Home | History | Annotate | Line # | Download | only in kern
uipc_syscalls.c revision 1.59
      1 /*	$NetBSD: uipc_syscalls.c,v 1.59 2001/06/14 20:32:47 thorpej 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_SET_MATURE(fp);
    103 		FILE_UNUSE(fp, p);
    104 		*retval = fd;
    105 	}
    106 	return (error);
    107 }
    108 
    109 /* ARGSUSED */
    110 int
    111 sys_bind(struct proc *p, void *v, register_t *retval)
    112 {
    113 	struct sys_bind_args /* {
    114 		syscallarg(int)				s;
    115 		syscallarg(const struct sockaddr *)	name;
    116 		syscallarg(unsigned int)		namelen;
    117 	} */ *uap = v;
    118 	struct file	*fp;
    119 	struct mbuf	*nam;
    120 	int		error;
    121 
    122 	/* getsock() will use the descriptor for us */
    123 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    124 		return (error);
    125 	error = sockargs(&nam, SCARG(uap, name), SCARG(uap, namelen),
    126 	    MT_SONAME);
    127 	if (error) {
    128 		FILE_UNUSE(fp, p);
    129 		return (error);
    130 	}
    131 	error = sobind((struct socket *)fp->f_data, nam, p);
    132 	m_freem(nam);
    133 	FILE_UNUSE(fp, p);
    134 	return (error);
    135 }
    136 
    137 /* ARGSUSED */
    138 int
    139 sys_listen(struct proc *p, void *v, register_t *retval)
    140 {
    141 	struct sys_listen_args /* {
    142 		syscallarg(int)	s;
    143 		syscallarg(int)	backlog;
    144 	} */ *uap = v;
    145 	struct file	*fp;
    146 	int		error;
    147 
    148 	/* getsock() will use the descriptor for us */
    149 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    150 		return (error);
    151 	error = solisten((struct socket *)fp->f_data, SCARG(uap, backlog));
    152 	FILE_UNUSE(fp, p);
    153 	return (error);
    154 }
    155 
    156 int
    157 sys_accept(struct proc *p, void *v, register_t *retval)
    158 {
    159 	struct sys_accept_args /* {
    160 		syscallarg(int)			s;
    161 		syscallarg(struct sockaddr *)	name;
    162 		syscallarg(unsigned int *)	anamelen;
    163 	} */ *uap = v;
    164 	struct filedesc	*fdp;
    165 	struct file	*fp;
    166 	struct mbuf	*nam;
    167 	unsigned int	namelen;
    168 	int		error, s, fd;
    169 	struct socket	*so;
    170 
    171 	fdp = p->p_fd;
    172 	if (SCARG(uap, name) && (error = copyin((caddr_t)SCARG(uap, anamelen),
    173 	    (caddr_t)&namelen, sizeof(namelen))))
    174 		return (error);
    175 	if (SCARG(uap, name) != NULL &&
    176 	    uvm_useracc((caddr_t)SCARG(uap, name), sizeof(struct sockaddr),
    177 	     B_WRITE) == FALSE)
    178 		return (EFAULT);
    179 
    180 	/* getsock() will use the descriptor for us */
    181 	if ((error = getsock(fdp, SCARG(uap, s), &fp)) != 0)
    182 		return (error);
    183 	s = splsoftnet();
    184 	so = (struct socket *)fp->f_data;
    185 	FILE_UNUSE(fp, p);
    186 	if (!(so->so_proto->pr_flags & PR_LISTEN)) {
    187 		splx(s);
    188 		return (EOPNOTSUPP);
    189 	}
    190 	if ((so->so_options & SO_ACCEPTCONN) == 0) {
    191 		splx(s);
    192 		return (EINVAL);
    193 	}
    194 	if ((so->so_state & SS_NBIO) && so->so_qlen == 0) {
    195 		splx(s);
    196 		return (EWOULDBLOCK);
    197 	}
    198 	while (so->so_qlen == 0 && so->so_error == 0) {
    199 		if (so->so_state & SS_CANTRCVMORE) {
    200 			so->so_error = ECONNABORTED;
    201 			break;
    202 		}
    203 		error = tsleep((caddr_t)&so->so_timeo, PSOCK | PCATCH,
    204 			       netcon, 0);
    205 		if (error) {
    206 			splx(s);
    207 			return (error);
    208 		}
    209 	}
    210 	if (so->so_error) {
    211 		error = so->so_error;
    212 		so->so_error = 0;
    213 		splx(s);
    214 		return (error);
    215 	}
    216 	/* falloc() will use the descriptor for us */
    217 	if ((error = falloc(p, &fp, &fd)) != 0) {
    218 		splx(s);
    219 		return (error);
    220 	}
    221 	*retval = fd;
    222 	{ struct socket *aso = so->so_q.tqh_first;
    223 	  if (soqremque(aso, 1) == 0)
    224 		panic("accept");
    225 	  so = aso;
