), retval);
    431 done:
    432 	if (iov != aiov)
    433 		free(iov, M_IOV);
    434 	return (error);
    435 }
    436 
    437 int
    438 sendit(struct proc *p, int s, struct msghdr *mp, int flags, register_t *retsize)
    439 {
    440 	struct file	*fp;
    441 	struct uio	auio;
    442 	struct iovec	*iov;
    443 	int		i, len, error;
    444 	struct mbuf	*to, *control;
    445 	struct socket	*so;
    446 #ifdef KTRACE
    447 	struct iovec	*ktriov;
    448 #endif
    449 
    450 #ifdef KTRACE
    451 	ktriov = NULL;
    452 #endif
    453 	/* getsock() will use the descriptor for us */
    454 	if ((error = getsock(p->p_fd, s, &fp)) != 0)
    455 		return (error);
    456 	auio.uio_iov = mp->msg_iov;
    457 	auio.uio_iovcnt = mp->msg_iovlen;
    458 	auio.uio_segflg = UIO_USERSPACE;
    459 	auio.uio_rw = UIO_WRITE;
    460 	auio.uio_procp = p;
    461 	auio.uio_offset = 0;			/* XXX */
    462 	auio.uio_resid = 0;
    463 	iov = mp->msg_iov;
    464 	for (i = 0; i < mp->msg_iovlen; i++, iov++) {
    465 #if 0
    466 		/* cannot happen; iov_len is unsigned */
    467 		if (iov->iov_len < 0) {
    468 			error = EINVAL;
    469 			goto out;
    470 		}
    471 #endif
    472 		/*
    473 		 * Writes return ssize_t because -1 is returned on error.
    474 		 * Therefore, we must restrict the length to SSIZE_MAX to
    475 		 * avoid garbage return values.
    476 		 */
    477 		auio.uio_resid += iov->iov_len;
    478 		if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
    479 			error = EINVAL;
    480 			goto out;
    481 		}
    482 	}
    483 	if (mp->msg_name) {
    484 		error = sockargs(&to, mp->msg_name, mp->msg_namelen,
    485 				 MT_SONAME);
    486 		if (error)
    487 			goto out;
    488 	} else
    489 		to = 0;
    490 	if (mp->msg_control) {
    491 		if (mp->msg_controllen < sizeof(struct cmsghdr)
    492 #ifdef COMPAT_OLDSOCK
    493 		    && mp->msg_flags != MSG_COMPAT
    494 #endif
    495 		) {
    496 			error = EINVAL;
    497 			goto bad;
    498 		}
    499 		error = sockargs(&control, mp->msg_control,
    500 				 mp->msg_controllen, MT_CONTROL);
    501 		if (error)
    502 			goto bad;
    503 #ifdef COMPAT_OLDSOCK
    504 		if (mp->msg_flags == MSG_COMPAT) {
    505 			struct cmsghdr *cm;
    506 
    507 			M_PREPEND(control, sizeof(*cm), M_WAIT);
    508 			if (control == 0) {
    509 				error = ENOBUFS;
    510 				goto bad;
    511 			} else {
    512 				cm = mtod(control, struct cmsghdr *);
    513 				cm->cmsg_len = control->m_len;
    514 				cm->cmsg_level = SOL_SOCKET;
    515 				cm->cmsg_type = SCM_RIGHTS;
    516 			}
    517 		}
    518 #endif
    519 	} else
    520 		control = 0;
    521 #ifdef KTRACE
    522 	if (KTRPOINT(p, KTR_GENIO)) {
    523 		int iovlen = auio.uio_iovcnt * sizeof(struct iovec);
    524 
    525 		ktriov = malloc(iovlen, M_TEMP, M_WAITOK);
    526 		memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen);
    527 	}
    528 #endif
    529 	len = auio.uio_resid;
