Home | History | Annotate | Line # | Download | only in kern
uipc_syscalls.c revision 1.41.2.3
      1 /*	$NetBSD: uipc_syscalls.c,v 1.41.2.3 2001/07/29 20:53:41 he 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 <vm/vm.h>
     73 #include <uvm/uvm_extern.h>
     74 #include <vm/vm_param.h>
     75 
     76 /*
     77  * System call interface to the socket abstraction.
     78  */
     79 extern	struct fileops socketops;
     80 
     81 int
     82 sys_socket(p, v, retval)
     83 	struct proc *p;
     84 	void *v;
     85 	register_t *retval;
     86 {
     87 	register struct sys_socket_args /* {
     88 		syscallarg(int) domain;
     89 		syscallarg(int) type;
     90 		syscallarg(int) protocol;
     91 	} */ *uap = v;
     92 	struct filedesc *fdp = p->p_fd;
     93 	struct socket *so;
     94 	struct file *fp;
     95 	int fd, error;
     96 
     97 	if ((error = falloc(p, &fp, &fd)) != 0)
     98 		return (error);
     99 	fp->f_flag = FREAD|FWRITE;
    100 	fp->f_type = DTYPE_SOCKET;
    101 	fp->f_ops = &socketops;
    102 	error = socreate(SCARG(uap, domain), &so, SCARG(uap, type),
    103 			 SCARG(uap, protocol));
    104 	if (error) {
    105 		fdp->fd_ofiles[fd] = 0;
    106 		ffree(fp);
    107 	} else {
    108 		fp->f_data = (caddr_t)so;
    109 		*retval = fd;
    110 	}
    111 	return (error);
    112 }
    113 
    114 /* ARGSUSED */
    115 int
    116 sys_bind(p, v, retval)
    117 	struct proc *p;
    118 	void *v;
    119 	register_t *retval;
    120 {
    121 	register struct sys_bind_args /* {
    122 		syscallarg(int) s;
    123 		syscallarg(const struct sockaddr *) name;
    124 		syscallarg(unsigned int) namelen;
    125 	} */ *uap = v;
    126 	struct file *fp;
    127 	struct mbuf *nam;
    128 	int error;
    129 
    130 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    131 		return (error);
    132 	error = sockargs(&nam, SCARG(uap, name), SCARG(uap, namelen),
    133 	    MT_SONAME);
    134 	if (error)
    135 		return (error);
    136 	error = sobind((struct socket *)fp->f_data, nam);
    137 	m_freem(nam);
    138 	return (error);
    139 }
    140 
    141 /* ARGSUSED */
    142 int
    143 sys_listen(p, v, retval)
    144 	struct proc *p;
    145 	void *v;
    146 	register_t *retval;
    147 {
    148 	register struct sys_listen_args /* {
    149 		syscallarg(int) s;
    150 		syscallarg(int) backlog;
    151 	} */ *uap = v;
    152 	struct file *fp;
    153 	int error;
    154 
    155 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    156 		return (error);
    157 	return (solisten((struct socket *)fp->f_data, SCARG(uap, backlog)));
    158 }
    159 
    160 int
    161 sys_accept(p, v, retval)
    162 	struct proc *p;
    163 	void *v;
    164 	register_t *retval;
    165 {
    166 	register struct sys_accept_args /* {
    167 		syscallarg(int) s;
    168 		syscallarg(struct sockaddr *) name;
    169 		syscallarg(unsigned int *) anamelen;
    170 	} */ *uap = v;
    171 	struct file *fp;
    172 	struct mbuf *nam;
    173 	unsigned int namelen;
    174 	int error, s, tmpfd;
    175 	register struct socket *so;
    176 
    177 	if (SCARG(uap, name) && (error = copyin((caddr_t)SCARG(uap, anamelen),
    178 	    (caddr_t)&namelen, sizeof(namelen))))
    179 		return (error);
    180 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    181 		return (error);
    182 	s = splsoftnet();
    183 	so = (struct socket *)fp->f_data;
    184 	if ((so->so_options & SO_ACCEPTCONN) == 0) {
    185 		splx(s);
    186 		return (EINVAL);
    187 	}
    188 	if ((so->so_state & SS_NBIO) && so->so_qlen == 0) {
    189 		splx(s);
    190 		return (EWOULDBLOCK);
    191 	}
    192 	while (so->so_qlen == 0 && so->so_error == 0) {
    193 		if (so->so_state & SS_CANTRCVMORE) {
    194 			so->so_error = ECONNABORTED;
    195 			break;
    196 		}
    197 		error = tsleep((caddr_t)&so->so_timeo, PSOCK | PCATCH,
    198 			       netcon, 0);
    199 		if (error) {
    200 			splx(s);
    201 			return (error);
    202 		}
    203 	}
    204 	if (so->so_error) {
    205 		error = so->so_error;