    226 	}
    227 	fp->f_type = DTYPE_SOCKET;
    228 	fp->f_flag = FREAD|FWRITE;
    229 	fp->f_ops = &socketops;
    230 	fp->f_data = (caddr_t)so;
    231 	FILE_UNUSE(fp, p);
    232 	nam = m_get(M_WAIT, MT_SONAME);
    233 	if ((error = soaccept(so, nam)) == 0 && SCARG(uap, name)) {
    234 		if (namelen > nam->m_len)
    235 			namelen = nam->m_len;
    236 		/* SHOULD COPY OUT A CHAIN HERE */
    237 		if ((error = copyout(mtod(nam, caddr_t),
    238 		    (caddr_t)SCARG(uap, name), namelen)) == 0)
    239 			error = copyout((caddr_t)&namelen,
    240 			    (caddr_t)SCARG(uap, anamelen),
    241 			    sizeof(*SCARG(uap, anamelen)));
    242 	}
    243 	/* if an error occured, free the file descriptor */
    244 	if (error) {
    245 		fdremove(fdp, fd);
    246 		ffree(fp);
    247 	}
    248 	m_freem(nam);
    249 	splx(s);
    250 	FILE_SET_MATURE(fp);
    251 	return (error);
    252 }
    253 
    254 /* ARGSUSED */
    255 int
    256 sys_connect(struct proc *p, void *v, register_t *retval)
    257 {
    258 	struct sys_connect_args /* {
    259 		syscallarg(int)				s;
    260 		syscallarg(const struct sockaddr *)	name;
    261 		syscallarg(unsigned int)		namelen;
    262 	} */ *uap = v;
    263 	struct file	*fp;
    264 	struct socket	*so;
    265 	struct mbuf	*nam;
    266 	int		error, s;
    267 
    268 	/* getsock() will use the descriptor for us */
    269 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    270 		return (error);
    271 	so = (struct socket *)fp->f_data;
    272 	FILE_UNUSE(fp, p);
    273 	if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING))
    274 		return (EALREADY);
    275 	error = sockargs(&nam, SCARG(uap, name), SCARG(uap, namelen),
    276 	    MT_SONAME);
    277 	if (error)
    278 		return (error);
    279 	error = soconnect(so, nam);
    280 	if (error)
    281 		goto bad;
    282 	if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING)) {
    283 		m_freem(nam);
    284 		return (EINPROGRESS);
    285 	}
    286 	s = splsoftnet();
    287 	while ((so->so_state & SS_ISCONNECTING) && so->so_error == 0) {
    288 		error = tsleep((caddr_t)&so->so_timeo, PSOCK | PCATCH,
    289 			       netcon, 0);
    290 		if (error)
    291 			break;
    292 	}
    293 	if (error == 0) {
    294 		error = so->so_error;
    295 		so->so_error = 0;
    296 	}
    297 	splx(s);
    298  bad:
    299 	so->so_state &= ~SS_ISCONNECTING;
    300 	m_freem(nam);
    301 	if (error == ERESTART)
    302 		error = EINTR;
    303 	return (error);
    304 }
    305 
    306 int
    307 sys_socketpair(struct proc *p, void *v, register_t *retval)
    308 {
    309 	struct sys_socketpair_args /* {
    310 		syscallarg(int)		domain;
    311 		syscallarg(int)		type;
    312 		syscallarg(int)		protocol;
    313 		syscallarg(int *)	rsv;
    314 	} */ *uap = v;
    315 	struct filedesc	*fdp;
    316 	struct file	*fp1, *fp2;
    317 	struct socket	*so1, *so2;
    318 	int		fd, error, sv[2];
    319 
    320 	fdp = p->p_fd;
    321 	error = socreate(SCARG(uap, domain), &so1, SCARG(uap, type),
    322 			 SCARG(uap, protocol));
    323 	if (error)
    324 		return (error);
    325 	error = socreate(SCARG(uap, domain), &so2, SCARG(uap, type),
    326 			 SCARG(uap, protocol));
    327 	if (error)
    328 		goto free1;
    329 	/* falloc() will use the descriptor for us */
    330 	if ((error = falloc(p, &fp1, &fd)) != 0)
    331 		goto free2;
    332 	sv[0] = fd;
    333 	fp1->f_flag = FREAD|FWRITE;
    334 	fp1->f_type = DTYPE_SOCKET;
    335 	fp1->f_ops = &socketops;
    336 	fp1->f_data = (caddr_t)so1;
    337 	if ((error = falloc(p, &fp2, &fd)) != 0)
    338 		goto free3;
    339 	fp2->f_flag = FREAD|FWRITE;
    340 	fp2->f_type = DTYPE_SOCKET;
    341 	fp2->f_ops = &socketops;
    342 	fp2->f_data = (caddr_t)so2;
    343 	sv[1] = fd;
    344 	if ((error = soconnect2(so1, so2)) != 0)
    345 		goto free4;
    346 	if (SCARG(uap, type) == SOCK_DGRAM) {
    347 		/*
    348 		 * Datagram socket connection is asymmetric.