    530 	so = (struct socket *)fp->f_data;
    531 	error = (*so->so_send)(so, to, &auio, NULL, control, flags);
    532 	if (error) {
    533 		if (auio.uio_resid != len && (error == ERESTART ||
    534 		    error == EINTR || error == EWOULDBLOCK))
    535 			error = 0;
    536 		if (error == EPIPE)
    537 			psignal(p, SIGPIPE);
    538 	}
    539 	if (error == 0)
    540 		*retsize = len - auio.uio_resid;
    541 #ifdef KTRACE
    542 	if (ktriov != NULL) {
    543 		if (error == 0)
    544 			ktrgenio(p, s, UIO_WRITE, ktriov, *retsize, error);
    545 		free(ktriov, M_TEMP);
    546 	}
    547 #endif
    548  bad:
    549 	if (to)
    550 		m_freem(to);
    551  out:
    552 	FILE_UNUSE(fp, p);
    553 	return (error);
    554 }
    555 
    556 int
    557 sys_recvfrom(struct proc *p, void *v, register_t *retval)
    558 {
    559 	struct sys_recvfrom_args /* {
    560 		syscallarg(int)			s;
    561 		syscallarg(void *)		buf;
    562 		syscallarg(size_t)		len;
    563 		syscallarg(int)			flags;
    564 		syscallarg(struct sockaddr *)	from;
    565 		syscallarg(unsigned int *)	fromlenaddr;
    566 	} */ *uap = v;
    567 	struct msghdr	msg;
    568 	struct iovec	aiov;
    569 	int		error;
    570 
    571 	if (SCARG(uap, fromlenaddr)) {
    572 		error = copyin((caddr_t)SCARG(uap, fromlenaddr),
    573 			       (caddr_t)&msg.msg_namelen,
    574 			       sizeof(msg.msg_namelen));
    575 		if (error)
    576 			return (error);
    577 	} else
    578 		msg.msg_namelen = 0;
    579 	msg.msg_name = (caddr_t)SCARG(uap, from);
    580 	msg.msg_iov = &aiov;
    581 	msg.msg_iovlen = 1;
    582 	aiov.iov_base = SCARG(uap, buf);
    583 	aiov.iov_len = SCARG(uap, len);
    584 	msg.msg_control = 0;
    585 	msg.msg_flags = SCARG(uap, flags);
    586 	return (recvit(p, SCARG(uap, s), &msg,
    587 		       (caddr_t)SCARG(uap, fromlenaddr), retval));
    588 }
    589 
    590 int
    591 sys_recvmsg(struct proc *p, void *v, register_t *retval)
    592 {
    593 	struct sys_recvmsg_args /* {
    594 		syscallarg(int)			s;
    595 		syscallarg(struct msghdr *)	msg;
    596 		syscallarg(int)			flags;
    597 	} */ *uap = v;
    598 	struct msghdr	msg;
    599 	struct iovec	aiov[UIO_SMALLIOV], *uiov, *iov;
    600 	int		error;
    601 
    602 	error = copyin((caddr_t)SCARG(uap, msg), (caddr_t)&msg,
    603 		       sizeof(msg));
    604 	if (error)
    605 		return (error);
    606 	if ((unsigned int)msg.msg_iovlen > UIO_SMALLIOV) {
    607 		if ((unsigned int)msg.msg_iovlen > IOV_MAX)
    608 			return (EMSGSIZE);
    609 		iov = malloc(sizeof(struct iovec) * msg.msg_iovlen,
    610 		    M_IOV, M_WAITOK);
    611 	} else
    612 		iov = aiov;
    613 	if ((unsigned int)msg.msg_iovlen > 0) {
    614 		error = copyin((caddr_t)msg.msg_iov, (caddr_t)iov,
    615 		    (size_t)(msg.msg_iovlen * sizeof(struct iovec)));
    616 		if (error)
    617 			goto done;
    618 	}
    619 	uiov = msg.msg_iov;
    620 	msg.msg_iov = iov;
    621 #ifdef COMPAT_OLDSOCK