    206 		so->so_error = 0;
    207 		splx(s);
    208 		return (error);
    209 	}
    210 	if ((error = falloc(p, &fp, &tmpfd)) != 0) {
    211 		splx(s);
    212 		return (error);
    213 	}
    214 	*retval = tmpfd;
    215 	{ struct socket *aso = so->so_q.tqh_first;
    216 	  if (soqremque(aso, 1) == 0)
    217 		panic("accept");
    218 	  so = aso;
    219 	}
    220 	fp->f_type = DTYPE_SOCKET;
    221 	fp->f_flag = FREAD|FWRITE;
    222 	fp->f_ops = &socketops;
    223 	fp->f_data = (caddr_t)so;
    224 	nam = m_get(M_WAIT, MT_SONAME);
    225 	(void) soaccept(so, nam);
    226 	if (SCARG(uap, name)) {
    227 		if (namelen > nam->m_len)
    228 			namelen = nam->m_len;
    229 		/* SHOULD COPY OUT A CHAIN HERE */
    230 		if ((error = copyout(mtod(nam, caddr_t),
    231 		    (caddr_t)SCARG(uap, name), namelen)) == 0)
    232 			error = copyout((caddr_t)&namelen,
    233 			    (caddr_t)SCARG(uap, anamelen),
    234 			    sizeof(*SCARG(uap, anamelen)));
    235 	}
    236 	m_freem(nam);
    237 	splx(s);
    238 	return (error);
    239 }
    240 
    241 /* ARGSUSED */
    242 int
    243 sys_connect(p, v, retval)
    244 	struct proc *p;
    245 	void *v;
    246 	register_t *retval;
    247 {
    248 	register struct sys_connect_args /* {
    249 		syscallarg(int) s;
    250 		syscallarg(const struct sockaddr *) name;
    251 		syscallarg(unsigned int) namelen;
    252 	} */ *uap = v;
    253 	struct file *fp;
    254 	register struct socket *so;
    255 	struct mbuf *nam;
    256 	int error, s;
    257 
    258 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    259 		return (error);
    260 	so = (struct socket *)fp->f_data;
    261 	if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING))
    262 		return (EALREADY);
    263 	error = sockargs(&nam, SCARG(uap, name), SCARG(uap, namelen),
    264 	    MT_SONAME);
    265 	if (error)
    266 		return (error);
    267 	error = soconnect(so, nam);
    268 	if (error)
    269 		goto bad;
    270 	if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING)) {
    271 		m_freem(nam);
    272 		return (EINPROGRESS);
    273 	}
    274 	s = splsoftnet();
    275 	while ((so->so_state & SS_ISCONNECTING) && so->so_error == 0) {
    276 		error = tsleep((caddr_t)&so->so_timeo, PSOCK | PCATCH,
    277 			       netcon, 0);
    278 		if (error)
    279 			break;
    280 	}
    281 	if (error == 0) {
    282 		error = so->so_error;
    283 		so->so_error = 0;
    284 	}
    285 	splx(s);
    286 bad:
    287 	so->so_state &= ~SS_ISCONNECTING;
    288 	m_freem(nam);
    289 	if (error == ERESTART)
    290 		error = EINTR;
    291 	return (error);
    292 }
    293 
    294 int
    295 sys_socketpair(p, v, retval)
    296 	struct proc *p;
    297 	void *v;
    298 	register_t *retval;
    299 {
    300 	register struct sys_socketpair_args /* {
    301 		syscallarg(int) domain;
    302 		syscallarg(int) type;
    303 		syscallarg(int) protocol;
    304 		syscallarg(int *) rsv;
    305 	} */ *uap = v;
    306 	register struct filedesc *fdp = p->p_fd;
    307 	struct file *fp1, *fp2;
    308 	struct socket *so1, *so2;
    309 	int fd, error, sv[2];
    310 
    311 	error = socreate(SCARG(uap, domain), &so1, SCARG(uap, type),
    312 			 SCARG(uap, protocol));
    313 	if (error)
    314 		return (error);
    315 	error = socreate(SCARG(uap, domain), &so2, SCARG(uap, type),
    316 			 SCARG(uap, protocol));
    317 	if (error)
    318 		goto free1;
    319 	if ((error = falloc(p, &fp1, &fd)) != 0)
    320 		goto free2;
    321 	sv[0] = fd;
    322 	fp1->f_flag = FREAD|FWRITE;
    323 	fp1->f_type = DTYPE_SOCKET;
    324 	fp1->f_ops = &socketops;
    325 	fp1->f_data = (caddr_t)so1;
    326 	if ((error = falloc(p, &fp2, &fd)) != 0)
    327 		goto free3;
    328 	fp2->f_flag = FREAD|FWRITE;
    329 	fp2->f_type = DTYPE_SOCKET;
    330 	fp2->f_ops = &socketops;
    331 	fp2->f_data = (caddr_t)so2;
    332 	sv[1] = fd;
    333 	if ((error = soconnect2(so1, so2)) != 0)
    334 		goto free4;
    335 	if (SCARG(uap, type) == SOCK_DGRAM) {
    336 		/*
    337 		 * Datagram socket connection is asymmetric.