    349 		 */
    350 		 if ((error = soconnect2(so2, so1)) != 0)
    351 			goto free4;
    352 	}
    353 	error = copyout((caddr_t)sv, (caddr_t)SCARG(uap, rsv),
    354 	    2 * sizeof(int));
    355 	FILE_SET_MATURE(fp1);
    356 	FILE_SET_MATURE(fp2);
    357 	FILE_UNUSE(fp1, p);
    358 	FILE_UNUSE(fp2, p);
    359 	return (error);
    360  free4:
    361 	FILE_UNUSE(fp2, p);
    362 	ffree(fp2);
    363 	fdremove(fdp, sv[1]);
    364  free3:
    365 	FILE_UNUSE(fp1, p);
    366 	ffree(fp1);
    367 	fdremove(fdp, sv[0]);
    368  free2:
    369 	(void)soclose(so2);
    370  free1:
    371 	(void)soclose(so1);
    372 	return (error);
    373 }
    374 
    375 int
    376 sys_sendto(struct proc *p, void *v, register_t *retval)
    377 {
    378 	struct sys_sendto_args /* {
    379 		syscallarg(int)				s;
    380 		syscallarg(const void *)		buf;
    381 		syscallarg(size_t)			len;
    382 		syscallarg(int)				flags;
    383 		syscallarg(const struct sockaddr *)	to;
    384 		syscallarg(unsigned int)		tolen;
    385 	} */ *uap = v;
    386 	struct msghdr	msg;
    387 	struct iovec	aiov;
    388 
    389 	msg.msg_name = (caddr_t)SCARG(uap, to);		/* XXX kills const */
    390 	msg.msg_namelen = SCARG(uap, tolen);
    391 	msg.msg_iov = &aiov;
    392 	msg.msg_iovlen = 1;
    393 	msg.msg_control = 0;
    394 #ifdef COMPAT_OLDSOCK
    395 	msg.msg_flags = 0;
    396 #endif
    397 	aiov.iov_base = (char *)SCARG(uap, buf);	/* XXX kills const */
    398 	aiov.iov_len = SCARG(uap, len);
    399 	return (sendit(p, SCARG(uap, s), &msg, SCARG(uap, flags), retval));
    400 }
    401 
    402 int
    403 sys_sendmsg(struct proc *p, void *v, register_t *retval)
    404 {
    405 	struct sys_sendmsg_args /* {
    406 		syscallarg(int)				s;
    407 		syscallarg(const struct msghdr *)	msg;
    408 		syscallarg(int)				flags;
    409 	} */ *uap = v;
    410 	struct msghdr	msg;
    411 	struct iovec	aiov[UIO_SMALLIOV], *iov;
    412 	int		error;
    413 
    414 	error = copyin(SCARG(uap, msg), (caddr_t)&msg, sizeof(msg));
    415 	if (error)
    416 		return (error);
    417 	if ((unsigned int)msg.msg_iovlen > UIO_SMALLIOV) {
    418 		if ((unsigned int)msg.msg_iovlen > IOV_MAX)
    419 			return (EMSGSIZE);
    420 		iov = malloc(sizeof(struct iovec) * msg.msg_iovlen,
    421 		    M_IOV, M_WAITOK);
    422 	} else
    423 		iov = aiov;
    424 	if ((unsigned int)msg.msg_iovlen > 0) {
    425 		error = copyin((caddr_t)msg.msg_iov, (caddr_t)iov,
    426 		    (size_t)(msg.msg_iovlen * sizeof(struct iovec)));
    427 		if (error)
    428 			goto done;
    429 	}
    430 	msg.msg_iov = iov;
    431 #ifdef COMPAT_OLDSOCK
    432 	msg.msg_flags = 0;
    433 #endif
    434 	error = sendit(p, SCARG(uap, s), &msg, SCARG(uap, flags), retval);
    435 done:
    436 	if (iov != aiov)
    437 		free(iov, M_IOV);
    438 	return (error);
    439 }
    440 
    441 int
    442 sendit(struct proc *p, int s, struct msghdr *mp, int flags, register_t *retsize)
    443 {
    444 	struct file	*fp;
    445 	struct uio	auio;
    446 	struct iovec	*iov;
    447 	int		i, len, error;
    448 	struct mbuf	*to, *control;
    449 	struct socket	*so;
    450 #ifdef KTRACE
    451 	struct iovec	*ktriov;
    452 #endif
    453 
    454 #ifdef KTRACE
    455 	ktriov = NULL;
    456 #endif
    457 	/* getsock() will use the descriptor for us */
    458 	if ((error = getsock(p->p_fd, s, &fp)) != 0)
    459 		return (error);
    460 	auio.uio_iov = mp->msg_iov;
    461 	auio.uio_iovcnt = mp->msg_iovlen;
    462 	auio.uio_segflg = UIO_USERSPACE;
    463 	auio.uio_rw = UIO_WRITE;
    464 	auio.uio_procp = p;
    465 	auio.uio_offset = 0;			/* XXX */
    466 	auio.uio_resid = 0;
    467 	iov = mp->msg_iov;
    468 	for (i = 0; i < mp->msg_iovlen; i++, iov++) {
    469 #if 0
    470 		/* cannot happen; iov_len is unsigned */
    471 		if (iov->iov_len < 0) {
    472 			error = EINVAL;
    473 			goto out;
    474 		}
    475 #endif
    476 		/*
    477 		 * Writes return ssize_t because -1 is returned on error.
    478 		 * Therefore, we must restrict the length to SSIZE_MAX to
    479 		 * avoid garbage return values.
