Home | History | Annotate | Line # | Download | only in kern
uipc_syscalls.c revision 1.61
      1 /*	$NetBSD: uipc_syscalls.c,v 1.61 2001/06/25 19:24:03 jdolecek 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_new_pipe.h"
     40 
     41 /*
     42  * Though COMPAT_OLDSOCK is needed only for COMPAT_43, SunOS, Linux,
     43  * HP-UX, FreeBSD, Ultrix, OSF1, we define it unconditionally so that
     44  * this would be LKM-safe.
     45  */
     46 #define COMPAT_OLDSOCK		/* used by <sys/socket.h> */
     47 
     48 #include <sys/param.h>
     49 #include <sys/systm.h>
     50 #include <sys/filedesc.h>
     51 #include <sys/proc.h>
     52 #include <sys/file.h>
     53 #include <sys/buf.h>
     54 #include <sys/malloc.h>
     55 #include <sys/mbuf.h>
     56 #include <sys/protosw.h>
     57 #include <sys/socket.h>
     58 #include <sys/socketvar.h>
     59 #include <sys/signalvar.h>
     60 #include <sys/un.h>
     61 #ifdef KTRACE
     62 #include <sys/ktrace.h>
     63 #endif
     64 
     65 #include <sys/mount.h>
     66 #include <sys/syscallargs.h>
     67 
     68 #include <uvm/uvm_extern.h>
     69 
     70 /*
     71  * System call interface to the socket abstraction.
     72  */
     73 extern	struct fileops socketops;
     74 
     75 int
     76 sys_socket(struct proc *p, void *v, register_t *retval)
     77 {
     78 	struct sys_socket_args /* {
     79 		syscallarg(int)	domain;
     80 		syscallarg(int)	type;
     81 		syscallarg(int)	protocol;
     82 	} */ *uap = v;
     83 	struct filedesc	*fdp;
     84 	struct socket	*so;
     85 	struct file	*fp;
     86 	int		fd, error;
     87 
     88 	fdp = p->p_fd;
     89 	/* falloc() will use the desciptor for us */
     90 	if ((error = falloc(p, &fp, &fd)) != 0)
     91 		return (error);
     92 	fp->f_flag = FREAD|FWRITE;
     93 	fp->f_type = DTYPE_SOCKET;
     94 	fp->f_ops = &socketops;
     95 	error = socreate(SCARG(uap, domain), &so, SCARG(uap, type),
     96 			 SCARG(uap, protocol));
     97 	if (error) {
     98 		FILE_UNUSE(fp, p);
     99 		fdremove(fdp, fd);
    100 		ffree(fp);
    101 	} else {
    102 		fp->f_data = (caddr_t)so;
    103 		FILE_SET_MATURE(fp);
    104 		FILE_UNUSE(fp, p);
    105 		*retval = fd;
    106 	}
    107 	return (error);
    108 }
    109 
    110 /* ARGSUSED */
    111 int
    112 sys_bind(struct proc *p, void *v, register_t *retval)
    113 {
    114 	struct sys_bind_args /* {
    115 		syscallarg(int)				s;
    116 		syscallarg(const struct sockaddr *)	name;
    117 		syscallarg(unsigned int)		namelen;
    118 	} */ *uap = v;
    119 	struct file	*fp;
    120 	struct mbuf	*nam;
    121 	int		error;
    122 
    123 	/* getsock() will use the descriptor for us */
    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 		FILE_UNUSE(fp, p);
    130 		return (error);
    131 	}
    132 	error = sobind((struct socket *)fp->f_data, nam, p);
    133 	m_freem(nam);
    134 	FILE_UNUSE(fp, p);
    135 	return (error);
    136 }
    137 
    138 /* ARGSUSED */
    139 int
    140 sys_listen(struct proc *p, void *v, register_t *retval)
    141 {
    142 	struct sys_listen_args /* {
    143 		syscallarg(int)	s;
    144 		syscallarg(int)	backlog;
    145 	} */ *uap = v;
    146 	struct file	*fp;
    147 	int		error;
    148 
    149 	/* getsock() will use the descriptor for us */
    150 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    151 		return (error);
    152 	error = solisten((struct socket *)fp->f_data, SCARG(uap, backlog));
    153 	FILE_UNUSE(fp, p);
    154 	return (error);
    155 }
    156 
    157 int
    158 sys_accept(struct proc *p, void *v, register_t *retval)
    159 {
    160 	struct sys_accept_args /* {
    161 		syscallarg(int)			s;
    162 		syscallarg(struct sockaddr *)	name;
    163 		syscallarg(unsigned int *)	anamelen;
    164 	} */ *uap = v;
    165 	struct filedesc	*fdp;
    166 	struct file	*fp;
    167 	struct mbuf	*nam;
    168 	unsigned int	namelen;
    169 	int		error, s, fd;
    170 	struct socket	*so;
    171 
    172 	fdp = p->p_fd;
    173 	if (SCARG(uap, name) && (error = copyin((caddr_t)SCARG(uap, anamelen),
    174 	    (caddr_t)&namelen, sizeof(namelen))))
    175 		return (error);
    176 	if (SCARG(uap, name) != NULL &&
    177 	    uvm_useracc((caddr_t)SCARG(uap, name), sizeof(struct sockaddr),
    178 	     B_WRITE) == FALSE)
    179 		return (EFAULT);
    180 
    181 	/* getsock() will use the descriptor for us */
    182 	if ((error = getsock(fdp, SCARG(uap, s), &fp)) != 0)
    183 		return (error);
    184 	s = splsoftnet();
    185 	so = (struct socket *)fp->f_data;
    186 	FILE_UNUSE(fp, p);
    187 	if (!(so->so_proto->pr_flags & PR_LISTEN)) {
    188 		splx(s);
    189 		return (EOPNOTSUPP);
    190 	}
    191 	if ((so->so_options & SO_ACCEPTCONN) == 0) {
    192 		splx(s);
    193 		return (EINVAL);
    194 	}
    195 	if ((so->so_state & SS_NBIO) && so->so_qlen == 0) {
