Home | History | Annotate | Line # | Download | only in kern
uipc_syscalls.c revision 1.102.2.1
      1 /*	$NetBSD: uipc_syscalls.c,v 1.102.2.1 2006/10/20 21:33:55 ad 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. Neither the name of the University nor the names of its contributors
     16  *    may be used to endorse or promote products derived from this software
     17  *    without specific prior written permission.
     18  *
     19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     29  * SUCH DAMAGE.
     30  *
     31  *	@(#)uipc_syscalls.c	8.6 (Berkeley) 2/14/95
     32  */
     33 
     34 #include <sys/cdefs.h>
     35 __KERNEL_RCSID(0, "$NetBSD: uipc_syscalls.c,v 1.102.2.1 2006/10/20 21:33:55 ad Exp $");
     36 
     37 #include "opt_ktrace.h"
     38 #include "opt_pipe.h"
     39 
     40 #include <sys/param.h>
     41 #include <sys/systm.h>
     42 #include <sys/filedesc.h>
     43 #include <sys/proc.h>
     44 #include <sys/file.h>
     45 #include <sys/buf.h>
     46 #include <sys/malloc.h>
     47 #include <sys/mbuf.h>
     48 #include <sys/protosw.h>
     49 #include <sys/socket.h>
     50 #include <sys/socketvar.h>
     51 #include <sys/signalvar.h>
     52 #include <sys/un.h>
     53 #ifdef KTRACE
     54 #include <sys/ktrace.h>
     55 #endif
     56 #include <sys/event.h>
     57 
     58 #include <sys/mount.h>
     59 #include <sys/sa.h>
     60 #include <sys/syscallargs.h>
     61 
     62 #include <uvm/uvm_extern.h>
     63 
     64 static void adjust_rights(struct mbuf *m, int len, struct lwp *l);
     65 
     66 /*
     67  * System call interface to the socket abstraction.
     68  */
     69 extern const struct fileops socketops;
     70 
     71 int
     72 sys___socket30(struct lwp *l, void *v, register_t *retval)
     73 {
     74 	struct sys___socket30_args /* {
     75 		syscallarg(int)	domain;
     76 		syscallarg(int)	type;
     77 		syscallarg(int)	protocol;
     78 	} */ *uap = v;
     79 
     80 	struct filedesc	*fdp;
     81 	struct socket	*so;
     82 	struct file	*fp;
     83 	int		fd, error;
     84 
     85 	fdp = l->l_proc->p_fd;
     86 	/* falloc() will use the desciptor for us */
     87 	if ((error = falloc(l, &fp, &fd)) != 0)
     88 		return (error);
     89 	fp->f_flag = FREAD|FWRITE;
     90 	fp->f_type = DTYPE_SOCKET;
     91 	fp->f_ops = &socketops;
     92 	error = socreate(SCARG(uap, domain), &so, SCARG(uap, type),
     93 			 SCARG(uap, protocol), l);
     94 	if (error) {
     95 		FILE_UNUSE(fp, l);
     96 		fdremove(fdp, fd);
     97 		ffree(fp);
     98 	} else {
     99 		fp->f_data = (caddr_t)so;
    100 		FILE_SET_MATURE(fp);
    101 		FILE_UNUSE(fp, l);
    102 		*retval = fd;
    103 	}
    104 	return (error);
    105 }
    106 
    107 /* ARGSUSED */
    108 int
    109 sys_bind(struct lwp *l, void *v, register_t *retval)
    110 {
    111 	struct sys_bind_args /* {
    112 		syscallarg(int)				s;
    113 		syscallarg(const struct sockaddr *)	name;
    114 		syscallarg(unsigned int)		namelen;
    115 	} */ *uap = v;
    116 	struct proc	*p;
    117 	struct file	*fp;
    118 	struct mbuf	*nam;
    119 	int		error;
    120 
    121 	p = l->l_proc;
    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, l);
    129 		return (error);
    130 	}
    131 	MCLAIM(nam, ((struct socket *)fp->f_data)->so_mowner);
    132 	error = sobind((struct socket *)fp->f_data, nam, l);
    133 	m_freem(nam);
    134 	FILE_UNUSE(fp, l);
    135 	return (error);
    136 }
    137 
    138 /* ARGSUSED */
    139 int
    140 sys_listen(struct lwp *l, void *v, register_t *retval)
    141 {
    142 	struct sys_listen_args /* {
    143 		syscallarg(int)	s;
    144 		syscallarg(int)	backlog;
    145 	} */ *uap = v;
    146 	struct proc	*p;
    147 	struct file	*fp;
    148 	int		error;
    149 
    150 	p = l->l_proc;
    151 	/* getsock() will use the descriptor for us */
    152 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    153 		return (error);
    154 	error = solisten((struct socket *)fp->f_data, SCARG(uap, backlog));
    155 	FILE_UNUSE(fp, l);
    156 	return (error);
    157 }
    158 
    159 int
    160 sys_accept(struct lwp *l, void *v, register_t *retval)
    161 {
    162 	struct sys_accept_args /* {
    163 		syscallarg(int)			s;
    164 		syscallarg(struct sockaddr *)	name;
    165 		syscallarg(unsigned int *)	anamelen;
    166 	} */ *uap = v;
    167 	struct filedesc	*fdp;
    168 	struct file	*fp;
    169 	struct mbuf	*nam;
    170 	unsigned int	namelen;
    171 	int		error, s, fd;
    172 	struct socket	*so;
    173 	int		fflag;
    174 
    175 	fdp = l->l_proc->p_fd;
    176 	if (SCARG(uap, name) && (error = copyin(SCARG(uap, anamelen),
    177 	    &namelen, sizeof(namelen))))
    178 		return (error);
    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, l);