    480 		 */
    481 		auio.uio_resid += iov->iov_len;
    482 		if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
    483 			error = EINVAL;
    484 			goto out;
    485 		}
    486 	}
    487 	if (mp->msg_name) {
    488 		error = sockargs(&to, mp->msg_name, mp->msg_namelen,
    489 				 MT_SONAME);
    490 		if (error)
    491 			goto out;
    492 	} else
    493 		to = 0;
    494 	if (mp->msg_control) {
    495 		if (mp->msg_controllen < sizeof(struct cmsghdr)
    496 #ifdef COMPAT_OLDSOCK
    497 		    && mp->msg_flags != MSG_COMPAT
    498 #endif
    499 		) {
    500 			error = EINVAL;
    501 			goto bad;
    502 		}
    503 		error = sockargs(&control, mp->msg_control,
    504 				 mp->msg_controllen, MT_CONTROL);
    505 		if (error)
    506 			goto bad;
    507 #ifdef COMPAT_OLDSOCK
    508 		if (mp->msg_flags == MSG_COMPAT) {
    509 			struct cmsghdr *cm;
    510 
    511 			M_PREPEND(control, sizeof(*cm), M_WAIT);
    512 			if (control == 0) {
    513 				error = ENOBUFS;
    514 				goto bad;
    515 			} else {
    516 				cm = mtod(control, struct cmsghdr *);
    517 				cm->cmsg_len = control->m_len;
    518 				cm->cmsg_level = SOL_SOCKET;
    519 				cm->cmsg_type = SCM_RIGHTS;
    520 			}
    521 		}
    522 #endif
    523 	} else
    524 		control = 0;
    525 #ifdef KTRACE
    526 	if (KTRPOINT(p, KTR_GENIO)) {
    527 		int iovlen = auio.uio_iovcnt * sizeof(struct iovec);
    528 
    529 		ktriov = malloc(iovlen, M_TEMP, M_WAITOK);
    530 		memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen);
    531 	}
    532 #endif
    533 	len = auio.uio_resid;
    534 	so = (struct socket *)fp->f_data;
    535 	error = (*so->so_send)(so, to, &auio, NULL, control, flags);
    536 	if (error) {
    537 		if (auio.uio_resid != len && (error == ERESTART ||
    538 		    error == EINTR || error == EWOULDBLOCK))
    539 			error = 0;
    540 		if (error == EPIPE)
    541 			psignal(p, SIGPIPE);
    542 	}
    543 	if (error == 0)
    544 		*retsize = len - auio.uio_resid;
    545 #ifdef KTRACE
    546 	if (ktriov != NULL) {
    547 		if (error == 0)
    548 			ktrgenio(p, s, UIO_WRITE, ktriov, *retsize, error);
    549 		free(ktriov, M_TEMP);
    550 	}
    551 #endif
    552  bad:
    553 	if (to)
    554 		m_freem(to);
    555  out:
    556 	FILE_UNUSE(fp, p);
    557 	return (error);
    558 }
    559 
    560 int
    561 sys_recvfrom(struct proc *p, void *v, register_t *retval)
    562 {
    563 	struct sys_recvfrom_args /* {
    564 		syscallarg(int)			s;
    565 		syscallarg(void *)		buf;
    566 		syscallarg(size_t)		len;
    567 		syscallarg(int)			flags;
    568 		syscallarg(struct sockaddr *)	from;
    569 		syscallarg(unsigned int *)	fromlenaddr;
    570 	} */ *uap = v;
    571 	struct msghdr	msg;
    572 	struct iovec	aiov;
    573 	int		error;
    574 
    575 	if (SCARG(uap, fromlenaddr)) {
    576 		error = copyin((caddr_t)SCARG(uap, fromlenaddr),
    577 			       (caddr_t)&msg.msg_namelen,
    578 			       sizeof(msg.msg_namelen));
    579 		if (error)
    580 			return (error);
    581 	} else
    582 		msg.msg_namelen = 0;
    583 	msg.msg_name = (caddr_t)SCARG(uap, from);
    584 	msg.msg_iov = &aiov;
    585 	msg.msg_iovlen = 1;
    586 	aiov.iov_base = SCARG(uap, buf);
    587 	aiov.iov_len = SCARG(uap, len);
    588 	msg.msg_control = 0;
    589 	msg.msg_flags = SCARG(uap, flags);
    590 	return (recvit(p, SCARG(uap, s), &msg,
    591 		       (caddr_t)SCARG(uap, fromlenaddr), retval));
    592 }
    593 
    594 int
    595 sys_recvmsg(struct proc *p, void *v, register_t *retval)
    596 {
    597 	struct sys_recvmsg_args /* {
    598 		syscallarg(int)			s;
    599 		syscallarg(struct msghdr *)	msg;
    600 		syscallarg(int)			flags;
    601 	} */ *uap = v;
    602 	struct msghdr	msg;
    603 	struct iovec	aiov[UIO_SMALLIOV], *uiov, *iov;
    604 	int		error;
    605 
    606 	error = copyin((caddr_t)SCARG(uap, msg), (caddr_t)&msg,
    607 		       sizeof(msg));
    608 	if (error)
    609 		return (error);
    610 	if ((unsigned int)msg.msg_iovlen > UIO_SMALLIOV) {
    611 		if ((unsigned int)msg.msg_iovlen > IOV_MAX)
    612 			return (EMSGSIZE);
    613 		iov = malloc(sizeof(struct iovec) * msg.msg_iovlen,
    614 		    M_IOV, M_WAITOK);
    615 	} else
    616 		iov = aiov;
    617 	if ((unsigned int)msg.msg_iovlen > 0) {
    618 		error = copyin((caddr_t)msg.msg_iov, (caddr_t)iov,