    338 		 */
    339 		 if ((error = soconnect2(so2, so1)) != 0)
    340 			goto free4;
    341 	}
    342 	error = copyout((caddr_t)sv, (caddr_t)SCARG(uap, rsv),
    343 	    2 * sizeof(int));
    344 	return (error);
    345 free4:
    346 	ffree(fp2);
    347 	fdp->fd_ofiles[sv[1]] = 0;
    348 free3:
    349 	ffree(fp1);
    350 	fdp->fd_ofiles[sv[0]] = 0;
    351 free2:
    352 	(void)soclose(so2);
    353 free1:
    354 	(void)soclose(so1);
    355 	return (error);
    356 }
    357 
    358 int
    359 sys_sendto(p, v, retval)
    360 	struct proc *p;
    361 	void *v;
    362 	register_t *retval;
    363 {
    364 	register struct sys_sendto_args /* {
    365 		syscallarg(int) s;
    366 		syscallarg(const void *) buf;
    367 		syscallarg(size_t) len;
    368 		syscallarg(int) flags;
    369 		syscallarg(const struct sockaddr *) to;
    370 		syscallarg(unsigned int) tolen;
    371 	} */ *uap = v;
    372 	struct msghdr msg;
    373 	struct iovec aiov;
    374 
    375 	msg.msg_name = (caddr_t)SCARG(uap, to);		/* XXX kills const */
    376 	msg.msg_namelen = SCARG(uap, tolen);
    377 	msg.msg_iov = &aiov;
    378 	msg.msg_iovlen = 1;
    379 	msg.msg_control = 0;
    380 #ifdef COMPAT_OLDSOCK
    381 	msg.msg_flags = 0;
    382 #endif
    383 	aiov.iov_base = (char *)SCARG(uap, buf);	/* XXX kills const */
    384 	aiov.iov_len = SCARG(uap, len);
    385 	return (sendit(p, SCARG(uap, s), &msg, SCARG(uap, flags), retval));
    386 }
    387 
    388 int
    389 sys_sendmsg(p, v, retval)
    390 	struct proc *p;
    391 	void *v;
    392 	register_t *retval;
    393 {
    394 	register struct sys_sendmsg_args /* {
    395 		syscallarg(int) s;
    396 		syscallarg(const struct msghdr *) msg;
    397 		syscallarg(int) flags;
    398 	} */ *uap = v;
    399 	struct msghdr msg;
    400 	struct iovec aiov[UIO_SMALLIOV], *iov;
    401 	int error;
    402 
    403 	error = copyin(SCARG(uap, msg), (caddr_t)&msg, sizeof(msg));
    404 	if (error)
    405 		return (error);
    406 	if ((unsigned int)msg.msg_iovlen > UIO_SMALLIOV) {
    407 		if ((unsigned int)msg.msg_iovlen > IOV_MAX)
    408 			return (EMSGSIZE);
    409 		MALLOC(iov, struct iovec *,
    410 		    sizeof(struct iovec) * msg.msg_iovlen, M_IOV, M_WAITOK);
    411 	} else
    412 		iov = aiov;
    413 	if ((unsigned int)msg.msg_iovlen > 0) {
    414 		error = copyin((caddr_t)msg.msg_iov, (caddr_t)iov,
    415 		    (size_t)(msg.msg_iovlen * sizeof(struct iovec)));
    416 		if (error)
    417 			goto done;
    418 	}
    419 	msg.msg_iov = iov;
    420 #ifdef COMPAT_OLDSOCK
    421 	msg.msg_flags = 0;
    422 #endif
    423 	error = sendit(p, SCARG(uap, s), &msg, SCARG(uap, flags), retval);
    424 done:
    425 	if (iov != aiov)
    426 		FREE(iov, M_IOV);
    427 	return (error);
    428 }
    429 
    430 int
    431 sendit(p, s, mp, flags, retsize)
    432 	register struct proc *p;
    433 	int s;
    434 	register struct msghdr *mp;
    435 	int flags;
    436 	register_t *retsize;
    437 {
    438 	struct file *fp;
    439 	struct uio auio;
    440 	register struct iovec *iov;
    441 	register int i;
    442 	struct mbuf *to, *control;
    443 	int len, error;
    444 	struct socket *so;
    445 #ifdef KTRACE
    446 	struct iovec *ktriov = NULL;
    447 #endif
    448 
    449 	if ((error = getsock(p->p_fd, s, &fp)) != 0)
    450 		return (error);
    451 	auio.uio_iov = mp->msg_iov;
    452 	auio.uio_iovcnt = mp->msg_iovlen;
    453 	auio.uio_segflg = UIO_USERSPACE;
    454 	auio.uio_rw = UIO_WRITE;
    455 	auio.uio_procp = p;
    456 	auio.uio_offset = 0;			/* XXX */
    457 	auio.uio_resid = 0;
    458 	iov = mp->msg_iov;
    459 	for (i = 0; i < mp->msg_iovlen; i++, iov++) {
    460 #if 0
    461 		/* cannot happen; iov_len is unsigned */
    462 		if (iov->iov_len < 0)
    463 			return (EINVAL);
    464 #endif
    465 		/*
    466 		 * Writes return ssize_t because -1 is returned on error.