    196 		splx(s);
    197 		return (EWOULDBLOCK);
    198 	}
    199 	while (so->so_qlen == 0 && so->so_error == 0) {
    200 		if (so->so_state & SS_CANTRCVMORE) {
    201 			so->so_error = ECONNABORTED;
    202 			break;
    203 		}
    204 		error = tsleep((caddr_t)&so->so_timeo, PSOCK | PCATCH,
    205 			       netcon, 0);
    206 		if (error) {
    207 			splx(s);
    208 			return (error);
    209 		}
    210 	}
    211 	if (so->so_error) {
    212 		error = so->so_error;
    213 		so->so_error = 0;
    214 		splx(s);
    215 		return (error);
    216 	}
    217 	/* falloc() will use the descriptor for us */
    218 	if ((error = falloc(p, &fp, &fd)) != 0) {
    219 		splx(s);
    220 		return (error);
    221 	}
    222 	*retval = fd;
    223 	{ struct socket *aso = so->so_q.tqh_first;
    224 	  if (soqremque(aso, 1) == 0)
    225 		panic("accept");
    226 	  so = aso;
    227 	}
    228 	fp->f_type = DTYPE_SOCKET;
    229 	fp->f_flag = FREAD|FWRITE;
    230 	fp->f_ops = &socketops;
    231 	fp->f_data = (caddr_t)so;
    232 	FILE_UNUSE(fp, p);
    233 	nam = m_get(M_WAIT, MT_SONAME);
    234 	if ((error = soaccept(so, nam)) == 0 && SCARG(uap, name)) {
    235 		if (namelen > nam->m_len)
    236 			namelen = nam->m_len;
    237 		/* SHOULD COPY OUT A CHAIN HERE */
    238 		if ((error = copyout(mtod(nam, caddr_t),
    239 		    (caddr_t)SCARG(uap, name), namelen)) == 0)
    240 			error = copyout((caddr_t)&namelen,
    241 			    (caddr_t)SCARG(uap, anamelen),
    242 			    sizeof(*SCARG(uap, anamelen)));
    243 	}
    244 	/* if an error occured, free the file descriptor */
    245 	if (error) {
    246 		fdremove(fdp, fd);
    247 		ffree(fp);
    248 	}
    249 	m_freem(nam);
    250 	splx(s);
    251 	FILE_SET_MATURE(fp);
    252 	return (error);
    253 }
    254 
    255 /* ARGSUSED */
    256 int
    257 sys_connect(struct proc *p, void *v, register_t *retval)
    258 {
    259 	struct sys_connect_args /* {
    260 		syscallarg(int)				s;
    261 		syscallarg(const struct sockaddr *)	name;
    262 		syscallarg(unsigned int)		namelen;
    263 	} */ *uap = v;
    264 	struct file	*fp;
    265 	struct socket	*so;
    266 	struct mbuf	*nam;
    267 	int		error, s;
    268 
    269 	/* getsock() will use the descriptor for us */
    270 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    271 		return (error);
    272 	so = (struct socket *)fp->f_data;
    273 	FILE_UNUSE(fp, p);
    274 	if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING))
    275 		return (EALREADY);
    276 	error = sockargs(&nam, SCARG(uap, name), SCARG(uap, namelen),
    277 	    MT_SONAME);
    278 	if (error)
    279 		return (error);
    280 	error = soconnect(so, nam);
    281 	if (error)
    282 		goto bad;
    283 	if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING)) {
    284 		m_freem(nam);
    285 		return (EINPROGRESS);
    286 	}
    287 	s = splsoftnet();
    288 	while ((so->so_state & SS_ISCONNECTING) && so->so_error == 0) {
    289 		error = tsleep((caddr_t)&so->so_timeo, PSOCK | PCATCH,
    290 			       netcon, 0);
    291 		if (error)
    292 			break;
    293 	}
    294 	if (error == 0) {
    295 		error = so->so_error;
    296 		so->so_error = 0;
    297 	}
    298 	splx(s);
    299  bad:
    300 	so->so_state &= ~SS_ISCONNECTING;
    301 	m_freem(nam);
    302 	if (error == ERESTART)
    303 		error = EINTR;
    304 	return (error);
    305 }
    306 
    307 int
    308 sys_socketpair(struct proc *p, void *v, register_t *retval)
    309 {
    310 	struct sys_socketpair_args /* {
    311 		syscallarg(int)		domain;
    312 		syscallarg(int)		type;
    313 		syscallarg(int)		protocol;
    314 		syscallarg(int *)	rsv;
    315 	} */ *uap = v;
    316 	struct filedesc	*fdp;
    317 	struct file	*fp1, *fp2;
    318 	struct socket	*so1, *so2;
    319 	int		fd, error, sv[2];
    320 
    321 	fdp = p->p_fd;
    322 	error = socreate(SCARG(uap, domain), &so1, SCARG(uap, type),
    323 			 SCARG(uap, protocol));
    324 	if (error)
    325 		return (error);
    326 	error = socreate(SCARG(uap, domain), &so2, SCARG(uap, type),
    327 			 SCARG(uap, protocol));
    328 	if (error)
    329 		goto free1;
    330 	/* falloc() will use the descriptor for us */
    331 	if ((error = falloc(p, &fp1, &fd)) != 0)
    332 		goto free2;
    333 	sv[0] = fd;
    334 	fp1->f_flag = FREAD|FWRITE;
    335 	fp1->f_type = DTYPE_SOCKET;
    336 	fp1->f_ops = &socketops;
    337 	fp1->f_data = (caddr_t)so1;
    338 	if ((error = falloc(p, &fp2, &fd)) != 0)
    339 		goto free3;
    340 	fp2->f_flag = FREAD|FWRITE;
    341 	fp2->f_type = DTYPE_SOCKET;
    342 	fp2->f_ops = &socketops;
    343 	fp2->f_data = (caddr_t)so2;
    344 	sv[1] = fd;
    345 	if ((error = soconnect2(so1, so2)) != 0)
    346 		goto free4;
    347 	if (SCARG(uap, type) == SOCK_DGRAM) {
    348 		/*
    349 		 * Datagram socket connection is asymmetric.