    619 		    (size_t)(msg.msg_iovlen * sizeof(struct iovec)));
    620 		if (error)
    621 			goto done;
    622 	}
    623 	uiov = msg.msg_iov;
    624 	msg.msg_iov = iov;
    625 #ifdef COMPAT_OLDSOCK
    626 	msg.msg_flags = SCARG(uap, flags) &~ MSG_COMPAT;
    627 #else
    628 	msg.msg_flags = SCARG(uap, flags);
    629 #endif
    630 	if ((error = recvit(p, SCARG(uap, s), &msg, (caddr_t)0, retval)) == 0) {
    631 		msg.msg_iov = uiov;
    632 		error = copyout((caddr_t)&msg, (caddr_t)SCARG(uap, msg),
    633 		    sizeof(msg));
    634 	}
    635 done:
    636 	if (iov != aiov)
    637 		free(iov, M_IOV);
    638 	return (error);
    639 }
    640 
    641 int
    642 recvit(struct proc *p, int s, struct msghdr *mp, caddr_t namelenp,
    643 	register_t *retsize)
    644 {
    645 	struct file	*fp;
    646 	struct uio	auio;
    647 	struct iovec	*iov;
    648 	int		i, len, error;
    649 	struct mbuf	*from, *control;
    650 	struct socket	*so;
    651 #ifdef KTRACE
    652 	struct iovec	*ktriov;
    653 #endif
    654 
    655 	from = 0;
    656 	control = 0;
    657 #ifdef KTRACE
    658 	ktriov = NULL;
    659 #endif
    660 
    661 	/* getsock() will use the descriptor for us */
    662 	if ((error = getsock(p->p_fd, s, &fp)) != 0)
    663 		return (error);
    664 	auio.uio_iov = mp->msg_iov;
    665 	auio.uio_iovcnt = mp->msg_iovlen;
    666 	auio.uio_segflg = UIO_USERSPACE;
    667 	auio.uio_rw = UIO_READ;
    668 	auio.uio_procp = p;
    669 	auio.uio_offset = 0;			/* XXX */
    670 	auio.uio_resid = 0;
    671 	iov = mp->msg_iov;
    672 	for (i = 0; i < mp->msg_iovlen; i++, iov++) {
    673 #if 0
    674 		/* cannot happen iov_len is unsigned */
    675 		if (iov->iov_len < 0) {
    676 			error = EINVAL;
    677 			goto out1;
    678 		}
    679 #endif
    680 		/*
    681 		 * Reads return ssize_t because -1 is returned on error.
    682 		 * Therefore we must restrict the length to SSIZE_MAX to
    683 		 * avoid garbage return values.
    684 		 */
    685 		auio.uio_resid += iov->iov_len;
    686 		if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
    687 			error = EINVAL;
    688 			goto out1;
    689 		}
    690 	}
    691 #ifdef KTRACE
    692 	if (KTRPOINT(p, KTR_GENIO)) {
    693 		int iovlen = auio.uio_iovcnt * sizeof(struct iovec);
    694 
    695 		ktriov = malloc(iovlen, M_TEMP, M_WAITOK);
    696 		memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen);
    697 	}
    698 #endif
    699 	len = auio.uio_resid;
    700 	so = (struct socket *)fp->f_data;
    701 	error = (*so->so_receive)(so, &from, &auio, NULL,
    702 			  mp->msg_control ? &control : NULL, &mp->msg_flags);
    703 	if (error) {
    704 		if (auio.uio_resid != len && (error == ERESTART ||
    705 		    error == EINTR || error == EWOULDBLOCK))
    706 			error = 0;
    707 	}
    708 #ifdef KTRACE
    709 	if (ktriov != NULL) {
    710 		if (error == 0)
    711 			ktrgenio(p, s, UIO_READ, ktriov,
    712 			    len - auio.uio_resid, error);
    713 		free(ktriov, M_TEMP);
    714 	}
    715 #endif
    716 	if (error)
    717 		goto out;
    718 	*retsize = len - auio.uio_resid;
    719 	if (mp->msg_name) {
    720 		len = mp->msg_namelen;
    721 		if (len <= 0 || from == 0)
    722 			len = 0;
    723 		else {
    724 #ifdef COMPAT_OLDSOCK
    725 			if (mp->msg_flags & MSG_COMPAT)
    726 				mtod(from, struct osockaddr *)->sa_family =
    727 				    mtod(from, struct sockaddr *)->sa_family;
    728 #endif
    729 			if (len > from->m_len)
    730 				len = from->m_len;
    731 			/* else if len < from->m_len ??? */
    732 			error = copyout(mtod(from, caddr_t),
    733 					(caddr_t)mp->msg_name, (unsigned)len);
    734 			if (error)
    735 				goto out;
    736 		}
    737 		mp->msg_namelen = len;
    738 		if (namelenp &&
    739 		    (error = copyout((caddr_t)&len, namelenp, sizeof(int)))) {
    740 #ifdef COMPAT_OLDSOCK
    741 			if (mp->msg_flags & MSG_COMPAT)
    742 				error = 0;	/* old recvfrom didn't check */
    743 			else
    744 #endif
    745 			goto out;
    746 		}
    747 	}
    748 	if (mp->msg_control) {
    749 #ifdef COMPAT_OLDSOCK
    750 		/*
    751 		 * We assume that old recvmsg calls won't receive access
    752 		 * rights and other control info, esp. as control info
    753 		 * is always optional and those options didn't exist in 4.3.