    467 		 * Therefore, we must restrict the length to SSIZE_MAX to
    468 		 * avoid garbage return values.
    469 		 */
    470 		auio.uio_resid += iov->iov_len;
    471 		if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX)
    472 			return (EINVAL);
    473 	}
    474 	if (mp->msg_name) {
    475 		error = sockargs(&to, mp->msg_name, mp->msg_namelen,
    476 				 MT_SONAME);
    477 		if (error)
    478 			return (error);
    479 	} else
    480 		to = 0;
    481 	if (mp->msg_control) {
    482 		if (mp->msg_controllen < sizeof(struct cmsghdr)
    483 #ifdef COMPAT_OLDSOCK
    484 		    && mp->msg_flags != MSG_COMPAT
    485 #endif
    486 		) {
    487 			error = EINVAL;
    488 			goto bad;
    489 		}
    490 		error = sockargs(&control, mp->msg_control,
    491 				 mp->msg_controllen, MT_CONTROL);
    492 		if (error)
    493 			goto bad;
    494 #ifdef COMPAT_OLDSOCK
    495 		if (mp->msg_flags == MSG_COMPAT) {
    496 			register struct cmsghdr *cm;
    497 
    498 			M_PREPEND(control, sizeof(*cm), M_WAIT);
    499 			if (control == 0) {
    500 				error = ENOBUFS;
    501 				goto bad;
    502 			} else {
    503 				cm = mtod(control, struct cmsghdr *);
    504 				cm->cmsg_len = control->m_len;
    505 				cm->cmsg_level = SOL_SOCKET;
    506 				cm->cmsg_type = SCM_RIGHTS;
    507 			}
    508 		}
    509 #endif
    510 	} else
    511 		control = 0;
    512 #ifdef KTRACE
    513 	if (KTRPOINT(p, KTR_GENIO)) {
    514 		int iovlen = auio.uio_iovcnt * sizeof(struct iovec);
    515 
    516 		MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK);
    517 		memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen);
    518 	}
    519 #endif
    520 	len = auio.uio_resid;
    521 	so = (struct socket *)fp->f_data;
    522 	error = (*so->so_send)(so, to, &auio, NULL, control, flags);
    523 	if (error) {
    524 		if (auio.uio_resid != len && (error == ERESTART ||
    525 		    error == EINTR || error == EWOULDBLOCK))
    526 			error = 0;
    527 		if (error == EPIPE)
    528 			psignal(p, SIGPIPE);
    529 	}
    530 	if (error == 0)
    531 		*retsize = len - auio.uio_resid;
    532 #ifdef KTRACE
    533 	if (ktriov != NULL) {
    534 		if (error == 0)
    535 			ktrgenio(p->p_tracep, s, UIO_WRITE,
    536 				ktriov, *retsize, error);
    537 		FREE(ktriov, M_TEMP);
    538 	}
    539 #endif
    540 bad:
    541 	if (to)
    542 		m_freem(to);
    543 	return (error);
    544 }
    545 
    546 int
    547 sys_recvfrom(p, v, retval)
    548 	struct proc *p;
    549 	void *v;
    550 	register_t *retval;
    551 {
    552 	register struct sys_recvfrom_args /* {
    553 		syscallarg(int) s;
    554 		syscallarg(void *) buf;
    555 		syscallarg(size_t) len;
    556 		syscallarg(int) flags;
    557 		syscallarg(struct sockaddr *) from;
    558 		syscallarg(unsigned int *) fromlenaddr;
    559 	} */ *uap = v;
    560 	struct msghdr msg;
    561 	struct iovec aiov;
    562 	int error;
    563 
    564 	if (SCARG(uap, fromlenaddr)) {
    565 		error = copyin((caddr_t)SCARG(uap, fromlenaddr),
    566 			       (caddr_t)&msg.msg_namelen,
    567 			       sizeof(msg.msg_namelen));
    568 		if (error)
    569 			return (error);
    570 	} else
    571 		msg.msg_namelen = 0;