    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 	fflag = fp->f_flag;
    217 	/* falloc() will use the descriptor for us */
    218 	if ((error = falloc(l, &fp, &fd)) != 0) {
    219 		splx(s);
    220 		return (error);
    221 	}
    222 	*retval = fd;
    223 
    224 	/* connection has been removed from the listen queue */
    225 	KNOTE(&so->so_rcv.sb_sel.sel_klist, 0);
    226 
    227 	{ struct socket *aso = TAILQ_FIRST(&so->so_q);
    228 	  if (soqremque(aso, 1) == 0)
    229 		panic("accept");
    230 	  so = aso;
    231 	}
    232 	fp->f_type = DTYPE_SOCKET;
    233 	fp->f_flag = fflag;
    234 	fp->f_ops = &socketops;
    235 	fp->f_data = (caddr_t)so;
    236 	FILE_UNUSE(fp, l);
    237 	nam = m_get(M_WAIT, MT_SONAME);
    238 	if ((error = soaccept(so, nam)) == 0 && SCARG(uap, name)) {
    239 		if (namelen > nam->m_len)
    240 			namelen = nam->m_len;
    241 		/* SHOULD COPY OUT A CHAIN HERE */
    242 		if ((error = copyout(mtod(nam, caddr_t),
    243 		    (caddr_t)SCARG(uap, name), namelen)) != 0 ||
    244 		    (error = copyout((caddr_t)&namelen,
    245 		    (caddr_t)SCARG(uap, anamelen),
    246 		    sizeof(*SCARG(uap, anamelen)))) != 0) {
    247 			soclose(so);
    248 		}
    249 	}
    250 	/* if an error occurred, free the file descriptor */
    251 	if (error) {
    252 		fdremove(fdp, fd);
    253 		ffree(fp);
    254 	} else {
    255 		FILE_SET_MATURE(fp);
    256 	}
    257 	m_freem(nam);
    258 	splx(s);
    259 	return (error);
    260 }
    261 
    262 /* ARGSUSED */
    263 int
    264 sys_connect(struct lwp *l, void *v, register_t *retval)
    265 {
    266 	struct sys_connect_args /* {
    267 		syscallarg(int)				s;
    268 		syscallarg(const struct sockaddr *)	name;
    269 		syscallarg(unsigned int)		namelen;
    270 	} */ *uap = v;
    271 	struct proc	*p;
    272 	struct file	*fp;
    273 	struct socket	*so;
    274 	struct mbuf	*nam;
    275 	int		error, s;
    276 	int		interrupted = 0;
    277 
    278 	p = l->l_proc;
    279 	/* getsock() will use the descriptor for us */
    280 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    281 		return (error);
    282 	so = (struct socket *)fp->f_data;
    283 	if (so->so_state & SS_ISCONNECTING) {
    284 		error = EALREADY;
    285 		goto out;
    286 	}
    287 	error = sockargs(&nam, SCARG(uap, name), SCARG(uap, namelen),
    288 	    MT_SONAME);
    289 	if (error)
    290 		goto out;
    291 	MCLAIM(nam, so->so_mowner);
    292 	error = soconnect(so, nam, l);
    293 	if (error)
    294 		goto bad;
    295 	if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING)) {
    296 		m_freem(nam);
    297 		error = EINPROGRESS;
    298 		goto out;
    299 	}
    300 	s = splsoftnet();
    301 	while ((so->so_state & SS_ISCONNECTING) && so->so_error == 0) {
    302 		error = tsleep((caddr_t)&so->so_timeo, PSOCK | PCATCH,
    303 			       netcon, 0);
    304 		if (error) {
    305 			if (error == EINTR || error == ERESTART)
    306 				interrupted = 1;
    307 			break;
    308 		}
    309 	}
    310 	if (error == 0) {
    311 		error = so->so_error;
    312 		so->so_error = 0;
    313 	}
    314 	splx(s);
    315  bad:
    316 	if (!interrupted)
    317 		so->so_state &= ~SS_ISCONNECTING;
    318 	m_freem(nam);
    319 	if (error == ERESTART)
    320 		error = EINTR;
    321  out:
    322 	FILE_UNUSE(fp, l);
    323 	return (error);
    324 }
    325 
    326 int
    327 sys_socketpair(struct lwp *l, void *v, register_t *retval)
    328 {
    329 	struct sys_socketpair_args /* {
    330 		syscallarg(int)		domain;
    331 		syscallarg(int)		type;
    332 		syscallarg(int)		protocol;
    333 		syscallarg(int *)	rsv;
    334 	} */ *uap = v;
    335 	struct filedesc	*fdp;
    336 	struct file	*fp1, *fp2;
    337 	struct socket	*so1, *so2;
    338 	int		fd, error, sv[2];
    339 
    340 	fdp = l->l_proc->p_fd;
    341 	error = socreate(SCARG(uap, domain), &so1, SCARG(uap, type),
    342 	    SCARG(uap, protocol), l);
    343 	if (error)
    344 		return (error);
    345 	error = socreate(SCARG(uap, domain), &so2, SCARG(uap, type),
    346 	    SCARG(uap, protocol), l);
    347 	if (error)
    348 		goto free1;
    349 	/* falloc() will use the descriptor for us */
    350 	if ((error = falloc(l, &fp1, &fd)) != 0)
    351 		goto free2;
    352 	sv[0] = fd;
    353 	fp1->f_flag = FREAD|FWRITE;
    354 	fp1->f_type = DTYPE_SOCKET;
    355 	fp1->f_ops = &socketops;
    356 	fp1->f_data = (caddr_t)so1;
    357 	if ((error = falloc(l, &fp2, &fd)) != 0)
    358 		goto free3;
    359 	fp2->f_flag = FREAD|FWRITE;
    360 	fp2->f_type = DTYPE_SOCKET;
    361 	fp2->f_ops = &socketops;
    362 	fp2->f_data = (caddr_t)so2;
    363 	sv[1] = fd;
    364 	if ((error = soconnect2(so1, so2)) != 0)
    365 		goto free4;
    366 	if (SCARG(uap, type) == SOCK_DGRAM) {
    367 		/*
    368 		 * Datagram socket connection is asymmetric.