    754 		 * If we receive rights, trim the cmsghdr; anything else
    755 		 * is tossed.
    756 		 */
    757 		if (control && mp->msg_flags & MSG_COMPAT) {
    758 			if (mtod(control, struct cmsghdr *)->cmsg_level !=
    759 			    SOL_SOCKET ||
    760 			    mtod(control, struct cmsghdr *)->cmsg_type !=
    761 			    SCM_RIGHTS) {
    762 				mp->msg_controllen = 0;
    763 				goto out;
    764 			}
    765 			control->m_len -= sizeof(struct cmsghdr);
    766 			control->m_data += sizeof(struct cmsghdr);
    767 		}
    768 #endif
    769 		len = mp->msg_controllen;
    770 		if (len <= 0 || control == 0)
    771 			len = 0;
    772 		else {
    773 			struct mbuf *m = control;
    774 			caddr_t p = (caddr_t)mp->msg_control;
    775 
    776 			do {
    777 				i = m->m_len;
    778 				if (len < i) {
    779 					mp->msg_flags |= MSG_CTRUNC;
    780 					i = len;
    781 				}
    782 				error = copyout(mtod(m, caddr_t), p,
    783 				    (unsigned)i);
    784 				if (m->m_next)
    785 					i = ALIGN(i);
    786 				p += i;
    787 				len -= i;
    788 				if (error != 0 || len <= 0)
    789 					break;
    790 			} while ((m = m->m_next) != NULL);
    791 			len = p - (caddr_t)mp->msg_control;
    792 		}
    793 		mp->msg_controllen = len;
    794 	}
    795  out:
    796 	if (from)
    797 		m_freem(from);
    798 	if (control)
    799 		m_freem(control);
    800  out1:
    801 	FILE_UNUSE(fp, p);
    802 	return (error);
    803 }
    804 
    805 /* ARGSUSED */
    806 int
    807 sys_shutdown(struct proc *p, void *v, register_t *retval)
    808 {
    809 	struct sys_shutdown_args /* {
    810 		syscallarg(int)	s;
    811 		syscallarg(int)	how;
    812 	} */ *uap = v;
    813 	struct file	*fp;
    814 	int		error;
    815 
    816 	/* getsock() will use the descriptor for us */
    817 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    818 		return (error);
    819 	error = soshutdown((struct socket *)fp->f_data, SCARG(uap, how));
    820 	FILE_UNUSE(fp, p);
    821 	return (error);
    822 }
    823 
    824 /* ARGSUSED */
    825 int
    826 sys_setsockopt(struct proc *p, void *v, register_t *retval)
    827 {
    828 	struct sys_setsockopt_args /* {
    829 		syscallarg(int)			s;
    830 		syscallarg(int)			level;
    831 		syscallarg(int)			name;
    832 		syscallarg(const void *)	val;
    833 		syscallarg(unsigned int)	valsize;
    834 	} */ *uap = v;
    835 	struct file	*fp;
    836 	struct mbuf	*m;
    837 	int		error;
    838 
    839 	m = NULL;
    840 	/* getsock() will use the descriptor for us */
    841 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    842 		return (error);
    843 	if (SCARG(uap, valsize) > MLEN) {
    844 		error = EINVAL;
    845 		goto out;
    846 	}
    847 	if (SCARG(uap, val)) {
    848 		m = m_get(M_WAIT, MT_SOOPTS);
    849 		error = copyin(SCARG(uap, val), mtod(m, caddr_t),
    850 			       SCARG(uap, valsize));
    851 		if (error) {
    852 			(void) m_free(m);
    853 			goto out;
    854 		}
    855 		m->m_len = SCARG(uap, valsize);
    856 	}
    857 	error = sosetopt((struct socket *)fp->f_data, SCARG(uap, level),
    858 			 SCARG(uap, name), m);
    859  out:
    860 	FILE_UNUSE(fp, p);
    861 	return (error);
    862 }
    863 
    864 /* ARGSUSED */
    865 int
    866 sys_getsockopt(struct proc *p, void *v, register_t *retval)
    867 {
    868 	struct sys_getsockopt_args /* {
    869 		syscallarg(int)			s;
    870 		syscallarg(int)			level;
    871 		syscallarg(int)			name;
    872 		syscallarg(void *)		val;
    873 		syscallarg(unsigned int *)	avalsize;
    874 	} */ *uap = v;
    875 	struct file	*fp;
    876 	struct mbuf	*m, *m0;
    877 	unsigned int	op, i, valsize;
    878 	int		error;
    879 
    880 	m = NULL;
    881 	/* getsock() will use the descriptor for us */
    882 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    883 		return (error);
    884 	if (SCARG(uap, val)) {
    885 		error = copyin((caddr_t)SCARG(uap, avalsize),