    622 	msg.msg_flags = SCARG(uap, flags) &~ MSG_COMPAT;
    623 #else
    624 	msg.msg_flags = SCARG(uap, flags);
    625 #endif
    626 	if ((error = recvit(p, SCARG(uap, s), &msg, (caddr_t)0, retval)) == 0) {
    627 		msg.msg_iov = uiov;
    628 		error = copyout((caddr_t)&msg, (caddr_t)SCARG(uap, msg),
    629 		    sizeof(msg));
    630 	}
    631 done:
    632 	if (iov != aiov)
    633 		free(iov, M_IOV);
    634 	return (error);
    635 }
    636 
    637 int
    638 recvit(struct proc *p, int s, struct msghdr *mp, caddr_t namelenp,
    639 	register_t *retsize)
    640 {
    641 	struct file	*fp;
    642 	struct uio	auio;
    643 	struct iovec	*iov;
    644 	int		i, len, error;
    645 	struct mbuf	*from, *control;
    646 	struct socket	*so;
    647 #ifdef KTRACE
    648 	struct iovec	*ktriov;
    649 #endif
    650 
    651 	from = 0;
    652 	control = 0;
    653 #ifdef KTRACE
    654 	ktriov = NULL;
    655 #endif
    656 
    657 	/* getsock() will use the descriptor for us */
    658 	if ((error = getsock(p->p_fd, s, &fp)) != 0)
    659 		return (error);
    660 	auio.uio_iov = mp->msg_iov;
    661 	auio.uio_iovcnt = mp->msg_iovlen;
    662 	auio.uio_segflg = UIO_USERSPACE;
    663 	auio.uio_rw = UIO_READ;
    664 	auio.uio_procp = p;
    665 	auio.uio_offset = 0;			/* XXX */
    666 	auio.uio_resid = 0;
    667 	iov = mp->msg_iov;
    668 	for (i = 0; i < mp->msg_iovlen; i++, iov++) {
    669 #if 0
    670 		/* cannot happen iov_len is unsigned */
    671 		if (iov->iov_len < 0) {
    672 			error = EINVAL;
    673 			goto out1;
    674 		}
    675 #endif
    676 		/*
    677 		 * Reads return ssize_t because -1 is returned on error.
    678 		 * Therefore we must restrict the length to SSIZE_MAX to
    679 		 * avoid garbage return values.
    680 		 */
    681 		auio.uio_resid += iov->iov_len;
    682 		if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
    683 			error = EINVAL;
    684 			goto out1;
    685 		}
    686 	}
    687 #ifdef KTRACE
    688 	if (KTRPOINT(p, KTR_GENIO)) {
    689 		int iovlen = auio.uio_iovcnt * sizeof(struct iovec);
    690 
    691 		ktriov = malloc(iovlen, M_TEMP, M_WAITOK);
    692 		memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen);
    693 	}
    694 #endif
    695 	len = auio.uio_resid;
    696 	so = (struct socket *)fp->f_data;
    697 	error = (*so->so_receive)(so, &from, &auio, NULL,
    698 			  mp->msg_control ? &control : NULL, &mp->msg_flags);
    699 	if (error) {
    700 		if (auio.uio_resid != len && (error == ERESTART ||
    701 		    error == EINTR || error == EWOULDBLOCK))
    702 			error = 0;
    703 	}
    704 #ifdef KTRACE
    705 	if (ktriov != NULL) {
    706 		if (error == 0)
    707 			ktrgenio(p, s, UIO_READ, ktriov,
    708 			    len - auio.uio_resid, error);
    709 		free(ktriov, M_TEMP);
    710 	}
    711 #endif
    712 	if (error)
    713 		goto out;
    714 	*retsize = len - auio.uio_resid;
    715 	if (mp->msg_name) {
    716 		len = mp->msg_namelen;