    350 		 */
    351 		 if ((error = soconnect2(so2, so1)) != 0)
    352 			goto free4;
    353 	}
    354 	error = copyout((caddr_t)sv, (caddr_t)SCARG(uap, rsv),
    355 	    2 * sizeof(int));
    356 	FILE_SET_MATURE(fp1);
    357 	FILE_SET_MATURE(fp2);
    358 	FILE_UNUSE(fp1, p);
    359 	FILE_UNUSE(fp2, p);
    360 	return (error);
    361  free4:
    362 	FILE_UNUSE(fp2, p);
    363 	ffree(fp2);
    364 	fdremove(fdp, sv[1]);
    365  free3:
    366 	FILE_UNUSE(fp1, p);
    367 	ffree(fp1);
    368 	fdremove(fdp, sv[0]);
    369  free2:
    370 	(void)soclose(so2);
    371  free1:
    372 	(void)soclose(so1);
    373 	return (error);
    374 }
    375 
    376 int
    377 sys_sendto(struct proc *p, void *v, register_t *retval)
    378 {
    379 	struct sys_sendto_args /* {
    380 		syscallarg(int)				s;
    381 		syscallarg(const void *)		buf;
    382 		syscallarg(size_t)			len;
    383 		syscallarg(int)				flags;
    384 		syscallarg(const struct sockaddr *)	to;
    385 		syscallarg(unsigned int)		tolen;
    386 	} */ *uap = v;
    387 	struct msghdr	msg;
    388 	struct iovec	aiov;
    389 
    390 	msg.msg_name = (caddr_t)SCARG(uap, to);		/* XXX kills const */
    391 	msg.msg_namelen = SCARG(uap, tolen);
    392 	msg.msg_iov = &aiov;
    393 	msg.msg_iovlen = 1;
    394 	msg.msg_control = 0;
    395 #ifdef COMPAT_OLDSOCK
    396 	msg.msg_flags = 0;
    397 #endif
    398 	aiov.iov_base = (char *)SCARG(uap, buf);	/* XXX kills const */
    399 	aiov.iov_len = SCARG(uap, len);
    400 	return (sendit(p, SCARG(uap, s), &msg, SCARG(uap, flags), retval, 0));
    401 }
    402 
    403 int
    404 sys_sendmsg(struct proc *p, void *v, register_t *retval)
    405 {
    406 	struct sys_sendmsg_args /* {
    407 		syscallarg(int)				s;
    408 		syscallarg(const struct msghdr *)	msg;
    409 		syscallarg(int)				flags;
    410 	} */ *uap = v;
    411 	struct msghdr	msg;
    412 	struct iovec	aiov[UIO_SMALLIOV], *iov;
    413 	int		error;
    414 
    415 	error = copyin(SCARG(uap, msg), (caddr_t)&msg, sizeof(msg));
    416 	if (error)
    417 		return (error);
    418 	if ((unsigned int)msg.msg_iovlen > UIO_SMALLIOV) {
    419 		if ((unsigned int)msg.msg_iovlen > IOV_MAX)
    420 			return (EMSGSIZE);
    421 		iov = malloc(sizeof(struct iovec) * msg.msg_iovlen,
    422 		    M_IOV, M_WAITOK);
    423 	} else
    424 		iov = aiov;
    425 	if ((unsigned int)msg.msg_iovlen > 0) {
    426 		error = copyin((caddr_t)msg.msg_iov, (caddr_t)iov,
    427 		    (size_t)(msg.msg_iovlen * sizeof(struct iovec)));
    428 		if (error)
    429 			goto done;
    430 	}
    431 	msg.msg_iov = iov;
    432 #ifdef COMPAT_OLDSOCK
    433 	msg.msg_flags = 0;
    434 #endif
    435 	error = sendit(p, SCARG(uap, s), &msg, SCARG(uap, flags), retval, 0);
    436 done:
    437 	if (iov != aiov)
    438 		free(iov, M_IOV);
    439 	return (error);
    440 }
    441 
    442 int
    443 sendit(struct proc *p, int s, struct msghdr *mp, int flags, register_t *retsize,
    444 	int kernsa)
    445 {
    446 	struct file	*fp;
    447 	struct uio	auio;
    448 	struct iovec	*iov;
    449 	int		i, len, error;
    450 	struct mbuf	*to, *control;
    451 	struct socket	*so;
    452 #ifdef KTRACE
    453 	struct iovec	*ktriov;
    454 #endif
    455 
    456 #ifdef KTRACE
    457 	ktriov = NULL;
    458 #endif
    459 	/* getsock() will use the descriptor for us */
    460 	if ((error = getsock(p->p_fd, s, &fp)) != 0)
    461 		return (error);
    462 	auio.uio_iov = mp->msg_iov;
    463 	auio.uio_iovcnt = mp->msg_iovlen;
    464 	auio.uio_segflg = UIO_USERSPACE;
    465 	auio.uio_rw = UIO_WRITE;
    466 	auio.uio_procp = p;
    467 	auio.uio_offset = 0;			/* XXX */
    468 	auio.uio_resid = 0;
    469 	iov = mp->msg_iov;
    470 	for (i = 0; i < mp->msg_iovlen; i++, iov++) {
    471 #if 0
    472 		/* cannot happen; iov_len is unsigned */
    473 		if (iov->iov_len < 0) {
    474 			error = EINVAL;
    475 			goto out;
    476 		}
    477 #endif
    478 		/*
    479 		 * Writes return ssize_t because -1 is returned on error.