    886 			       (caddr_t)&valsize, sizeof(valsize));
    887 		if (error)
    888 			goto out;
    889 	} else
    890 		valsize = 0;
    891 	if ((error = sogetopt((struct socket *)fp->f_data, SCARG(uap, level),
    892 	    SCARG(uap, name), &m)) == 0 && SCARG(uap, val) && valsize &&
    893 	    m != NULL) {
    894 		op = 0;
    895 		while (m && !error && op < valsize) {
    896 			i = min(m->m_len, (valsize - op));
    897 			error = copyout(mtod(m, caddr_t), SCARG(uap, val), i);
    898 			op += i;
    899 			SCARG(uap, val) = ((u_int8_t *)SCARG(uap, val)) + i;
    900 			m0 = m;
    901 			MFREE(m0, m);
    902 		}
    903 		valsize = op;
    904 		if (error == 0)
    905 			error = copyout(&valsize,
    906 					SCARG(uap, avalsize), sizeof(valsize));
    907 	}
    908 	if (m != NULL)
    909 		(void) m_free(m);
    910  out:
    911 	FILE_UNUSE(fp, p);
    912 	return (error);
    913 }
    914 
    915 /* ARGSUSED */
    916 int
    917 sys_pipe(struct proc *p, void *v, register_t *retval)
    918 {
    919 	struct filedesc	*fdp;
    920 	struct file	*rf, *wf;
    921 	struct socket	*rso, *wso;
    922 	int		fd, error;
    923 
    924 	fdp = p->p_fd;
    925 	if ((error = socreate(AF_LOCAL, &rso, SOCK_STREAM, 0)) != 0)
    926 		return (error);
    927 	if ((error = socreate(AF_LOCAL, &wso, SOCK_STREAM, 0)) != 0)
    928 		goto free1;
    929 	/* remember this socket pair implements a pipe */
    930 	wso->so_state |= SS_ISAPIPE;
    931 	rso->so_state |= SS_ISAPIPE;
    932 	/* falloc() will use the descriptor for us */
    933 	if ((error = falloc(p, &rf, &fd)) != 0)
    934 		goto free2;
    935 	retval[0] = fd;
    936 	rf->f_flag = FREAD;
    937 	rf->f_type = DTYPE_SOCKET;
    938 	rf->f_ops = &socketops;
    939 	rf->f_data = (caddr_t)rso;
    940 	if ((error = falloc(p, &wf, &fd)) != 0)
    941 		goto free3;
    942 	wf->f_flag = FWRITE;
    943 	wf->f_type = DTYPE_SOCKET;
    944 	wf->f_ops = &socketops;
    945 	wf->f_data = (caddr_t)wso;
    946 	retval[1] = fd;
    947 	if ((error = unp_connect2(wso, rso)) != 0)
    948 		goto free4;
    949 	FILE_SET_MATURE(rf);
    950 	FILE_SET_MATURE(wf);
    951 	FILE_UNUSE(rf, p);
    952 	FILE_UNUSE(wf, p);
    953 	return (0);
    954  free4:
    955 	FILE_UNUSE(wf, p);
    956 	ffree(wf);
    957 	fdremove(fdp, retval[1]);
    958  free3:
    959 	FILE_UNUSE(rf, p);
    960 	ffree(rf);
    961 	fdremove(fdp, retval[0]);
    962  free2:
    963 	(void)soclose(wso);
    964  free1:
    965 	(void)soclose(rso);
    966 	return (error);
    967 }
    968 
    969 /*
    970  * Get socket name.
    971  */
    972 /* ARGSUSED */
    973 int
    974 sys_getsockname(struct proc *p, void *v, register_t *retval)
    975 {
    976 	struct sys_getsockname_args /* {
    977 		syscallarg(int)			fdes;
    978 		syscallarg(struct sockaddr *)	asa;
    979 		syscallarg(unsigned int *)	alen;
    980 	} */ *uap = v;
    981 	struct file	*fp;
    982 	struct socket	*so;
    983 	struct mbuf	*m;
    984 	unsigned int	len;
    985 	int		error;
    986 
    987 	/* getsock() will use the descriptor for us */
    988 	if ((error = getsock(p->p_fd, SCARG(uap, fdes), &fp)) != 0)
    989 		return (error);
    990 	error = copyin((caddr_t)SCARG(uap, alen), (caddr_t)&len, sizeof(len));
    991 	if (error)
    992 		goto out;
    993 	so = (struct socket *)fp->f_data;
    994 	m = m_getclr(M_WAIT, MT_SONAME);
    995 	error = (*so->so_proto->pr_usrreq)(so, PRU_SOCKADDR, (struct mbuf *)0,
    996 	    m, (struct mbuf *)0, (struct proc *)0);
    997 	if (error)
    998 		goto bad;
    999 	if (len > m->m_len)
   1000 		len = m->m_len;
   1001 	error = copyout(mtod(m, caddr_t), (caddr_t)SCARG(uap, asa), len);
   1002 	if (error == 0)
   1003 		error = copyout((caddr_t)&len, (caddr_t)SCARG(uap, alen),
   1004 		    sizeof(len));
   1005  bad:
   1006 	m_freem(m);
   1007  out:
   1008 	FILE_UNUSE(fp, p);
   1009 	return (error);
   1010 }
   1011 
   1012 /*
   1013  * Get name of peer for connected socket.
