xref: /netbsd-src/sys/kern/uipc_syscalls.c (revision 3b01aba77a7a698587faaae455bbfe740923c1f5)
1 /*	$NetBSD: uipc_syscalls.c,v 1.65 2001/07/17 11:49:41 jdolecek Exp $	*/
2 
3 /*
4  * Copyright (c) 1982, 1986, 1989, 1990, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. All advertising materials mentioning features or use of this software
16  *    must display the following acknowledgement:
17  *	This product includes software developed by the University of
18  *	California, Berkeley and its contributors.
19  * 4. Neither the name of the University nor the names of its contributors
20  *    may be used to endorse or promote products derived from this software
21  *    without specific prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33  * SUCH DAMAGE.
34  *
35  *	@(#)uipc_syscalls.c	8.6 (Berkeley) 2/14/95
36  */
37 
38 #include "opt_ktrace.h"
39 #include "opt_new_pipe.h"
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/filedesc.h>
44 #include <sys/proc.h>
45 #include <sys/file.h>
46 #include <sys/buf.h>
47 #include <sys/malloc.h>
48 #include <sys/mbuf.h>
49 #include <sys/protosw.h>
50 #include <sys/socket.h>
51 #include <sys/socketvar.h>
52 #include <sys/signalvar.h>
53 #include <sys/un.h>
54 #ifdef KTRACE
55 #include <sys/ktrace.h>
56 #endif
57 
58 #include <sys/mount.h>
59 #include <sys/syscallargs.h>
60 
61 #include <uvm/uvm_extern.h>
62 
63 /*
64  * System call interface to the socket abstraction.
65  */
66 extern	struct fileops socketops;
67 
68 int
69 sys_socket(struct proc *p, void *v, register_t *retval)
70 {
71 	struct sys_socket_args /* {
72 		syscallarg(int)	domain;
73 		syscallarg(int)	type;
74 		syscallarg(int)	protocol;
75 	} */ *uap = v;
76 	struct filedesc	*fdp;
77 	struct socket	*so;
78 	struct file	*fp;
79 	int		fd, error;
80 
81 	fdp = p->p_fd;
82 	/* falloc() will use the desciptor for us */
83 	if ((error = falloc(p, &fp, &fd)) != 0)
84 		return (error);
85 	fp->f_flag = FREAD|FWRITE;
86 	fp->f_type = DTYPE_SOCKET;
87 	fp->f_ops = &socketops;
88 	error = socreate(SCARG(uap, domain), &so, SCARG(uap, type),
89 			 SCARG(uap, protocol));
90 	if (error) {
91 		FILE_UNUSE(fp, p);
92 		fdremove(fdp, fd);
93 		ffree(fp);
94 	} else {
95 		fp->f_data = (caddr_t)so;
96 		FILE_SET_MATURE(fp);
97 		FILE_UNUSE(fp, p);
98 		*retval = fd;
99 	}
100 	return (error);
101 }
102 
103 /* ARGSUSED */
104 int
105 sys_bind(struct proc *p, void *v, register_t *retval)
106 {
107 	struct sys_bind_args /* {
108 		syscallarg(int)				s;
109 		syscallarg(const struct sockaddr *)	name;
110 		syscallarg(unsigned int)		namelen;
111 	} */ *uap = v;
112 	struct file	*fp;
113 	struct mbuf	*nam;
114 	int		error;
115 
116 	/* getsock() will use the descriptor for us */
117 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
118 		return (error);
119 	error = sockargs(&nam, SCARG(uap, name), SCARG(uap, namelen),
120 	    MT_SONAME);
121 	if (error) {
122 		FILE_UNUSE(fp, p);
123 		return (error);
124 	}
125 	error = sobind((struct socket *)fp->f_data, nam, p);
126 	m_freem(nam);
127 	FILE_UNUSE(fp, p);
128 	return (error);
129 }
130 
131 /* ARGSUSED */
132 int
133 sys_listen(struct proc *p, void *v, register_t *retval)
134 {
135 	struct sys_listen_args /* {
136 		syscallarg(int)	s;
137 		syscallarg(int)	backlog;
138 	} */ *uap = v;
139 	struct file	*fp;
140 	int		error;
141 
142 	/* getsock() will use the descriptor for us */
143 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
144 		return (error);
145 	error = solisten((struct socket *)fp->f_data, SCARG(uap, backlog));