    369 		 */
    370 		 if ((error = soconnect2(so2, so1)) != 0)
    371 			goto free4;
    372 	}
    373 	error = copyout((caddr_t)sv, (caddr_t)SCARG(uap, rsv),
    374 	    2 * sizeof(int));
    375 	FILE_SET_MATURE(fp1);
    376 	FILE_SET_MATURE(fp2);
    377 	FILE_UNUSE(fp1, l);
    378 	FILE_UNUSE(fp2, l);
    379 	return (error);
    380  free4:
    381 	FILE_UNUSE(fp2, l);
    382 	ffree(fp2);
    383 	fdremove(fdp, sv[1]);
    384  free3:
    385 	FILE_UNUSE(fp1, l);
    386 	ffree(fp1);
    387 	fdremove(fdp, sv[0]);
    388  free2:
    389 	(void)soclose(so2);
    390  free1:
    391 	(void)soclose(so1);
    392 	return (error);
    393 }
    394 
    395 int
    396 sys_sendto(struct lwp *l, void *v, register_t *retval)
    397 {
    398 	struct sys_sendto_args /* {
    399 		syscallarg(int)				s;
    400 		syscallarg(const void *)		buf;
    401 		syscallarg(size_t)			len;
    402 		syscallarg(int)				flags;
    403 		syscallarg(const struct sockaddr *)	to;
    404 		syscallarg(unsigned int)		tolen;
    405 	} */ *uap = v;
    406 	struct msghdr	msg;
    407 	struct iovec	aiov;
    408 
    409 	msg.msg_name = __UNCONST(SCARG(uap, to)); /* XXXUNCONST kills const */
    410 	msg.msg_namelen = SCARG(uap, tolen);
    411 	msg.msg_iov = &aiov;
    412 	msg.msg_iovlen = 1;
    413 	msg.msg_control = 0;
    414 	msg.msg_flags = 0;
    415 	aiov.iov_base = __UNCONST(SCARG(uap, buf)); /* XXXUNCONST kills const */
    416 	aiov.iov_len = SCARG(uap, len);
    417 	return (sendit(l, SCARG(uap, s), &msg, SCARG(uap, flags), retval));
    418 }
    419 
    420 int
    421 sys_sendmsg(struct lwp *l, void *v, register_t *retval)
    422 {
    423 	struct sys_sendmsg_args /* {
    424 		syscallarg(int)				s;
    425 		syscallarg(const struct msghdr *)	msg;
    426 		syscallarg(int)				flags;
    427 	} */ *uap = v;
    428 	struct msghdr	msg;
    429 	struct iovec	aiov[UIO_SMALLIOV], *iov;
    430 	int		error;
    431 
    432 	error = copyin(SCARG(uap, msg), (caddr_t)&msg, sizeof(msg));
    433 	if (error)
    434 		return (error);
    435 	if ((unsigned int)msg.msg_iovlen > UIO_SMALLIOV) {
    436 		if ((unsigned int)msg.msg_iovlen > IOV_MAX)
    437 			return (EMSGSIZE);
    438 		iov = malloc(sizeof(struct iovec) * msg.msg_iovlen,
    439 		    M_IOV, M_WAITOK);
    440 	} else
    441 		iov = aiov;
    442 	if ((unsigned int)msg.msg_iovlen > 0) {
    443 		error = copyin((caddr_t)msg.msg_iov, (caddr_t)iov,
    444 		    (size_t)(msg.msg_iovlen * sizeof(struct iovec)));
    445 		if (error)
    446 			goto done;
    447 	}
    448 	msg.msg_iov = iov;
    449 	msg.msg_flags = 0;
    450 	error = sendit(l, SCARG(uap, s), &msg, SCARG(uap, flags), retval);
    451 done:
    452 	if (iov != aiov)
    453 		free(iov, M_IOV);
    454 	return (error);
    455 }
    456 
    457 int
    458 sendit(struct lwp *l, int s, struct msghdr *mp, int flags, register_t *retsize)
    459 {
    460 	struct proc	*p;
    461 	struct file	*fp;
    462 	struct uio	auio;
    463 	struct iovec	*iov;
    464 	int		i, len, error;
    465 	struct mbuf	*to, *control;
    466 	struct socket	*so;
    467 #ifdef KTRACE
    468 	struct iovec	*ktriov;
    469 #endif
    470 
    471 #ifdef KTRACE
    472 	ktriov = NULL;
    473 #endif
    474 	p = l->l_proc;
    475 	/* getsock() will use the descriptor for us */
    476 	if ((error = getsock(p->p_fd, s, &fp)) != 0)
    477 		return (error);
    478 	so = (struct socket *)fp->f_data;
    479 	auio.uio_iov = mp->msg_iov;
    480 	auio.uio_iovcnt = mp->msg_iovlen;
    481 	auio.uio_rw = UIO_WRITE;
    482 	auio.uio_offset = 0;			/* XXX */
    483 	auio.uio_resid = 0;
    484 	KASSERT(l == curlwp);
    485 	auio.uio_vmspace = l->l_proc->p_vmspace;
    486 	iov = mp->msg_iov;
    487 	for (i = 0; i < mp->msg_iovlen; i++, iov++) {
    488 #if 0
    489 		/* cannot happen; iov_len is unsigned */
    490 		if (iov->iov_len < 0) {
    491 			error = EINVAL;
    492 			goto out;
    493 		}
    494 #endif
    495 		/*
    496 		 * Writes return ssize_t because -1 is returned on error.