    480 		 * Therefore, we must restrict the length to SSIZE_MAX to
    481 		 * avoid garbage return values.
    482 		 */
    483 		auio.uio_resid += iov->iov_len;
    484 		if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
    485 			error = EINVAL;
    486 			goto out;
    487 		}
    488 	}
    489 	if (mp->msg_name) {
    490 		if (kernsa)
    491 			to = (struct mbuf *) mp->msg_name;
    492 		else {
    493 			error = sockargs(&to, mp->msg_name, mp->msg_namelen,
    494 				 MT_SONAME);
    495 			if (error)
    496 				goto out;
    497 		}
    498 	} else
    499 		to = 0;
    500 	if (mp->msg_control) {
    501 		if (mp->msg_controllen < sizeof(struct cmsghdr)
    502 #ifdef COMPAT_OLDSOCK
    503 		    && mp->msg_flags != MSG_COMPAT
    504 #endif
    505 		) {
    506 			error = EINVAL;
    507 			goto bad;
    508 		}
    509 		error = sockargs(&control, mp->msg_control,
    510 				 mp->msg_controllen, MT_CONTROL);
    511 		if (error)
    512 			goto bad;
    513 #ifdef COMPAT_OLDSOCK
    514 		if (mp->msg_flags == MSG_COMPAT) {
    515 			struct cmsghdr *cm;
    516 
    517 			M_PREPEND(control, sizeof(*cm), M_WAIT);
    518 			if (control == 0) {
    519 				error = ENOBUFS;
    520 				goto bad;
    521 			} else {
    522 				cm = mtod(control, struct cmsghdr *);
    523 				cm->cmsg_len = control->m_len;
    524 				cm->cmsg_level = SOL_SOCKET;
    525 				cm->cmsg_type = SCM_RIGHTS;
    526 			}
    527 		}
    528 #endif
    529 	} else
    530 		control = 0;
    531 #ifdef KTRACE
    532 	if (KTRPOINT(p, KTR_GENIO)) {
    533 		int iovlen = auio.uio_iovcnt * sizeof(struct iovec);
    534 
    535 		ktriov = malloc(iovlen, M_TEMP, M_WAITOK);
    536 		memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen);
    537 	}
    538 #endif
    539 	len = auio.uio_resid;
    540 	so = (struct socket *)fp->f_data;
    541 	error = (*so->so_send)(so, to, &auio, NULL, control, flags);
    542 	if (error) {
    543 		if (auio.uio_resid != len && (error == ERESTART ||
    544 		    error == EINTR || error == EWOULDBLOCK))
    545 			error = 0;
    546 		if (error == EPIPE)
    547 			psignal(p, SIGPIPE);
    548 	}
    549 	if (error == 0)
    550 		*retsize = len - auio.uio_resid;
    551 #ifdef KTRACE
    552 	if (ktriov != NULL) {
    553 		if (error == 0)
    554 			ktrgenio(p, s, UIO_WRITE, ktriov, *retsize, error);
    555 		free(ktriov, M_TEMP);
    556 	}
    557 #endif
    558  bad:
    559 	if (to)
    560 		m_freem(to);
    561  out:
    562 	FILE_UNUSE(fp, p);
    563 	return (error);
    564 }
    565 
    566 int
    567 sys_recvfrom(struct proc *p, void *v, register_t *retval)
    568 {
    569 	struct sys_recvfrom_args /* {
    570 		syscallarg(int)			s;
    571 		syscallarg(void *)		buf;
    572 		syscallarg(size_t)		len;
    573 		syscallarg(int)			flags;
    574 		syscallarg(struct sockaddr *)	from;
    575 		syscallarg(unsigned int *)	fromlenaddr;
    576 	} */ *uap = v;
    577 	struct msghdr	msg;
    578 	struct iovec	aiov;
    579 	int		error;
    580 
    581 	if (SCARG(uap, fromlenaddr)) {
    582 		error = copyin((caddr_t)SCARG(uap, fromlenaddr),
    583 			       (caddr_t)&msg.msg_namelen,
    584 			       sizeof(msg.msg_namelen));
    585 		if (error)
    586 			return (error);
    587 	} else