    717 		if (len <= 0 || from == 0)
    718 			len = 0;
    719 		else {
    720 #ifdef COMPAT_OLDSOCK
    721 			if (mp->msg_flags & MSG_COMPAT)
    722 				mtod(from, struct osockaddr *)->sa_family =
    723 				    mtod(from, struct sockaddr *)->sa_family;
    724 #endif
    725 			if (len > from->m_len)
    726 				len = from->m_len;
    727 			/* else if len < from->m_len ??? */
    728 			error = copyout(mtod(from, caddr_t),
    729 					(caddr_t)mp->msg_name, (unsigned)len);
    730 			if (error)
    731 				goto out;
    732 		}
    733 		mp->msg_namelen = len;
    734 		if (namelenp &&
    735 		    (error = copyout((caddr_t)&len, namelenp, sizeof(int)))) {
    736 #ifdef COMPAT_OLDSOCK
    737 			if (mp->msg_flags & MSG_COMPAT)
    738 				error = 0;	/* old recvfrom didn't check */
    739 			else
    740 #endif
    741 			goto out;
    742 		}
    743 	}
    744 	if (mp->msg_control) {
    745 #ifdef COMPAT_OLDSOCK
    746 		/*
    747 		 * We assume that old recvmsg calls won't receive access
    748 		 * rights and other control info, esp. as control info
    749 		 * is always optional and those options didn't exist in 4.3.
    750 		 * If we receive rights, trim the cmsghdr; anything else
    751 		 * is tossed.
    752 		 */
    753 		if (control && mp->msg_flags & MSG_COMPAT) {
    754 			if (mtod(control, struct cmsghdr *)->cmsg_level !=
    755 			    SOL_SOCKET ||
    756 			    mtod(control, struct cmsghdr *)->cmsg_type !=
    757 			    SCM_RIGHTS) {
    758 				mp->msg_controllen = 0;
    759 				goto out;
    760 			}
    761 			control->m_len -= sizeof(struct cmsghdr);
    762 			control->m_data += sizeof(struct cmsghdr);
    763 		}
    764 #endif
    765 		len = mp->msg_controllen;
    766 		if (len <= 0 || control == 0)
    767 			len = 0;
    768 		else {
    769 			struct mbuf *m = control;
    770 			caddr_t p = (caddr_t)mp->msg_control;
    771 
    772 			do {
    773 				i = m->m_len;
    774 				if (len < i) {
    775 					mp->msg_flags |= MSG_CTRUNC;
    776 					i = len;
    777 				}
    778 				error = copyout(mtod(m, caddr_t), p,
    779 				    (unsigned)i);
    780 				if (m->m_next)
    781 					i = ALIGN(i);
    782 				p += i;
    783 				len -= i;
    784 				if (error != 0 || len <= 0)
    785 					break;
    786 			} while ((m = m->m_next) != NULL);
    787 			len = p - (caddr_t)mp->msg_control;
    788 		}
    789 		mp->msg_controllen = len;
    790 	}
    791  out:
    792 	if (from)
    793 		m_freem(from);
    794 	if (control)
    795 		m_freem(control);
    796  out1:
    797 	FILE_UNUSE(fp, p);
    798 	return (error);
    799 }
    800 
    801 /* ARGSUSED */
    802 int
    803 sys_shutdown(struct proc *p, void *v, register_t *retval)
    804 {
    805 	struct sys_shutdown_args /* {
    806 		syscallarg(int)	s;
    807 		syscallarg(int)	how;
    808 	} */ *uap = v;
    809 	struct file	*fp;
    810 	int		error;
    811 
    812 	/* getsock() will use the descriptor for us */