    497 		 * Therefore, we must restrict the length to SSIZE_MAX to
    498 		 * avoid garbage return values.
    499 		 */
    500 		auio.uio_resid += iov->iov_len;
    501 		if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
    502 			error = EINVAL;
    503 			goto out;
    504 		}
    505 	}
    506 	if (mp->msg_name) {
    507 		error = sockargs(&to, mp->msg_name, mp->msg_namelen,
    508 				 MT_SONAME);
    509 		if (error)
    510 			goto out;
    511 		MCLAIM(to, so->so_mowner);
    512 	} else
    513 		to = 0;
    514 	if (mp->msg_control) {
    515 		if (mp->msg_controllen < sizeof(struct cmsghdr)) {
    516 			error = EINVAL;
    517 			goto bad;
    518 		}
    519 		error = sockargs(&control, mp->msg_control,
    520 				 mp->msg_controllen, MT_CONTROL);
    521 		if (error)
    522 			goto bad;
    523 		MCLAIM(control, so->so_mowner);
    524 	} else
    525 		control = 0;
    526 #ifdef KTRACE
    527 	if (KTRPOINT(p, KTR_GENIO)) {
    528 		int iovlen = auio.uio_iovcnt * sizeof(struct iovec);
    529 
    530 		ktriov = malloc(iovlen, M_TEMP, M_WAITOK);
    531 		memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen);
    532 	}
    533 #endif
    534 	len = auio.uio_resid;
    535 	error = (*so->so_send)(so, to, &auio, NULL, control, flags, l);
    536 	if (error) {
    537 		if (auio.uio_resid != len && (error == ERESTART ||
    538 		    error == EINTR || error == EWOULDBLOCK))
    539 			error = 0;
    540 		if (error == EPIPE && (flags & MSG_NOSIGNAL) == 0) {
    541 			rw_enter(&proclist_lock, RW_READER);
    542 			psignal(p, SIGPIPE);
    543 			rw_exit(&proclist_lock);
    544 		}
    545 	}
    546 	if (error == 0)
    547 		*retsize = len - auio.uio_resid;
    548 #ifdef KTRACE
    549 	if (ktriov != NULL) {
    550 		if (error == 0)
    551 			ktrgenio(l, s, UIO_WRITE, ktriov, *retsize, error);
    552 		free(ktriov, M_TEMP);
    553 	}
    554 #endif
    555  bad:
    556 	if (to)
    557 		m_freem(to);
    558  out:
    559 	FILE_UNUSE(fp, l);
    560 	return (error);
    561 }
    562 
    563 int
    564 sys_recvfrom(struct lwp *l, void *v, register_t *retval)
    565 {
    566 	struct sys_recvfrom_args /* {
    567 		syscallarg(int)			s;
    568 		syscallarg(void *)		buf;
    569 		syscallarg(size_t)		len;
    570 		syscallarg(int)			flags;
    571 		syscallarg(struct sockaddr *)	from;
    572 		syscallarg(unsigned int *)	fromlenaddr;
    573 	} */ *uap = v;
    574 	struct msghdr	msg;
    575 	struct iovec	aiov;
    576 	int		error;
    577 
    578 	if (SCARG(uap, fromlenaddr)) {
    579 		error = copyin((caddr_t)SCARG(uap, fromlenaddr),
    580 			       (caddr_t)&msg.msg_namelen,
    581 			       sizeof(msg.msg_namelen));
    582 		if (error)
    583 			return (error);
    584 	} else
    585 		msg.msg_namelen = 0;
    586 	msg.msg_name = (caddr_t)SCARG(uap, from);
    587 	msg.msg_iov = &aiov;
    588 	msg.msg_iovlen = 1;
    589 	aiov.iov_base = SCARG(uap, buf);
    590 	aiov.iov_len = SCARG(uap, len);
    591 	msg.msg_control = 0;
    592 	msg.msg_flags = SCARG(uap, flags);
    593 	return (recvit(l, SCARG(uap, s), &msg,
    594 		       (caddr_t)SCARG(uap, fromlenaddr), retval));
    595 }
    596 
    597 int
    598 sys_recvmsg(struct lwp *l, void *v, register_t *retval)
    599 {
    600 	struct sys_recvmsg_args /* {
    601 		syscallarg(int)			s;
    602 		syscallarg(struct msghdr *)	msg;
    603 		syscallarg(int)			flags;
    604 	} */ *uap = v;