146 	FILE_UNUSE(fp, p);
147 	return (error);
148 }
149 
150 int
151 sys_accept(struct proc *p, void *v, register_t *retval)
152 {
153 	struct sys_accept_args /* {
154 		syscallarg(int)			s;
155 		syscallarg(struct sockaddr *)	name;
156 		syscallarg(unsigned int *)	anamelen;
157 	} */ *uap = v;
158 	struct filedesc	*fdp;
159 	struct file	*fp;
160 	struct mbuf	*nam;
161 	unsigned int	namelen;
162 	int		error, s, fd;
163 	struct socket	*so;
164 
165 	fdp = p->p_fd;
166 	if (SCARG(uap, name) && (error = copyin((caddr_t)SCARG(uap, anamelen),
167 	    (caddr_t)&namelen, sizeof(namelen))))
168 		return (error);
169 	if (SCARG(uap, name) != NULL &&
170 	    uvm_useracc((caddr_t)SCARG(uap, name), sizeof(struct sockaddr),
171 	     B_WRITE) == FALSE)
172 		return (EFAULT);
173 
174 	/* getsock() will use the descriptor for us */
175 	if ((error = getsock(fdp, SCARG(uap, s), &fp)) != 0)
176 		return (error);
177 	s = splsoftnet();
178 	so = (struct socket *)fp->f_data;
179 	FILE_UNUSE(fp, p);
180 	if (!(so->so_proto->pr_flags & PR_LISTEN)) {
181 		splx(s);
182 		return (EOPNOTSUPP);
183 	}
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 	/* falloc() will use the descriptor for us */
211 	if ((error = falloc(p, &fp, &fd)) != 0) {
212 		splx(s);
213 		return (error);
214 	}
215 	*retval = fd;
216 	{ struct socket *aso = so->so_q.tqh_first;
217 	  if (soqremque(aso, 1) == 0)
218 		panic("accept");
219 	  so = aso;
220 	}
221 	fp->f_type = DTYPE_SOCKET;
222 	fp->f_flag = FREAD|FWRITE;
223 	fp->f_ops = &socketops;
224 	fp->f_data = (caddr_t)so;
225 	FILE_UNUSE(fp, p);
226 	nam = m_get(M_WAIT, MT_SONAME);
227 	if ((error = soaccept(so, nam)) == 0 && SCARG(uap, name)) {
228 		if (namelen > nam->m_len)
229 			namelen = nam->m_len;
230 		/* SHOULD COPY OUT A CHAIN HERE */
231 		if ((error = copyout(mtod(nam, caddr_t),
232 		    (caddr_t)SCARG(uap, name), namelen)) == 0)
233 			error = copyout((caddr_t)&namelen,
234 			    (caddr_t)SCARG(uap, anamelen),
235 			    sizeof(*SCARG(uap, anamelen)));
236 	}
237 	/* if an error occured, free the file descriptor */
238 	if (error) {
239 		fdremove(fdp, fd);
240 		ffree(fp);
241 	}
242 	m_freem(nam);
243 	splx(s);
244 	FILE_SET_MATURE(fp);
245 	return (error);
246 }
247 
248 /* ARGSUSED */
249 int
250 sys_connect(struct proc *p, void *v, register_t *retval)
251 {
252 	struct sys_connect_args /* {
253 		syscallarg(int)				s;
254 		syscallarg(const struct sockaddr *)	name;
255 		syscallarg(unsigned int)		namelen;
256 	} */ *uap = v;
257 	struct file	*fp;
258 	struct socket	*so;
259 	struct mbuf	*nam;
260 	int		error, s;
261 
262 	/* getsock() will use the descriptor for us */
263 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
264 		return (error);
265 	so = (struct socket *)fp->f_data;
266 	if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING)) {
267 		error = EALREADY;
268 		goto out;
269 	}
270 	error = sockargs(&nam, SCARG(uap, name), SCARG(uap, namelen),
271 	    MT_SONAME);
272 	if (error)
273 		goto out;
274 	error = soconnect(so, nam);
275 	if (error)
276 		goto bad;
277 	if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING)) {
278 		m_freem(nam);
279 		error = EINPROGRESS;
280 		goto out;
281 	}
282 	s = splsoftnet();
283 	while ((so->so_state & SS_ISCONNECTING) && so->so_error == 0) {
284 		error = tsleep((caddr_t)&so->so_timeo, PSOCK | PCATCH,
285 			       netcon, 0);
286 		if (error)
287 			break;
288 	}
289 	if (error == 0) {
290 		error = so->so_error;
291 		so->so_error = 0;
292 	}