    572 	msg.msg_name = (caddr_t)SCARG(uap, from);
    573 	msg.msg_iov = &aiov;
    574 	msg.msg_iovlen = 1;
    575 	aiov.iov_base = SCARG(uap, buf);
    576 	aiov.iov_len = SCARG(uap, len);
    577 	msg.msg_control = 0;
    578 	msg.msg_flags = SCARG(uap, flags);
    579 	return (recvit(p, SCARG(uap, s), &msg,
    580 		       (caddr_t)SCARG(uap, fromlenaddr), retval));
    581 }
    582 
    583 int
    584 sys_recvmsg(p, v, retval)
    585 	struct proc *p;
    586 	void *v;
    587 	register_t *retval;
    588 {
    589 	register struct sys_recvmsg_args /* {
    590 		syscallarg(int) s;
    591 		syscallarg(struct msghdr *) msg;
    592 		syscallarg(int) flags;
    593 	} */ *uap = v;
    594 	struct msghdr msg;
    595 	struct iovec aiov[UIO_SMALLIOV], *uiov, *iov;
    596 	register int error;
    597 
    598 	error = copyin((caddr_t)SCARG(uap, msg), (caddr_t)&msg,
    599 		       sizeof(msg));
    600 	if (error)
    601 		return (error);
    602 	if ((unsigned int)msg.msg_iovlen > UIO_SMALLIOV) {
    603 		if ((unsigned int)msg.msg_iovlen > IOV_MAX)
    604 			return (EMSGSIZE);
    605 		MALLOC(iov, struct iovec *,
    606 		    sizeof(struct iovec) * msg.msg_iovlen, M_IOV, M_WAITOK);
    607 	} else
    608 		iov = aiov;
    609 	if ((unsigned int)msg.msg_iovlen > 0) {
    610 		error = copyin((caddr_t)msg.msg_iov, (caddr_t)iov,
    611 		    (size_t)(msg.msg_iovlen * sizeof(struct iovec)));
    612 		if (error)
    613 			goto done;
    614 	}
    615 	uiov = msg.msg_iov;
    616 	msg.msg_iov = iov;
    617 #ifdef COMPAT_OLDSOCK
    618 	msg.msg_flags = SCARG(uap, flags) &~ MSG_COMPAT;
    619 #else
    620 	msg.msg_flags = SCARG(uap, flags);
    621 #endif
    622 	if ((error = recvit(p, SCARG(uap, s), &msg, (caddr_t)0, retval)) == 0) {
    623 		msg.msg_iov = uiov;
    624 		error = copyout((caddr_t)&msg, (caddr_t)SCARG(uap, msg),
    625 		    sizeof(msg));
    626 	}
    627 done:
    628 	if (iov != aiov)
    629 		FREE(iov, M_IOV);
    630 	return (error);
    631 }
    632 
    633 int
    634 recvit(p, s, mp, namelenp, retsize)
    635 	register struct proc *p;
    636 	int s;
    637 	register struct msghdr *mp;
    638 	caddr_t namelenp;
    639 	register_t *retsize;
    640 {
    641 	struct file *fp;
    642 	struct uio auio;
    643 	register struct iovec *iov;
    644 	register int i;
    645 	int len, error;
    646 	struct mbuf *from = 0, *control = 0;
    647 	struct socket *so;
    648 #ifdef KTRACE
    649 	struct iovec *ktriov = NULL;
    650 #endif
    651 
    652 	if ((error = getsock(p->p_fd, s, &fp)) != 0)
    653 		return (error);
    654 	auio.uio_iov = mp->msg_iov;
    655 	auio.uio_iovcnt = mp->msg_iovlen;
    656 	auio.uio_segflg = UIO_USERSPACE;
    657 	auio.uio_rw = UIO_READ;
    658 	auio.uio_procp = p;
    659 	auio.uio_offset = 0;			/* XXX */
    660 	auio.uio_resid = 0;
    661 	iov = mp->msg_iov;
    662 	for (i = 0; i < mp->msg_iovlen; i++, iov++) {
    663 #if 0
    664 		/* cannot happen iov_len is unsigned */
    665 		if (iov->iov_len < 0)
    666 			return (EINVAL);
    667 #endif
    668 		/*
    669 		 * Reads return ssize_t because -1 is returned on error.
    670 		 * Therefore we must restrict the length to SSIZE_MAX to
    671 		 * avoid garbage return values.