    605 	struct iovec	aiov[UIO_SMALLIOV], *uiov, *iov;
    606 	struct msghdr	msg;
    607 	int		error;
    608 
    609 	error = copyin((caddr_t)SCARG(uap, msg), (caddr_t)&msg,
    610 		       sizeof(msg));
    611 	if (error)
    612 		return (error);
    613 	if ((unsigned int)msg.msg_iovlen > UIO_SMALLIOV) {
    614 		if ((unsigned int)msg.msg_iovlen > IOV_MAX)
    615 			return (EMSGSIZE);
    616 		iov = malloc(sizeof(struct iovec) * msg.msg_iovlen,
    617 		    M_IOV, M_WAITOK);
    618 	} else
    619 		iov = aiov;
    620 	if ((unsigned int)msg.msg_iovlen > 0) {
    621 		error = copyin((caddr_t)msg.msg_iov, (caddr_t)iov,
    622 		    (size_t)(msg.msg_iovlen * sizeof(struct iovec)));
    623 		if (error)
    624 			goto done;
    625 	}
    626 	uiov = msg.msg_iov;
    627 	msg.msg_iov = iov;
    628 	msg.msg_flags = SCARG(uap, flags);
    629 	if ((error = recvit(l, SCARG(uap, s), &msg, (caddr_t)0, retval)) == 0) {
    630 		msg.msg_iov = uiov;
    631 		error = copyout((caddr_t)&msg, (caddr_t)SCARG(uap, msg),
    632 		    sizeof(msg));
    633 	}
    634 done:
    635 	if (iov != aiov)
    636 		free(iov, M_IOV);
    637 	return (error);
    638 }
    639 
    640 /*
    641  * Adjust for a truncated SCM_RIGHTS control message.  This means
    642  *  closing any file descriptors that aren't entirely present in the
    643  *  returned buffer.  m is the mbuf holding the (already externalized)
    644  *  SCM_RIGHTS message; len is the length it is being truncated to.  p
    645  *  is the affected process.
    646  */
    647 static void
    648 adjust_rights(struct mbuf *m, int len, struct lwp *l)
    649 {
    650 	int nfd;
    651 	int i;
    652 	int nok;
    653 	int *fdv;
    654 
    655 	nfd = m->m_len < CMSG_SPACE(sizeof(int)) ? 0
    656 	    : (m->m_len - CMSG_SPACE(sizeof(int))) / sizeof(int) + 1;
    657 	nok = (len < CMSG_LEN(0)) ? 0 : ((len - CMSG_LEN(0)) / sizeof(int));
    658 	fdv = (int *) CMSG_DATA(mtod(m,struct cmsghdr *));
    659 	for (i = nok; i < nfd; i++)
    660 		fdrelease(l,fdv[i]);
    661 }
    662 
    663 int
    664 recvit(struct lwp *l, int s, struct msghdr *mp, caddr_t namelenp,
    665 	register_t *retsize)
    666 {
    667 	struct proc	*p;
    668 	struct file	*fp;
    669 	struct uio	auio;
    670 	struct iovec	*iov;
    671 	int		i, len, error;
    672 	struct mbuf	*from, *control;
    673 	struct socket	*so;
    674 #ifdef KTRACE
    675 	struct iovec	*ktriov;
    676 #endif
    677 
    678 	p = l->l_proc;
    679 	from = 0;
    680 	control = 0;
    681 #ifdef KTRACE
    682 	ktriov = NULL;
    683 #endif
    684 
    685 	/* getsock() will use the descriptor for us */
    686 	if ((error = getsock(p->p_fd, s, &fp)) != 0)
    687 		return (error);
    688 	so = (struct socket *)fp->f_data;
    689 	auio.uio_iov = mp->msg_iov;
    690 	auio.uio_iovcnt = mp->msg_iovlen;
    691 	auio.uio_rw = UIO_READ;
    692 	auio.uio_offset = 0;			/* XXX */
    693 	auio.uio_resid = 0;
    694 	KASSERT(l == curlwp);
    695 	auio.uio_vmspace = l->l_proc->p_vmspace;
    696 	iov = mp->msg_iov;
    697 	for (i = 0; i < mp->msg_iovlen; i++, iov++) {
    698 #if 0
    699 		/* cannot happen iov_len is unsigned */
    700 		if (iov->iov_len < 0) {
    701 			error = EINVAL;
    702 			goto out1;
    703 		}
    704 #endif
    705 		/*
    706 		 * Reads return ssize_t because -1 is returned on error.
    707 		 * Therefore we must restrict the length to SSIZE_MAX to
    708 		 * avoid garbage return values.