   1014  */
   1015 /* ARGSUSED */
   1016 int
   1017 sys_getpeername(struct proc *p, void *v, register_t *retval)
   1018 {
   1019 	struct sys_getpeername_args /* {
   1020 		syscallarg(int)			fdes;
   1021 		syscallarg(struct sockaddr *)	asa;
   1022 		syscallarg(unsigned int *)	alen;
   1023 	} */ *uap = v;
   1024 	struct file	*fp;
   1025 	struct socket	*so;
   1026 	struct mbuf	*m;
   1027 	unsigned int	len;
   1028 	int		error;
   1029 
   1030 	/* getsock() will use the descriptor for us */
   1031 	if ((error = getsock(p->p_fd, SCARG(uap, fdes), &fp)) != 0)
   1032 		return (error);
   1033 	so = (struct socket *)fp->f_data;
   1034 	if ((so->so_state & (SS_ISCONNECTED|SS_ISCONFIRMING)) == 0) {
   1035 		error = ENOTCONN;
   1036 		goto out;
   1037 	}
   1038 	error = copyin((caddr_t)SCARG(uap, alen), (caddr_t)&len, sizeof(len));
   1039 	if (error)
   1040 		goto out;
   1041 	m = m_getclr(M_WAIT, MT_SONAME);
   1042 	error = (*so->so_proto->pr_usrreq)(so, PRU_PEERADDR, (struct mbuf *)0,
   1043 	    m, (struct mbuf *)0, (struct proc *)0);
   1044 	if (error)
   1045 		goto bad;
   1046 	if (len > m->m_len)
   1047 		len = m->m_len;
   1048 	error = copyout(mtod(m, caddr_t), (caddr_t)SCARG(uap, asa), len);
   1049 	if (error)
   1050 		goto bad;
   1051 	error = copyout((caddr_t)&len, (caddr_t)SCARG(uap, alen), sizeof(len));
   1052  bad:
   1053 	m_freem(m);
   1054  out:
   1055 	FILE_UNUSE(fp, p);
   1056 	return (error);
   1057 }
   1058 
   1059 /*
   1060  * XXX In a perfect world, we wouldn't pass around socket control
   1061  * XXX arguments in mbufs, and this could go away.
   1062  */
   1063 int
   1064 sockargs(struct mbuf **mp, const void *buf, int buflen, int type)
   1065 {
   1066 	struct sockaddr	*sa;
   1067 	struct mbuf	*m;
   1068 	int		error;
   1069 
   1070 	/*
   1071 	 * We can't allow socket names > UCHAR_MAX in length, since that
   1072 	 * will overflow sa_len.
   1073 	 */
   1074 	if (type == MT_SONAME && (u_int)buflen > UCHAR_MAX)
   1075 		return (EINVAL);
   1076 
   1077 	/* Allocate an mbuf to hold the arguments. */
   1078 	m = m_get(M_WAIT, type);
   1079 	if ((u_int)buflen > MLEN) {
   1080 		/*
   1081 		 * Won't fit into a regular mbuf, so we allocate just
   1082 		 * enough external storage to hold the argument.
   1083 		 */
   1084 		MEXTMALLOC(m, buflen, M_WAITOK);
   1085 	}
   1086 	m->m_len = buflen;
   1087 	error = copyin(buf, mtod(m, caddr_t), (u_int)buflen);
   1088 	if (error) {
   1089 		(void) m_free(m);
   1090 		return (error);
   1091 	}
   1092 	*mp = m;
   1093 	if (type == MT_SONAME) {
   1094 		sa = mtod(m, struct sockaddr *);
   1095 
   1096 #if defined(COMPAT_OLDSOCK) && BYTE_ORDER != BIG_ENDIAN
   1097 		if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
   1098 			sa->sa_family = sa->sa_len;
   1099 #endif
   1100 		sa->sa_len = buflen;
   1101 	}
   1102 	return (0);
   1103 }
   1104 
   1105 int
   1106 getsock(struct filedesc *fdp, int fdes, struct file **fpp)
   1107 {
   1108 	struct file	*fp;
   1109 
   1110 	if ((fp = fd_getfile(fdp, fdes)) == NULL)
   1111 		return (EBADF);
   1112 
   1113 	FILE_USE(fp);
   1114 
   1115 	if (fp->f_type != DTYPE_SOCKET) {
   1116 		FILE_UNUSE(fp, NULL);
   1117 		return (ENOTSOCK);
   1118 	}
   1119 	*fpp = fp;
   1120 	return (0);
   1121 }
   1122