    588 		msg.msg_namelen = 0;
    589 	msg.msg_name = (caddr_t)SCARG(uap, from);
    590 	msg.msg_iov = &aiov;
    591 	msg.msg_iovlen = 1;
    592 	aiov.iov_base = SCARG(uap, buf);
    593 	aiov.iov_len = SCARG(uap, len);
    594 	msg.msg_control = 0;
    595 	msg.msg_flags = SCARG(uap, flags);
    596 	return (recvit(p, SCARG(uap, s), &msg,
    597 		       (caddr_t)SCARG(uap, fromlenaddr), retval, 0));
    598 }
    599 
    600 int
    601 sys_recvmsg(struct proc *p, void *v, register_t *retval)
    602 {
    603 	struct sys_recvmsg_args /* {
    604 		syscallarg(int)			s;
    605 		syscallarg(struct msghdr *)	msg;
    606 		syscallarg(int)			flags;
    607 	} */ *uap = v;
    608 	struct msghdr	msg;
    609 	struct iovec	aiov[UIO_SMALLIOV], *uiov, *iov;
    610 	int		error;
    611 
    612 	error = copyin((caddr_t)SCARG(uap, msg), (caddr_t)&msg,
    613 		       sizeof(msg));
    614 	if (error)
    615 		return (error);
    616 	if ((unsigned int)msg.msg_iovlen > UIO_SMALLIOV) {
    617 		if ((unsigned int)msg.msg_iovlen > IOV_MAX)
    618 			return (EMSGSIZE);
    619 		iov = malloc(sizeof(struct iovec) * msg.msg_iovlen,
    620 		    M_IOV, M_WAITOK);
    621 	} else
    622 		iov = aiov;
    623 	if ((unsigned int)msg.msg_iovlen > 0) {
    624 		error = copyin((caddr_t)msg.msg_iov, (caddr_t)iov,
    625 		    (size_t)(msg.msg_iovlen * sizeof(struct iovec)));
    626 		if (error)
    627 			goto done;
    628 	}
    629 	uiov = msg.msg_iov;
    630 	msg.msg_iov = iov;
    631 #ifdef COMPAT_OLDSOCK
    632 	msg.msg_flags = SCARG(uap, flags) &~ MSG_COMPAT;
    633 #else
    634 	msg.msg_flags = SCARG(uap, flags);
    635 #endif
    636 	if ((error = recvit(p, SCARG(uap, s), &msg, (caddr_t)0, retval,0)) ==0){
    637 		msg.msg_iov = uiov;
    638 		error = copyout((caddr_t)&msg, (caddr_t)SCARG(uap, msg),
    639 		    sizeof(msg));
    640 	}
    641 done:
    642 	if (iov != aiov)
    643 		free(iov, M_IOV);
    644 	return (error);
    645 }
    646 
    647 int
    648 recvit(struct proc *p, int s, struct msghdr *mp, caddr_t namelenp,
    649 	register_t *retsize, int kernsa)
    650 {
    651 	struct file	*fp;
    652 	struct uio	auio;
    653 	struct iovec	*iov;
    654 	int		i, len, error;
    655 	struct mbuf	*from, *control;
    656 	struct socket	*so;
    657 #ifdef KTRACE
    658 	struct iovec	*ktriov;
    659 #endif
    660 
    661 	from = 0;
    662 	control = 0;
    663 #ifdef KTRACE
    664 	ktriov = NULL;
    665 #endif
    666 
    667 	/* getsock() will use the descriptor for us */
    668 	if ((error = getsock(p->p_fd, s, &fp)) != 0)
    669 		return (error);
    670 	auio.uio_iov = mp->msg_iov;
    671 	auio.uio_iovcnt = mp->msg_iovlen;
    672 	auio.uio_segflg = UIO_USERSPACE;
    673 	auio.uio_rw = UIO_READ;
    674 	auio.uio_procp = p;
    675 	auio.uio_offset = 0;			/* XXX */
    676 	auio.uio_resid = 0;
    677 	iov = mp->msg_iov;
    678 	for (i = 0; i < mp->msg_iovlen; i++, iov++) {
    679 #if 0
    680 		/* cannot happen iov_len is unsigned */
    681 		if (iov->iov_len < 0) {
    682 			error = EINVAL;
    683 			goto out1;
    684 		}
    685 #endif
    686 		/*
    687 		 * Reads return ssize_t because -1 is returned on error.
    688 		 * Therefore we must restrict the length to SSIZE_MAX to
    689 		 * avoid garbage return values.