    709 		 */
    710 		auio.uio_resid += iov->iov_len;
    711 		if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
    712 			error = EINVAL;
    713 			goto out1;
    714 		}
    715 	}
    716 #ifdef KTRACE
    717 	if (KTRPOINT(p, KTR_GENIO)) {
    718 		int iovlen = auio.uio_iovcnt * sizeof(struct iovec);
    719 
    720 		ktriov = malloc(iovlen, M_TEMP, M_WAITOK);
    721 		memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen);
    722 	}
    723 #endif
    724 	len = auio.uio_resid;
    725 	error = (*so->so_receive)(so, &from, &auio, NULL,
    726 			  mp->msg_control ? &control : NULL, &mp->msg_flags);
    727 	if (error) {
    728 		if (auio.uio_resid != len && (error == ERESTART ||
    729 		    error == EINTR || error == EWOULDBLOCK))
    730 			error = 0;
    731 	}
    732 #ifdef KTRACE
    733 	if (ktriov != NULL) {
    734 		if (error == 0)
    735 			ktrgenio(l, s, UIO_READ, ktriov,
    736 			    len - auio.uio_resid, error);
    737 		free(ktriov, M_TEMP);
    738 	}
    739 #endif
    740 	if (error)
    741 		goto out;
    742 	*retsize = len - auio.uio_resid;
    743 	if (mp->msg_name) {
    744 		len = mp->msg_namelen;
    745 		if (len <= 0 || from == 0)
    746 			len = 0;
    747 		else {
    748 			if (len > from->m_len)
    749 				len = from->m_len;
    750 			/* else if len < from->m_len ??? */
    751 			error = copyout(mtod(from, caddr_t),
    752 					(caddr_t)mp->msg_name, (unsigned)len);
    753 			if (error)
    754 				goto out;
    755 		}
    756 		mp->msg_namelen = len;
    757 		if (namelenp &&
    758 		    (error = copyout((caddr_t)&len, namelenp, sizeof(int))))
    759 			goto out;
    760 	}
    761 	if (mp->msg_control) {
    762 		len = mp->msg_controllen;
    763 		if (len <= 0 || control == 0)
    764 			len = 0;
    765 		else {
    766 			struct mbuf *m = control;
    767 			caddr_t q = (caddr_t)mp->msg_control;
    768 
    769 			do {
    770 				i = m->m_len;
    771 				if (len < i) {
    772 					mp->msg_flags |= MSG_CTRUNC;
    773 					i = len;
    774 					if (mtod(m, struct cmsghdr *)->
    775 					    cmsg_type == SCM_RIGHTS)
    776 						adjust_rights(m, len, l);
    777 				}
    778 				error = copyout(mtod(m, caddr_t), q,
    779 				    (unsigned)i);
    780 				m = m->m_next;
    781 				if (m)
    782 					i = ALIGN(i);
    783 				q += i;
    784 				len -= i;
    785 				if (error != 0 || len <= 0)
    786 					break;
    787 			} while (m != NULL);
    788 			while (m) {
    789 				if (mtod(m, struct cmsghdr *)->
    790 				    cmsg_type == SCM_RIGHTS)
    791 					adjust_rights(m, 0, l);
    792 				m = m->m_next;
    793 			}
    794 			len = q - (caddr_t)mp->msg_control;
    795 		}
    796 		mp->msg_controllen = len;
    797 	}
    798  out:
    799 	if (from)
    800 		m_freem(from);
    801 	if (control)
    802 		m_freem(control);
    803  out1:
    804 	FILE_UNUSE(fp, l);
    805 	return (error);
    806 }
    807 
    808 /* ARGSUSED */
    809 int
    810 sys_shutdown(struct lwp *l, void *v, register_t *retval)
    811 {
    812 	struct sys_shutdown_args /* {
    813 		syscallarg(int)	s;
    814 		syscallarg(int)	how;
    815 	} */ *uap = v;
    816 	struct proc	*p;
    817 	struct file	*fp;
    818 	int		error;
    819 
    820 	p = l->l_proc;
    821 	/* getsock() will use the descriptor for us */
    822 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    823 		return (error);
    824 	error = soshutdown((struct socket *)fp->f_data, SCARG(uap, how));
    825 	FILE_UNUSE(fp, l);
    826 	return (error);
    827 }
    828 
    829 /* ARGSUSED */
    830 int
    831 sys_setsockopt(struct lwp *l, void *v, register_t *retval)
    832 {
    833 	struct sys_setsockopt_args /* {
    834 		syscallarg(int)			s;
    835 		syscallarg(int)			level;
    836 		syscallarg(int)			name;
    837 		syscallarg(const void *)	val;
    838 		syscallarg(unsigned int)	valsize;
    839 	} */ *uap = v;
    840 	struct proc	*p;
    841 	struct file	*fp;
    842 	struct mbuf	*m;
    843 	struct socket	*so;
    844 	int		error;
    845 	unsigned int	len;
    846 
    847 	p = l->l_proc;
    848 	m = NULL;
    849 	/* getsock() will use the descriptor for us */
    850 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
    851 		return (error);
    852 	so = (struct socket *)fp->f_data;
    853 	len = SCARG(uap, valsize);
    854 	if (len > MCLBYTES) {
    855 		error = EINVAL;
    856 		goto out;
    857 	}
    858 	if (SCARG(uap, val)) {
    859 		m = m_get(M_WAIT, MT_SOOPTS);
    860 		MCLAIM(m, so->so_mowner);
    861 		if (len > MLEN)
    862 			m_clget(m, M_WAIT);
    863 		error = copyin(SCARG(uap, val), mtod(m, caddr_t), len);
    864 		if (error) {
    865 			(void) m_free(m);
    866 			goto out;
    867 		}
    868 		m->m_len = SCARG(uap, valsize);
    869 	}
    870 	error = sosetopt(so, SCARG(uap, level), SCARG(uap, name), m);
    871  out:
    872 	FILE_UNUSE(fp, l);