    813 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    814 		return (error);
    815 	error = soshutdown((struct socket *)fp->f_data, SCARG(uap, how));
    816 	FILE_UNUSE(fp, p);
    817 	return (error);
    818 }
    819 
    820 /* ARGSUSED */
    821 int
    822 sys_setsockopt(struct proc *p, void *v, register_t *retval)
    823 {
    824 	struct sys_setsockopt_args /* {
    825 		syscallarg(int)			s;
    826 		syscallarg(int)			level;
    827 		syscallarg(int)			name;
    828 		syscallarg(const void *)	val;
    829 		syscallarg(unsigned int)	valsize;
    830 	} */ *uap = v;
    831 	struct file	*fp;
    832 	struct mbuf	*m;
    833 	int		error;
    834 
    835 	m = NULL;
    836 	/* getsock() will use the descriptor for us */
    837 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    838 		return (error);
    839 	if (SCARG(uap, valsize) > MLEN) {
    840 		error = EINVAL;
    841 		goto out;
    842 	}
    843 	if (SCARG(uap, val)) {
    844 		m = m_get(M_WAIT, MT_SOOPTS);
    845 		error = copyin(SCARG(uap, val), mtod(m, caddr_t),
    846 			       SCARG(uap, valsize));
    847 		if (error) {
    848 			(void) m_free(m);
    849 			goto out;
    850 		}
    851 		m->m_len = SCARG(uap, valsize);
    852 	}
    853 	error = sosetopt((struct socket *)fp->f_data, SCARG(uap, level),
    854 			 SCARG(uap, name), m);
    855  out:
    856 	FILE_UNUSE(fp, p);
    857 	return (error);
    858 }
    859 
    860 /* ARGSUSED */
    861 int
    862 sys_getsockopt(struct proc *p, void *v, register_t *retval)
    863 {
    864 	struct sys_getsockopt_args /* {
    865 		syscallarg(int)			s;
    866 		syscallarg(int)			level;
    867 		syscallarg(int)			name;
    868 		syscallarg(void *)		val;
    869 		syscallarg(unsigned int *)	avalsize;
    870 	} */ *uap = v;
    871 	struct file	*fp;
    872 	struct mbuf	*m, *m0;
    873 	unsigned int	op, i, valsize;
    874 	int		error;
    875 
    876 	m = NULL;
    877 	/* getsock() will use the descriptor for us */
    878 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    879 		return (error);
    880 	if (SCARG(uap, val)) {
    881 		error = copyin((caddr_t)SCARG(uap, avalsize),
    882 			       (caddr_t)&valsize, sizeof(valsize));
    883 		if (error)
    884 			goto out;
    885 	} else
    886 		valsize = 0;
    887 	if ((error = sogetopt((struct socket *)fp->f_data, SCARG(uap, level),
    888 	    SCARG(uap, name), &m)) == 0 && SCARG(uap, val) && valsize &&
    889 	    m != NULL) {
    890 		op = 0;
    891 		while (m && !error && op < valsize) {
    892 			i = min(m->m_len, (valsize - op));
    893 			error = copyout(mtod(m, caddr_t), SCARG(uap, val), i);
    894 			op += i;
    895 			SCARG(uap, val) = ((u_int8_t *)SCARG(uap, val)) + i;
    896 			m0 = m;
    897 			MFREE(m0, m);
    898 		}
    899 		valsize = op;
    900 		if (error == 0)
    901 			error = copyout(&valsize,
    902 					SCARG(uap, avalsize), sizeof(valsize));
    903 	}
    904 	if (m != NULL)
    905 		(void) m_free(m);