    672 		 */
    673 		auio.uio_resid += iov->iov_len;
    674 		if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX)
    675 			return (EINVAL);
    676 	}
    677 #ifdef KTRACE
    678 	if (KTRPOINT(p, KTR_GENIO)) {
    679 		int iovlen = auio.uio_iovcnt * sizeof(struct iovec);
    680 
    681 		MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK);
    682 		memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen);
    683 	}
    684 #endif
    685 	len = auio.uio_resid;
    686 	so = (struct socket *)fp->f_data;
    687 	error = (*so->so_receive)(so, &from, &auio, NULL,
    688 			  mp->msg_control ? &control : NULL, &mp->msg_flags);
    689 	if (error) {
    690 		if (auio.uio_resid != len && (error == ERESTART ||
    691 		    error == EINTR || error == EWOULDBLOCK))
    692 			error = 0;
    693 	}
    694 #ifdef KTRACE
    695 	if (ktriov != NULL) {
    696 		if (error == 0)
    697 			ktrgenio(p->p_tracep, s, UIO_READ,
    698 				ktriov, len - auio.uio_resid, error);
    699 		FREE(ktriov, M_TEMP);
    700 	}
    701 #endif
    702 	if (error)
    703 		goto out;
    704 	*retsize = len - auio.uio_resid;
    705 	if (mp->msg_name) {
    706 		len = mp->msg_namelen;
    707 		if (len <= 0 || from == 0)
    708 			len = 0;
    709 		else {
    710 #ifdef COMPAT_OLDSOCK
    711 			if (mp->msg_flags & MSG_COMPAT)
    712 				mtod(from, struct osockaddr *)->sa_family =
    713 				    mtod(from, struct sockaddr *)->sa_family;
    714 #endif
    715 			if (len > from->m_len)
    716 				len = from->m_len;
    717 			/* else if len < from->m_len ??? */
    718 			error = copyout(mtod(from, caddr_t),
    719 					(caddr_t)mp->msg_name, (unsigned)len);
    720 			if (error)
    721 				goto out;
    722 		}
    723 		mp->msg_namelen = len;
    724 		if (namelenp &&
    725 		    (error = copyout((caddr_t)&len, namelenp, sizeof(int)))) {
    726 #ifdef COMPAT_OLDSOCK
    727 			if (mp->msg_flags & MSG_COMPAT)
    728 				error = 0;	/* old recvfrom didn't check */
    729 			else
    730 #endif
    731 			goto out;
    732 		}
    733 	}
    734 	if (mp->msg_control) {
    735 #ifdef COMPAT_OLDSOCK
    736 		/*
    737 		 * We assume that old recvmsg calls won't receive access
    738 		 * rights and other control info, esp. as control info
    739 		 * is always optional and those options didn't exist in 4.3.
    740 		 * If we receive rights, trim the cmsghdr; anything else
    741 		 * is tossed.
    742 		 */
    743 		if (control && mp->msg_flags & MSG_COMPAT) {
    744 			if (mtod(control, struct cmsghdr *)->cmsg_level !=
    745 			    SOL_SOCKET ||
    746 			    mtod(control, struct cmsghdr *)->cmsg_type !=
    747 			    SCM_RIGHTS) {
    748 				mp->msg_controllen = 0;
    749 				goto out;
    750 			}
    751 			control->m_len -= sizeof(struct cmsghdr);
    752 			control->m_data += sizeof(struct cmsghdr);
    753 		}
    754 #endif
    755 		len = mp->msg_controllen;
    756 		if (len <= 0 || control == 0)
    757 			len = 0;
    758 		else {
    759 			struct mbuf *m = control;
    760 			caddr_t p = (caddr_t)mp->msg_control;
    761 
    762 			do {
    763 				i = m->m_len;
    764 				if (len < i) {
    765 					mp->msg_flags |= MSG_CTRUNC;
    766 					i = len;
    767 				}
    768 				error = copyout(mtod(m, caddr_t), p,
    769 				    (unsigned)i);
    770 				if (m->m_next)
    771 					i = ALIGN(i);
    772 				p += i;
    773 				len -= i;
    774 				if (error != 0 || len <= 0)
    775 					break;
    776 			} while ((m = m->m_next) != NULL);
    777 			len = p - (caddr_t)mp->msg_control;
    778 		}
    779 		mp->msg_controllen = len;
    780 	}
    781 out:
    782 	if (from)
    783 		m_freem(from);
    784 	if (control)
    785 		m_freem(control);
    786 	return (error);
    787 }
    788 
    789 /* ARGSUSED */
    790 int
    791 sys_shutdown(p, v, retval)
    792 	struct proc *p;
    793 	void *v;
    794 	register_t *retval;
    795 {
    796 	register struct sys_shutdown_args /* {
    797 		syscallarg(int) s;
    798 		syscallarg(int) how;
    799 	} */ *uap = v;
    800 	struct file *fp;
    801 	int error;
    802 
    803 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    804 		return (error);
    805 	return (soshutdown((struct socket *)fp->f_data, SCARG(uap, how)));
    806 }
    807 
    808 /* ARGSUSED */
    809 int
    810 sys_setsockopt(p, v, retval)
    811 	struct proc *p;
    812 	void *v;
    813 	register_t *retval;
    814 {
    815 	register struct sys_setsockopt_args /* {
    816 		syscallarg(int) s;
    817 		syscallarg(int) level;
    818 		syscallarg(int) name;
    819 		syscallarg(const void *) val;
    820 		syscallarg(unsigned int) valsize;
    821 	} */ *uap = v;
    822 	struct file *fp;
    823 	struct mbuf *m = NULL;