    690 		 */
    691 		auio.uio_resid += iov->iov_len;
    692 		if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
    693 			error = EINVAL;
    694 			goto out1;
    695 		}
    696 	}
    697 #ifdef KTRACE
    698 	if (KTRPOINT(p, KTR_GENIO)) {
    699 		int iovlen = auio.uio_iovcnt * sizeof(struct iovec);
    700 
    701 		ktriov = malloc(iovlen, M_TEMP, M_WAITOK);
    702 		memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen);
    703 	}
    704 #endif
    705 	len = auio.uio_resid;
    706 	so = (struct socket *)fp->f_data;
    707 	error = (*so->so_receive)(so, &from, &auio, NULL,
    708 			  mp->msg_control ? &control : NULL, &mp->msg_flags);
    709 	if (error) {
    710 		if (auio.uio_resid != len && (error == ERESTART ||
    711 		    error == EINTR || error == EWOULDBLOCK))
    712 			error = 0;
    713 	}
    714 #ifdef KTRACE
    715 	if (ktriov != NULL) {
    716 		if (error == 0)
    717 			ktrgenio(p, s, UIO_READ, ktriov,
    718 			    len - auio.uio_resid, error);
    719 		free(ktriov, M_TEMP);
    720 	}
    721 #endif
    722 	if (error)
    723 		goto out;
    724 	*retsize = len - auio.uio_resid;
    725 	if (mp->msg_name) {
    726 		len = mp->msg_namelen;
    727 		if (len <= 0 || from == 0)
    728 			len = 0;
    729 		else {
    730 #ifdef COMPAT_OLDSOCK
    731 			if (mp->msg_flags & MSG_COMPAT)
    732 				mtod(from, struct osockaddr *)->sa_family =
    733 				    mtod(from, struct sockaddr *)->sa_family;
    734 #endif
    735 			if (len > from->m_len)
    736 				len = from->m_len;
    737 			/* else if len < from->m_len ??? */
    738 			if (kernsa)
    739 				memcpy(mp->msg_name, mtod(from, caddr_t), len);
    740 			else {
    741 				error = copyout(mtod(from, caddr_t),
    742 					(caddr_t)mp->msg_name, (unsigned)len);
    743 				if (error)
    744 					goto out;
    745 			}
    746 		}
    747 		mp->msg_namelen = len;
    748 		if (namelenp &&
    749 		    (error = copyout((caddr_t)&len, namelenp, sizeof(int)))) {
    750 #ifdef COMPAT_OLDSOCK
    751 			if (mp->msg_flags & MSG_COMPAT)
    752 				error = 0;	/* old recvfrom didn't check */
    753 			else
    754 #endif
    755 			goto out;
    756 		}
    757 	}
    758 	if (mp->msg_control) {
    759 #ifdef COMPAT_OLDSOCK
    760 		/*
    761 		 * We assume that old recvmsg calls won't receive access
    762 		 * rights and other control info, esp. as control info
    763 		 * is always optional and those options didn't exist in 4.3.
    764 		 * If we receive rights, trim the cmsghdr; anything else
    765 		 * is tossed.
    766 		 */
    767 		if (control && mp->msg_flags & MSG_COMPAT) {
    768 			if (mtod(control, struct cmsghdr *)->cmsg_level !=
    769 			    SOL_SOCKET ||
    770 			    mtod(control, struct cmsghdr *)->cmsg_type !=
    771 			    SCM_RIGHTS) {
    772 				mp->msg_controllen = 0;
    773 				goto out;
    774 			}
    775 			control->m_len -= sizeof(struct cmsghdr);
    776 			control->m_data += sizeof(struct cmsghdr);
    777 		}
    778 #endif
    779 		len = mp->msg_controllen;
    780 		if (len <= 0 || control == 0)
    781 			len = 0;
    782 		else {
    783 			struct mbuf *m = control;
    784 			caddr_t p = (caddr_t)mp->msg_control;
    785 
    786 			do {
    787 				i = m->m_len;
    788 				if (len < i) {
    789 					mp->msg_flags |= MSG_CTRUNC;
    790 					i = len;
    791 				}
    792 				error = copyout(mtod(m, caddr_t), p,
    793 				    (unsigned)i);
    794 				if (m->m_next)
    795 					i = ALIGN(i);
    796 				p += i;
    797 				len -= i;
    798 				if (error != 0 || len <= 0)
    799 					break;
    800 			} while ((m = m->m_next) != NULL);
    801 			len = p - (caddr_t)mp->msg_control;
    802 		}
    803 		mp->msg_controllen = len;
    804 	}
    805  out:
    806 	if (from)
    807 		m_freem(from);
    808 	if (control)
    809 		m_freem(control);
    810  out1:
    811 	FILE_UNUSE(fp, p);
    812 	return (error);
    813 }
    814 
    815 /* ARGSUSED */
    816 int
    817 sys_shutdown(struct proc *p, void *v, register_t *retval)
    818 {
    819 	struct sys_shutdown_args /* {
    820 		syscallarg(int)	s;
    821 		syscallarg(int)	how;
    822 	} */ *uap = v;
    823 	struct file	*fp;
    824 	int		error;
    825 
    826 	/* getsock() will use the descriptor for us */
    827 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    828 		return (error);
    829 	error = soshutdown((struct socket *)fp->f_data, SCARG(uap, how));
    830 	FILE_UNUSE(fp, p);
    831 	return (error);
    832 }
    833 
    834 /* ARGSUSED */