    873 	return (error);
    874 }
    875 
    876 /* ARGSUSED */
    877 int
    878 sys_getsockopt(struct lwp *l, void *v, register_t *retval)
    879 {
    880 	struct sys_getsockopt_args /* {
    881 		syscallarg(int)			s;
    882 		syscallarg(int)			level;
    883 		syscallarg(int)			name;
    884 		syscallarg(void *)		val;
    885 		syscallarg(unsigned int *)	avalsize;
    886 	} */ *uap = v;
    887 	struct file	*fp;
    888 	struct mbuf	*m;
    889 	unsigned int	op, i, valsize;
    890 	int		error;
    891 
    892 	m = NULL;
    893 	/* getsock() will use the descriptor for us */
    894 	if ((error = getsock(l->l_proc->p_fd, SCARG(uap, s), &fp)) != 0)
    895 		return (error);
    896 	if (SCARG(uap, val)) {
    897 		error = copyin((caddr_t)SCARG(uap, avalsize),
    898 			       (caddr_t)&valsize, sizeof(valsize));
    899 		if (error)
    900 			goto out;
    901 	} else
    902 		valsize = 0;
    903 	if ((error = sogetopt((struct socket *)fp->f_data, SCARG(uap, level),
    904 	    SCARG(uap, name), &m)) == 0 && SCARG(uap, val) && valsize &&
    905 	    m != NULL) {
    906 		op = 0;
    907 		while (m && !error && op < valsize) {
    908 			i = min(m->m_len, (valsize - op));
    909 			error = copyout(mtod(m, caddr_t), SCARG(uap, val), i);
    910 			op += i;
    911 			SCARG(uap, val) = ((uint8_t *)SCARG(uap, val)) + i;
    912 			m = m_free(m);
    913 		}
    914 		valsize = op;
    915 		if (error == 0)
    916 			error = copyout(&valsize,
    917 					SCARG(uap, avalsize), sizeof(valsize));
    918 	}
    919 	if (m != NULL)
    920 		(void) m_freem(m);
    921  out:
    922 	FILE_UNUSE(fp, l);
    923 	return (error);
    924 }
    925 
    926 #ifdef PIPE_SOCKETPAIR
    927 /* ARGSUSED */
    928 int
    929 sys_pipe(struct lwp *l, void *v, register_t *retval)
    930 {
    931 	struct filedesc	*fdp;
    932 	struct file	*rf, *wf;
    933 	struct socket	*rso, *wso;
    934 	int		fd, error;
    935 
    936 	fdp = l->l_proc->p_fd;
    937 	if ((error = socreate(AF_LOCAL, &rso, SOCK_STREAM, 0, l)) != 0)
    938 		return (error);
    939 	if ((error = socreate(AF_LOCAL, &wso, SOCK_STREAM, 0, l)) != 0)
    940 		goto free1;
    941 	/* remember this socket pair implements a pipe */
    942 	wso->so_state |= SS_ISAPIPE;
    943 	rso->so_state |= SS_ISAPIPE;
    944 	/* falloc() will use the descriptor for us */
    945 	if ((error = falloc(l, &rf, &fd)) != 0)
    946 		goto free2;
    947 	retval[0] = fd;
    948 	rf->f_flag = FREAD;
    949 	rf->f_type = DTYPE_SOCKET;
    950 	rf->f_ops = &socketops;
    951 	rf->f_data = (caddr_t)rso;
    952 	if ((error = falloc(l, &wf, &fd)) != 0)
    953 		goto free3;
    954 	wf->f_flag = FWRITE;
    955 	wf->f_type = DTYPE_SOCKET;
    956 	wf->f_ops = &socketops;
    957 	wf->f_data = (caddr_t)wso;
    958 	retval[1] = fd;
    959 	if ((error = unp_connect2(wso, rso, PRU_CONNECT2)) != 0)
    960 		goto free4;
    961 	FILE_SET_MATURE(rf);
    962 	FILE_SET_MATURE(wf);
    963 	FILE_UNUSE(rf, l);
    964 	FILE_UNUSE(wf, l);
    965 	return (0);
    966  free4:
    967 	FILE_UNUSE(wf, l);
    968 	ffree(wf);
    969 	fdremove(fdp, retval[1]);
    970  free3:
    971 	FILE_UNUSE(rf, l);
    972 	ffree(rf);
    973 	fdremove(fdp, retval[0]);
    974  free2:
    975 	(void)soclose(wso);
    976  free1:
    977 	(void)soclose(rso);
    978 	return (error);
    979 }
    980 #endif /* PIPE_SOCKETPAIR */
    981 
    982 /*
    983  * Get socket name.
    984  */
    985 /* ARGSUSED */
    986 int
    987 sys_getsockname(struct lwp *l, void *v, register_t *retval)
    988 {
    989 	struct sys_getsockname_args /* {
    990 		syscallarg(int)			fdes;
    991 		syscallarg(struct sockaddr *)	asa;
    992 		syscallarg(unsigned int *)	alen;
    993 	} */ *uap = v;
    994 	struct proc	*p;
    995 	struct file	*fp;
    996 	struct socket	*so;
    997 	struct mbuf	*m;
    998 	unsigned int	len;
    999 	int		error;
   1000 
   1001 	p = l->l_proc;
   1002 	/* getsock() will use the descriptor for us */
   1003 	if ((error = getsock(p->p_fd, SCARG(uap, fdes), &fp)) != 0)
   1004 		return (error);
   1005 	error = copyin((caddr_t)SCARG(uap, alen), (caddr_t)&len, sizeof(len));
   1006 	if (error)
   1007 		goto out;
   1008 	so = (struct socket *)fp->f_data;
   1009 	m = m_getclr(M_WAIT, MT_SONAME);
   1010 	MCLAIM(m, so->so_mowner);
   1011 	error = (*so->so_proto->pr_usrreq)(so, PRU_SOCKADDR, (struct mbuf *)0,
   1012 	    m, (struct mbuf *)0, (struct lwp *)0);
   1013 	if (error)
   1014 		goto bad;
   1015 	if (len > m->m_len)
   1016 		len = m->m_len;
   1017 	error = copyout(mtod(m, caddr_t), (caddr_t)SCARG(uap, asa), len);
   1018 	if (error == 0)
   1019 		error = copyout((caddr_t)&len, (caddr_t)SCARG(uap, alen),
   1020 		    sizeof(len));
   1021  bad:
   1022 	m_freem(m);
   1023  out:
   1024 	FILE_UNUSE(fp, l);
   1025 	return (error);
   1026 }
   1027 
   1028 /*
   1029  * Get name of peer for connected socket.