    824 	int error;
    825 
    826 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    827 		return (error);
    828 	if (SCARG(uap, valsize) > MLEN)
    829 		return (EINVAL);
    830 	if (SCARG(uap, val)) {
    831 		m = m_get(M_WAIT, MT_SOOPTS);
    832 		error = copyin(SCARG(uap, val), mtod(m, caddr_t),
    833 			       SCARG(uap, valsize));
    834 		if (error) {
    835 			(void) m_free(m);
    836 			return (error);
    837 		}
    838 		m->m_len = SCARG(uap, valsize);
    839 	}
    840 	return (sosetopt((struct socket *)fp->f_data, SCARG(uap, level),
    841 			 SCARG(uap, name), m));
    842 }
    843 
    844 /* ARGSUSED */
    845 int
    846 sys_getsockopt(p, v, retval)
    847 	struct proc *p;
    848 	void *v;
    849 	register_t *retval;
    850 {
    851 	register struct sys_getsockopt_args /* {
    852 		syscallarg(int) s;
    853 		syscallarg(int) level;
    854 		syscallarg(int) name;
    855 		syscallarg(void *) val;
    856 		syscallarg(unsigned int *) avalsize;
    857 	} */ *uap = v;
    858 	struct file *fp;
    859 	struct mbuf *m = NULL;
    860 	unsigned int valsize;
    861 	int error;
    862 
    863 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    864 		return (error);
    865 	if (SCARG(uap, val)) {
    866 		error = copyin((caddr_t)SCARG(uap, avalsize),
    867 			       (caddr_t)&valsize, sizeof(valsize));
    868 		if (error)
    869 			return (error);
    870 	} else
    871 		valsize = 0;
    872 	if ((error = sogetopt((struct socket *)fp->f_data, SCARG(uap, level),
    873 	    SCARG(uap, name), &m)) == 0 && SCARG(uap, val) && valsize &&
    874 	    m != NULL) {
    875 		if (valsize > m->m_len)
    876 			valsize = m->m_len;
    877 		error = copyout(mtod(m, caddr_t), SCARG(uap, val),
    878 		    valsize);
    879 		if (error == 0)
    880 			error = copyout((caddr_t)&valsize,
    881 			    (caddr_t)SCARG(uap, avalsize), sizeof(valsize));
    882 	}
    883 	if (m != NULL)
    884 		(void) m_free(m);
    885 	return (error);
    886 }
    887 
    888 /* ARGSUSED */
    889 int
    890 sys_pipe(p, v, retval)
    891 	struct proc *p;
    892 	void *v;
    893 	register_t *retval;
    894 {
    895 	register struct filedesc *fdp = p->p_fd;
    896 	struct file *rf, *wf;
    897 	struct socket *rso, *wso;
    898 	int fd, error;
    899 
    900 	if ((error = socreate(AF_LOCAL, &rso, SOCK_STREAM, 0)) != 0)
    901 		return (error);
    902 	if ((error = socreate(AF_LOCAL, &wso, SOCK_STREAM, 0)) != 0)
    903 		goto free1;
    904 	if ((error = falloc(p, &rf, &fd)) != 0)
    905 		goto free2;
    906 	retval[0] = fd;
    907 	rf->f_flag = FREAD;
    908 	rf->f_type = DTYPE_SOCKET;
    909 	rf->f_ops = &socketops;
    910 	rf->f_data = (caddr_t)rso;
    911 	if ((error = falloc(p, &wf, &fd)) != 0)
    912 		goto free3;
    913 	wf->f_flag = FWRITE;
    914 	wf->f_type = DTYPE_SOCKET;
    915 	wf->f_ops = &socketops;
    916 	wf->f_data = (caddr_t)wso;
    917 	retval[1] = fd;
    918 	if ((error = unp_connect2(wso, rso)) != 0)
    919 		goto free4;
    920 	return (0);
    921 free4:
    922 	ffree(wf);
    923 	fdp->fd_ofiles[retval[1]] = 0;
    924 free3:
    925 	ffree(rf);
    926 	fdp->fd_ofiles[retval[0]] = 0;
    927 free2:
    928 	(void)soclose(wso);
    929 free1:
    930 	(void)soclose(rso);
    931 	return (error);
    932 }
    933 
    934 /*
    935  * Get socket name.
    936  */
    937 /* ARGSUSED */
    938 int
    939 sys_getsockname(p, v, retval)
    940 	struct proc *p;
    941 	void *v;
    942 	register_t *retval;
    943 {
    944 	register struct sys_getsockname_args /* {
    945 		syscallarg(int) fdes;
    946 		syscallarg(struct sockaddr *) asa;
    947 		syscallarg(unsigned int *) alen;
    948 	} */ *uap = v;
    949 	struct file *fp;
    950 	register struct socket *so;
    951 	struct mbuf *m;
    952 	unsigned int len;
    953 	int error;
    954 
    955 	if ((error = getsock(p->p_fd, SCARG(uap, fdes), &fp)) != 0)
    956 		return (error);
    957 	error = copyin((caddr_t)SCARG(uap, alen), (caddr_t)&len, sizeof(len));
    958 	if (error)
    959 		return (error);
    960 	so = (struct socket *)fp->f_data;
    961 	m = m_getclr(M_WAIT, MT_SONAME);
    962 	error = (*so->so_proto->pr_usrreq)(so, PRU_SOCKADDR, (struct mbuf *)0,
    963 	    m, (struct mbuf *)0, (struct proc *)0);
    964 	if (error)
    965 		goto bad;
    966 	if (len > m->m_len)
    967 		len = m->m_len;
    968 	error = copyout(mtod(m, caddr_t), (caddr_t)SCARG(uap, asa), len);
    969 	if (error == 0)
    970 		error = copyout((caddr_t)&len, (caddr_t)SCARG(uap, alen),
    971 		    sizeof(len));
    972 bad:
    973 	m_freem(m);
    974 	return (error);
    975 }
    976 
    977 /*
    978  * Get name of peer for connected socket.