    835 int
    836 sys_setsockopt(struct proc *p, void *v, register_t *retval)
    837 {
    838 	struct sys_setsockopt_args /* {
    839 		syscallarg(int)			s;
    840 		syscallarg(int)			level;
    841 		syscallarg(int)			name;
    842 		syscallarg(const void *)	val;
    843 		syscallarg(unsigned int)	valsize;
    844 	} */ *uap = v;
    845 	struct file	*fp;
    846 	struct mbuf	*m;
    847 	int		error;
    848 
    849 	m = NULL;
    850 	/* getsock() will use the descriptor for us */
    851 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    852 		return (error);
    853 	if (SCARG(uap, valsize) > MLEN) {
    854 		error = EINVAL;
    855 		goto out;
    856 	}
    857 	if (SCARG(uap, val)) {
    858 		m = m_get(M_WAIT, MT_SOOPTS);
    859 		error = copyin(SCARG(uap, val), mtod(m, caddr_t),
    860 			       SCARG(uap, valsize));
    861 		if (error) {
    862 			(void) m_free(m);
    863 			goto out;
    864 		}
    865 		m->m_len = SCARG(uap, valsize);
    866 	}
    867 	error = sosetopt((struct socket *)fp->f_data, SCARG(uap, level),
    868 			 SCARG(uap, name), m);
    869  out:
    870 	FILE_UNUSE(fp, p);
    871 	return (error);
    872 }
    873 
    874 /* ARGSUSED */
    875 int
    876 sys_getsockopt(struct proc *p, void *v, register_t *retval)
    877 {
    878 	struct sys_getsockopt_args /* {
    879 		syscallarg(int)			s;
    880 		syscallarg(int)			level;
    881 		syscallarg(int)			name;
    882 		syscallarg(void *)		val;
    883 		syscallarg(unsigned int *)	avalsize;
    884 	} */ *uap = v;
    885 	struct file	*fp;
    886 	struct mbuf	*m, *m0;
    887 	unsigned int	op, i, valsize;
    888 	int		error;
    889 
    890 	m = NULL;
    891 	/* getsock() will use the descriptor for us */
    892 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    893 		return (error);
    894 	if (SCARG(uap, val)) {
    895 		error = copyin((caddr_t)SCARG(uap, avalsize),
    896 			       (caddr_t)&valsize, sizeof(valsize));
    897 		if (error)
    898 			goto out;
    899 	} else
    900 		valsize = 0;
    901 	if ((error = sogetopt((struct socket *)fp->f_data, SCARG(uap, level),
    902 	    SCARG(uap, name), &m)) == 0 && SCARG(uap, val) && valsize &&
    903 	    m != NULL) {
    904 		op = 0;
    905 		while (m && !error && op < valsize) {
    906 			i = min(m->m_len, (valsize - op));
    907 			error = copyout(mtod(m, caddr_t), SCARG(uap, val), i);
    908 			op += i;
    909 			SCARG(uap, val) = ((u_int8_t *)SCARG(uap, val)) + i;
    910 			m0 = m;
    911 			MFREE(m0, m);
    912 		}
    913 		valsize = op;
    914 		if (error == 0)
    915 			error = copyout(&valsize,
    916 					SCARG(uap, avalsize), sizeof(valsize));
    917 	}
    918 	if (m != NULL)
    919 		(void) m_free(m);
    920  out:
    921 	FILE_UNUSE(fp, p);
    922 	return (error);
    923 }
    924 
    925 #ifndef NEW_PIPE
    926 /* ARGSUSED */
    927 int
    928 sys_pipe(struct proc *p, void *v, register_t *retval)
    929 {
    930 	struct filedesc	*fdp;
    931 	struct file	*rf, *wf;
    932 	struct socket	*rso, *wso;
    933 	int		fd, error;
    934 
    935 	fdp = p->p_fd;
    936 	if ((error = socreate(AF_LOCAL, &rso, SOCK_STREAM, 0)) != 0)
    937 		return (error);
    938 	if ((error = socreate(AF_LOCAL, &wso, SOCK_STREAM, 0)) != 0)
    939 		goto free1;
    940 	/* remember this socket pair implements a pipe */
    941 	wso->so_state |= SS_ISAPIPE;
    942 	rso->so_state |= SS_ISAPIPE;
    943 	/* falloc() will use the descriptor for us */
    944 	if ((error = falloc(p, &rf, &fd)) != 0)
    945 		goto free2;
    946 	retval[0] = fd;
    947 	rf->f_flag = FREAD;
    948 	rf->f_type = DTYPE_SOCKET;
    949 	rf->f_ops = &socketops;
    950 	rf->f_data = (caddr_t)rso;
    951 	if ((error = falloc(p, &wf, &fd)) != 0)
    952 		goto free3;
    953 	wf->f_flag = FWRITE;
    954 	wf->f_type = DTYPE_SOCKET;
    955 	wf->f_ops = &socketops;
    956 	wf->f_data = (caddr_t)wso;
    957 	retval[1] = fd;
    958 	if ((error = unp_connect2(wso, rso)) != 0)
    959 		goto free4;
    960 	FILE_SET_MATURE(rf);
    961 	FILE_SET_MATURE(wf);
    962 	FILE_UNUSE(rf, p);
    963 	FILE_UNUSE(wf, p);
    964 	return (0);
    965  free4:
    966 	FILE_UNUSE(wf, p);
    967 	ffree(wf);
    968 	fdremove(fdp, retval[1]);
    969  free3:
    970 	FILE_UNUSE(rf, p);
    971 	ffree(rf);
    972 	fdremove(fdp, retval[0]);
    973  free2:
    974 	(void)soclose(wso);
    975  free1:
    976 	(void)soclose(rso);
    977 	return (error);
    978 }
    979 #endif /* !NEW_PIPE */
    980 
    981 /*
    982  * Get socket name.