   1030  */
   1031 /* ARGSUSED */
   1032 int
   1033 sys_getpeername(struct lwp *l, void *v, register_t *retval)
   1034 {
   1035 	struct sys_getpeername_args /* {
   1036 		syscallarg(int)			fdes;
   1037 		syscallarg(struct sockaddr *)	asa;
   1038 		syscallarg(unsigned int *)	alen;
   1039 	} */ *uap = v;
   1040 	struct proc	*p;
   1041 	struct file	*fp;
   1042 	struct socket	*so;
   1043 	struct mbuf	*m;
   1044 	unsigned int	len;
   1045 	int		error;
   1046 
   1047 	p = l->l_proc;
   1048 	/* getsock() will use the descriptor for us */
   1049 	if ((error = getsock(p->p_fd, SCARG(uap, fdes), &fp)) != 0)
   1050 		return (error);
   1051 	so = (struct socket *)fp->f_data;
   1052 	if ((so->so_state & (SS_ISCONNECTED|SS_ISCONFIRMING)) == 0) {
   1053 		error = ENOTCONN;
   1054 		goto out;
   1055 	}
   1056 	error = copyin((caddr_t)SCARG(uap, alen), (caddr_t)&len, sizeof(len));
   1057 	if (error)
   1058 		goto out;
   1059 	m = m_getclr(M_WAIT, MT_SONAME);
   1060 	MCLAIM(m, so->so_mowner);
   1061 	error = (*so->so_proto->pr_usrreq)(so, PRU_PEERADDR, (struct mbuf *)0,
   1062 	    m, (struct mbuf *)0, (struct lwp *)0);
   1063 	if (error)
   1064 		goto bad;
   1065 	if (len > m->m_len)
   1066 		len = m->m_len;
   1067 	error = copyout(mtod(m, caddr_t), (caddr_t)SCARG(uap, asa), len);
   1068 	if (error)
   1069 		goto bad;
   1070 	error = copyout((caddr_t)&len, (caddr_t)SCARG(uap, alen), sizeof(len));
   1071  bad:
   1072 	m_freem(m);
   1073  out:
   1074 	FILE_UNUSE(fp, l);
   1075 	return (error);
   1076 }
   1077 
   1078 /*
   1079  * XXX In a perfect world, we wouldn't pass around socket control
   1080  * XXX arguments in mbufs, and this could go away.
   1081  */
   1082 int
   1083 sockargs(struct mbuf **mp, const void *bf, size_t buflen, int type)
   1084 {
   1085 	struct sockaddr	*sa;
   1086 	struct mbuf	*m;
   1087 	int		error;
   1088 
   1089 	/*
   1090 	 * We can't allow socket names > UCHAR_MAX in length, since that
   1091 	 * will overflow sa_len.  Control data more than a page size in
   1092 	 * length is just too much.
   1093 	 */
   1094 	if (buflen > (type == MT_SONAME ? UCHAR_MAX : PAGE_SIZE))
   1095 		return (EINVAL);
   1096 
   1097 	/* Allocate an mbuf to hold the arguments. */
   1098 	m = m_get(M_WAIT, type);
   1099 	/* can't claim.  don't who to assign it to. */
   1100 	if (buflen > MLEN) {
   1101 		/*
   1102 		 * Won't fit into a regular mbuf, so we allocate just
   1103 		 * enough external storage to hold the argument.
   1104 		 */
   1105 		MEXTMALLOC(m, buflen, M_WAITOK);
   1106 	}
   1107 	m->m_len = buflen;
   1108 	error = copyin(bf, mtod(m, caddr_t), buflen);
   1109 	if (error) {
   1110 		(void) m_free(m);
   1111 		return (error);
   1112 	}
   1113 	*mp = m;
   1114 	if (type == MT_SONAME) {
   1115 		sa = mtod(m, struct sockaddr *);
   1116 #if BYTE_ORDER != BIG_ENDIAN
   1117 		/*
   1118 		 * 4.3BSD compat thing - need to stay, since bind(2),
   1119 		 * connect(2), sendto(2) were not versioned for COMPAT_43.
   1120 		 */
   1121 		if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
   1122 			sa->sa_family = sa->sa_len;
   1123 #endif
   1124 		sa->sa_len = buflen;
   1125 	}
   1126 	return (0);
   1127 }
   1128 
   1129 int
   1130 getsock(struct filedesc *fdp, int fdes, struct file **fpp)
   1131 {
   1132 	struct file	*fp;
   1133 
   1134 	if ((fp = fd_getfile(fdp, fdes)) == NULL)
   1135 		return (EBADF);
   1136 
   1137 	FILE_USE(fp);
   1138 
   1139 	if (fp->f_type != DTYPE_SOCKET) {
   1140 		FILE_UNUSE(fp, NULL);
   1141 		return (ENOTSOCK);
   1142 	}
   1143 	*fpp = fp;
   1144 	return (0);
   1145 }
   1146