    906  out:
    907 	FILE_UNUSE(fp, p);
    908 	return (error);
    909 }
    910 
    911 /* ARGSUSED */
    912 int
    913 sys_pipe(struct proc *p, void *v, register_t *retval)
    914 {
    915 	struct filedesc	*fdp;
    916 	struct file	*rf, *wf;
    917 	struct socket	*rso, *wso;
    918 	int		fd, error;
    919 
    920 	fdp = p->p_fd;
    921 	if ((error = socreate(AF_LOCAL, &rso, SOCK_STREAM, 0)) != 0)
    922 		return (error);
    923 	if ((error = socreate(AF_LOCAL, &wso, SOCK_STREAM, 0)) != 0)
    924 		goto free1;
    925 	/* falloc() will use the descriptor for us */
    926 	if ((error = falloc(p, &rf, &fd)) != 0)
    927 		goto free2;
    928 	retval[0] = fd;
    929 	rf->f_flag = FREAD;
    930 	rf->f_type = DTYPE_SOCKET;
    931 	rf->f_ops = &socketops;
    932 	rf->f_data = (caddr_t)rso;
    933 	if ((error = falloc(p, &wf, &fd)) != 0)
    934 		goto free3;
    935 	wf->f_flag = FWRITE;
    936 	wf->f_type = DTYPE_SOCKET;
    937 	wf->f_ops = &socketops;
    938 	wf->f_data = (caddr_t)wso;
    939 	retval[1] = fd;
    940 	if ((error = unp_connect2(wso, rso)) != 0)
    941 		goto free4;
    942 	FILE_UNUSE(rf, p);
    943 	FILE_UNUSE(wf, p);
    944 	return (0);
    945  free4:
    946 	FILE_UNUSE(wf, p);
    947 	ffree(wf);
    948 	fdremove(fdp, retval[1]);
    949  free3:
    950 	FILE_UNUSE(rf, p);
    951 	ffree(rf);
    952 	fdremove(fdp, retval[0]);
    953  free2:
    954 	(void)soclose(wso);
    955  free1:
    956 	(void)soclose(rso);
    957 	return (error);
    958 }
    959 
    960 /*
    961  * Get socket name.
    962  */
    963 /* ARGSUSED */
    964 int
    965 sys_getsockname(struct proc *p, void *v, register_t *retval)
    966 {
    967 	struct sys_getsockname_args /* {
    968 		syscallarg(int)			fdes;
    969 		syscallarg(struct sockaddr *)	asa;
    970 		syscallarg(unsigned int *)	alen;
    971 	} */ *uap = v;
    972 	struct file	*fp;
    973 	struct socket	*so;
    974 	struct mbuf	*m;
    975 	unsigned int	len;
    976 	int		error;
    977 
    978 	/* getsock() will use the descriptor for us */
    979 	if ((error = getsock(p->p_fd, SCARG(uap, fdes), &fp)) != 0)
    980 		return (error);
    981 	error = copyin((caddr_t)SCARG(uap, alen), (caddr_t)&len, sizeof(len));
    982 	if (error)
    983 		goto out;
    984 	so = (struct socket *)fp->f_data;
    985 	m = m_getclr(M_WAIT, MT_SONAME);
    986 	error = (*so->so_proto->pr_usrreq)(so, PRU_SOCKADDR, (struct mbuf *)0,
    987 	    m, (struct mbuf *)0, (struct proc *)0);
    988 	if (error)
    989 		goto bad;
    990 	if (len > m->m_len)
    991 		len = m->m_len;
    992 	error = copyout(mtod(m, caddr_t), (caddr_t)SCARG(uap, asa), len);
    993 	if (error == 0)
    994 		error = copyout((caddr_t)&len, (caddr_t)SCARG(uap, alen),
    995 		    sizeof(len));
    996  bad:
    997 	m_freem(m);
    998  out:
    999 	FILE_UNUSE(fp, p);
   1000 	return (error);
   1001 }
   1002 
   1003 /*
   1004  * Get name of peer for connected socket.