    983  */
    984 /* ARGSUSED */
    985 int
    986 sys_getsockname(struct proc *p, void *v, register_t *retval)
    987 {
    988 	struct sys_getsockname_args /* {
    989 		syscallarg(int)			fdes;
    990 		syscallarg(struct sockaddr *)	asa;
    991 		syscallarg(unsigned int *)	alen;
    992 	} */ *uap = v;
    993 	struct file	*fp;
    994 	struct socket	*so;
    995 	struct mbuf	*m;
    996 	unsigned int	len;
    997 	int		error;
    998 
    999 	/* getsock() will use the descriptor for us */
   1000 	if ((error = getsock(p->p_fd, SCARG(uap, fdes), &fp)) != 0)
   1001 		return (error);
   1002 	error = copyin((caddr_t)SCARG(uap, alen), (caddr_t)&len, sizeof(len));
   1003 	if (error)
   1004 		goto out;
   1005 	so = (struct socket *)fp->f_data;
   1006 	m = m_getclr(M_WAIT, MT_SONAME);
   1007 	error = (*so->so_proto->pr_usrreq)(so, PRU_SOCKADDR, (struct mbuf *)0,
   1008 	    m, (struct mbuf *)0, (struct proc *)0);
   1009 	if (error)
   1010 		goto bad;
   1011 	if (len > m->m_len)
   1012 		len = m->m_len;
   1013 	error = copyout(mtod(m, caddr_t), (caddr_t)SCARG(uap, asa), len);
   1014 	if (error == 0)
   1015 		error = copyout((caddr_t)&len, (caddr_t)SCARG(uap, alen),
   1016 		    sizeof(len));
   1017  bad:
   1018 	m_freem(m);
   1019  out:
   1020 	FILE_UNUSE(fp, p);
   1021 	return (error);
   1022 }
   1023 
   1024 /*
   1025  * Get name of peer for connected socket.
   1026  */
   1027 /* ARGSUSED */
   1028 int
   1029 sys_getpeername(struct proc *p, void *v, register_t *retval)
   1030 {
   1031 	struct sys_getpeername_args /* {
   1032 		syscallarg(int)			fdes;
   1033 		syscallarg(struct sockaddr *)	asa;
   1034 		syscallarg(unsigned int *)	alen;
   1035 	} */ *uap = v;
   1036 	struct file	*fp;
   1037 	struct socket	*so;
   1038 	struct mbuf	*m;
   1039 	unsigned int	len;
   1040 	int		error;
   1041 
   1042 	/* getsock() will use the descriptor for us */
   1043 	if ((error = getsock(p->p_fd, SCARG(uap, fdes), &fp)) != 0)
   1044 		return (error);
   1045 	so = (struct socket *)fp->f_data;
   1046 	if ((so->so_state & (SS_ISCONNECTED|SS_ISCONFIRMING)) == 0) {
   1047 		error = ENOTCONN;
   1048 		goto out;
   1049 	}
   1050 	error = copyin((caddr_t)SCARG(uap, alen), (caddr_t)&len, sizeof(len));
   1051 	if (error)
   1052 		goto out;
   1053 	m = m_getclr(M_WAIT, MT_SONAME);
   1054 	error = (*so->so_proto->pr_usrreq)(so, PRU_PEERADDR, (struct mbuf *)0,
   1055 	    m, (struct mbuf *)0, (struct proc *)0);
   1056 	if (error)
   1057 		goto bad;
   1058 	if (len > m->m_len)
   1059 		len = m->m_len;
   1060 	error = copyout(mtod(m, caddr_t), (caddr_t)SCARG(uap, asa), len);
   1061 	if (error)
   1062 		goto bad;
   1063 	error = copyout((caddr_t)&len, (caddr_t)SCARG(uap, alen), sizeof(len));
   1064  bad:
   1065 	m_freem(m);
   1066  out:
   1067 	FILE_UNUSE(fp, p);
   1068 	return (error);
   1069 }
   1070 
   1071 /*
   1072  * XXX In a perfect world, we wouldn't pass around socket control
   1073  * XXX arguments in mbufs, and this could go away.
   1074  */
   1075 int
   1076 sockargs(struct mbuf **mp, const void *buf, int buflen, int type)
   1077 {
   1078 	struct sockaddr	*sa;
   1079 	struct mbuf	*m;
   1080 	int		error;
   1081 
   1082 	/*
   1083 	 * We can't allow socket names > UCHAR_MAX in length, since that
   1084 	 * will overflow sa_len.
   1085 	 */
   1086 	if (type == MT_SONAME && (u_int)buflen > UCHAR_MAX)
   1087 		return (EINVAL);
   1088 
   1089 	/* Allocate an mbuf to hold the arguments. */
   1090 	m = m_get(M_WAIT, type);
   1091 	if ((u_int)buflen > MLEN) {
   1092 		/*
   1093 		 * Won't fit into a regular mbuf, so we allocate just
   1094 		 * enough external storage to hold the argument.
   1095 		 */
   1096 		MEXTMALLOC(m, buflen, M_WAITOK);
   1097 	}
   1098 	m->m_len = buflen;
   1099 	error = copyin(buf, mtod(m, caddr_t), (u_int)buflen);
   1100 	if (error) {
   1101 		(void) m_free(m);
   1102 		return (error);
   1103 	}
   1104 	*mp = m;
   1105 	if (type == MT_SONAME) {
   1106 		sa = mtod(m, struct sockaddr *);
   1107 
   1108 #if defined(COMPAT_OLDSOCK) && BYTE_ORDER != BIG_ENDIAN
   1109 		if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
   1110 			sa->sa_family = sa->sa_len;
   1111 #endif
   1112 		sa->sa_len = buflen;
   1113 	}
   1114 	return (0);
   1115 }
   1116 
   1117 int
   1118 getsock(struct filedesc *fdp, int fdes, struct file **fpp)
   1119 {
   1120 	struct file	*fp;
   1121 
   1122 	if ((fp = fd_getfile(fdp, fdes)) == NULL)
   1123 		return (EBADF);
   1124 
   1125 	FILE_USE(fp);
   1126 
   1127 	if (fp->f_type != DTYPE_SOCKET) {
   1128 		FILE_UNUSE(fp, NULL);
   1129 		return (ENOTSOCK);
   1130 	}
   1131 	*fpp = fp;
   1132 	return (0);
   1133 }
   1134