    979  */
    980 /* ARGSUSED */
    981 int
    982 sys_getpeername(p, v, retval)
    983 	struct proc *p;
    984 	void *v;
    985 	register_t *retval;
    986 {
    987 	register struct sys_getpeername_args /* {
    988 		syscallarg(int) fdes;
    989 		syscallarg(struct sockaddr *) asa;
    990 		syscallarg(unsigned int *) alen;
    991 	} */ *uap = v;
    992 	struct file *fp;
    993 	register struct socket *so;
    994 	struct mbuf *m;
    995 	unsigned int len;
    996 	int error;
    997 
    998 	if ((error = getsock(p->p_fd, SCARG(uap, fdes), &fp)) != 0)
    999 		return (error);
   1000 	so = (struct socket *)fp->f_data;
   1001 	if ((so->so_state & (SS_ISCONNECTED|SS_ISCONFIRMING)) == 0)
   1002 		return (ENOTCONN);
   1003 	error = copyin((caddr_t)SCARG(uap, alen), (caddr_t)&len, sizeof(len));
   1004 	if (error)
   1005 		return (error);
   1006 	m = m_getclr(M_WAIT, MT_SONAME);
   1007 	error = (*so->so_proto->pr_usrreq)(so, PRU_PEERADDR, (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)
   1015 		goto bad;
   1016 	error = copyout((caddr_t)&len, (caddr_t)SCARG(uap, alen), sizeof(len));
   1017 bad:
   1018 	m_freem(m);
   1019 	return (error);
   1020 }
   1021 
   1022 /*
   1023  * XXX In a perfect world, we wouldn't pass around socket control
   1024  * XXX arguments in mbufs, and this could go away.
   1025  */
   1026 int
   1027 sockargs(mp, buf, buflen, type)
   1028 	struct mbuf **mp;
   1029 	const void *buf;
   1030 	int buflen, type;
   1031 {
   1032 	register struct sockaddr *sa;
   1033 	register struct mbuf *m;
   1034 	int error;
   1035 
   1036 	/*
   1037 	 * We can't allow socket names > UCHAR_MAX in length, since that
   1038 	 * will overflow sa_len.   Control data more than a page size in
   1039 	 * length is just too much.
   1040 	 */
   1041 	if ((u_int)buflen > (type == MT_SONAME ? UCHAR_MAX : PAGE_SIZE))
   1042 		return (EINVAL);
   1043 
   1044 	/* Allocate an mbuf to hold the arguments. */
   1045 	m = m_get(M_WAIT, type);
   1046 	if ((u_int)buflen > MLEN) {
   1047 		/*
   1048 		 * Won't fit into a regular mbuf, so we allocate just
   1049 		 * enough external storage to hold the argument.
   1050 		 */
   1051 		MEXTMALLOC(m, buflen, M_WAITOK);
   1052 	}
   1053 	m->m_len = buflen;
   1054 	error = copyin(buf, mtod(m, caddr_t), (u_int)buflen);
   1055 	if (error) {
   1056 		(void) m_free(m);
   1057 		return (error);
   1058 	}
   1059 	*mp = m;
   1060 	if (type == MT_SONAME) {
   1061 		sa = mtod(m, struct sockaddr *);
   1062 
   1063 #if defined(COMPAT_OLDSOCK) && BYTE_ORDER != BIG_ENDIAN
   1064 		if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
   1065 			sa->sa_family = sa->sa_len;
   1066 #endif
   1067 		sa->sa_len = buflen;
   1068 	}
   1069 	return (0);
   1070 }
   1071 
   1072 int
   1073 getsock(fdp, fdes, fpp)
   1074 	struct filedesc *fdp;
   1075 	int fdes;
   1076 	struct file **fpp;
   1077 {
   1078 	register struct file *fp;
   1079 
   1080 	if ((unsigned)fdes >= fdp->fd_nfiles ||
   1081 	    (fp = fdp->fd_ofiles[fdes]) == NULL)
   1082 		return (EBADF);
   1083 	if (fp->f_type != DTYPE_SOCKET)
   1084 		return (ENOTSOCK);
   1085 	*fpp = fp;
   1086 	return (0);
   1087 }
   1088