   1005  */
   1006 /* ARGSUSED */
   1007 int
   1008 sys_getpeername(struct proc *p, void *v, register_t *retval)
   1009 {
   1010 	struct sys_getpeername_args /* {
   1011 		syscallarg(int)			fdes;
   1012 		syscallarg(struct sockaddr *)	asa;
   1013 		syscallarg(unsigned int *)	alen;
   1014 	} */ *uap = v;
   1015 	struct file	*fp;
   1016 	struct socket	*so;
   1017 	struct mbuf	*m;
   1018 	unsigned int	len;
   1019 	int		error;
   1020 
   1021 	/* getsock() will use the descriptor for us */
   1022 	if ((error = getsock(p->p_fd, SCARG(uap, fdes), &fp)) != 0)
   1023 		return (error);
   1024 	so = (struct socket *)fp->f_data;
   1025 	if ((so->so_state & (SS_ISCONNECTED|SS_ISCONFIRMING)) == 0) {
   1026 		error = ENOTCONN;
   1027 		goto out;
   1028 	}
   1029 	error = copyin((caddr_t)SCARG(uap, alen), (caddr_t)&len, sizeof(len));
   1030 	if (error)
   1031 		goto out;
   1032 	m = m_getclr(M_WAIT, MT_SONAME);
   1033 	error = (*so->so_proto->pr_usrreq)(so, PRU_PEERADDR, (struct mbuf *)0,
   1034 	    m, (struct mbuf *)0, (struct proc *)0);
   1035 	if (error)
   1036 		goto bad;
   1037 	if (len > m->m_len)
   1038 		len = m->m_len;
   1039 	error = copyout(mtod(m, caddr_t), (caddr_t)SCARG(uap, asa), len);
   1040 	if (error)
   1041 		goto bad;
   1042 	error = copyout((caddr_t)&len, (caddr_t)SCARG(uap, alen), sizeof(len));
   1043  bad:
   1044 	m_freem(m);
   1045  out:
   1046 	FILE_UNUSE(fp, p);
   1047 	return (error);
   1048 }
   1049 
   1050 /*
   1051  * XXX In a perfect world, we wouldn't pass around socket control
   1052  * XXX arguments in mbufs, and this could go away.
   1053  */
   1054 int
   1055 sockargs(struct mbuf **mp, const void *buf, int buflen, int type)
   1056 {
   1057 	struct sockaddr	*sa;
   1058 	struct mbuf	*m;
   1059 	int		error;
   1060 
   1061 	/*
   1062 	 * We can't allow socket names > UCHAR_MAX in length, since that
   1063 	 * will overflow sa_len.
   1064 	 */
   1065 	if (type == MT_SONAME && (u_int)buflen > UCHAR_MAX)
   1066 		return (EINVAL);
   1067 
   1068 	/* Allocate an mbuf to hold the arguments. */
   1069 	m = m_get(M_WAIT, type);
   1070 	if ((u_int)buflen > MLEN) {
   1071 		/*
   1072 		 * Won't fit into a regular mbuf, so we allocate just
   1073 		 * enough external storage to hold the argument.
   1074 		 */
   1075 		MEXTMALLOC(m, buflen, M_WAITOK);
   1076 	}
   1077 	m->m_len = buflen;
   1078 	error = copyin(buf, mtod(m, caddr_t), (u_int)buflen);
   1079 	if (error) {
   1080 		(void) m_free(m);
   1081 		return (error);
   1082 	}
   1083 	*mp = m;
   1084 	if (type == MT_SONAME) {
   1085 		sa = mtod(m, struct sockaddr *);
   1086 
   1087 #if defined(COMPAT_OLDSOCK) && BYTE_ORDER != BIG_ENDIAN
   1088 		if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
   1089 			sa->sa_family = sa->sa_len;
   1090 #endif
   1091 		sa->sa_len = buflen;
   1092 	}
   1093 	return (0);
   1094 }
   1095 
   1096 int
   1097 getsock(struct filedesc *fdp, int fdes, struct file **fpp)
   1098 {
   1099 	struct file	*fp;
   1100 
   1101 	if ((unsigned)fdes >= fdp->fd_nfiles ||
   1102 	    (fp = fdp->fd_ofiles[fdes]) == NULL ||
   1103 	    (fp->f_iflags & FIF_WANTCLOSE) != 0)
   1104 		return (EBADF);
   1105 
   1106 	FILE_USE(fp);
   1107 
   1108 	if (fp->f_type != DTYPE_SOCKET) {
   1109 		FILE_UNUSE(fp, NULL);
   1110 		return (ENOTSOCK);
   1111 	}
   1112 	*fpp = fp;
   1113 	return (0);
   1114 }
   1115