293 	splx(s);
294  bad:
295 	so->so_state &= ~SS_ISCONNECTING;
296 	m_freem(nam);
297 	if (error == ERESTART)
298 		error = EINTR;
299  out:
300 	FILE_UNUSE(fp, p);
301 	return (error);
302 }
303 
304 int
305 sys_socketpair(struct proc *p, void *v, register_t *retval)
306 {
307 	struct sys_socketpair_args /* {
308 		syscallarg(int)		domain;
309 		syscallarg(int)		type;
310 		syscallarg(int)		protocol;
311 		syscallarg(int *)	rsv;
312 	} */ *uap = v;
313 	struct filedesc	*fdp;
314 	struct file	*fp1, *fp2;
315 	struct socket	*so1, *so2;
316 	int		fd, error, sv[2];
317 
318 	fdp = p->p_fd;
319 	error = socreate(SCARG(uap, domain), &so1, SCARG(uap, type),
320 			 SCARG(uap, protocol));
321 	if (error)
322 		return (error);
323 	error = socreate(SCARG(uap, domain), &so2, SCARG(uap, type),
324 			 SCARG(uap, protocol));
325 	if (error)
326 		goto free1;
327 	/* falloc() will use the descriptor for us */
328 	if ((error = falloc(p, &fp1, &fd)) != 0)
329 		goto free2;
330 	sv[0] = fd;
331 	fp1->f_flag = FREAD|FWRITE;
332 	fp1->f_type = DTYPE_SOCKET;
333 	fp1->f_ops = &socketops;
334 	fp1->f_data = (caddr_t)so1;
335 	if ((error = falloc(p, &fp2, &fd)) != 0)
336 		goto free3;
337 	fp2->f_flag = FREAD|FWRITE;
338 	fp2->f_type = DTYPE_SOCKET;
339 	fp2->f_ops = &socketops;
340 	fp2->f_data = (caddr_t)so2;
341 	sv[1] = fd;
342 	if ((error = soconnect2(so1, so2)) != 0)
343 		goto free4;
344 	if (SCARG(uap, type) == SOCK_DGRAM) {
345 		/*
346 		 * Datagram socket connection is asymmetric.
347 		 */
348 		 if ((error = soconnect2(so2, so1)) != 0)
349 			goto free4;
350 	}
351 	error = copyout((caddr_t)sv, (caddr_t)SCARG(uap, rsv),
352 	    2 * sizeof(int));
353 	FILE_SET_MATURE(fp1);
354 	FILE_SET_MATURE(fp2);
355 	FILE_UNUSE(fp1, p);
356 	FILE_UNUSE(fp2, p);
357 	return (error);
358  free4:
359 	FILE_UNUSE(fp2, p);
360 	ffree(fp2);
361 	fdremove(fdp, sv[1]);
362  free3:
363 	FILE_UNUSE(fp1, p);
364 	ffree(fp1);
365 	fdremove(fdp, sv[0]);
366  free2:
367 	(void)soclose(so2);
368  free1:
369 	(void)soclose(so1);
370 	return (error);
371 }
372 
373 int
374 sys_sendto(struct proc *p, void *v, register_t *retval)
375 {
376 	struct sys_sendto_args /* {
377 		syscallarg(int)				s;
378 		syscallarg(const void *)		buf;
379 		syscallarg(size_t)			len;
380 		syscallarg(int)				flags;
381 		syscallarg(const struct sockaddr *)	to;
382 		syscallarg(unsigned int)		tolen;
383 	} */ *uap = v;
384 	struct msghdr	msg;
385 	struct iovec	aiov;
386 
387 	msg.msg_name = (caddr_t)SCARG(uap, to);		/* XXX kills const */
388 	msg.msg_namelen = SCARG(uap, tolen);
389 	msg.msg_iov = &aiov;
390 	msg.msg_iovlen = 1;
391 	msg.msg_control = 0;
392 	msg.msg_flags = 0;
393 	aiov.iov_base = (char *)SCARG(uap, buf);	/* XXX kills const */
394 	aiov.iov_len = SCARG(uap, len);
395 	return (sendit(p, SCARG(uap, s), &msg, SCARG(uap, flags), retval));
396 }
397 
398 int
399 sys_sendmsg(struct proc *p, void *v, register_t *retval)
400 {
401 	struct sys_sendmsg_args /* {
402 		syscallarg(int)				s;
403 		syscallarg(const struct msghdr *)	msg;
404 		syscallarg(int)				flags;
405 	} */ *uap = v;
406 	struct msghdr	msg;
407 	struct iovec	aiov[UIO_SMALLIOV], *iov;
408 	int		error;
409 
410 	error = copyin(SCARG(uap, msg), (caddr_t)&msg, sizeof(msg));
411 	if (error)
412 		return (error);
413 	if ((unsigned int)msg.msg_iovlen > UIO_SMALLIOV) {
414 		if ((unsigned int)msg.msg_iovlen > IOV_MAX)
415 			return (EMSGSIZE);
416 		iov = malloc(sizeof(struct iovec) * msg.msg_iovlen,
417 		    M_IOV, M_WAITOK);
418 	} else
419 		iov = aiov;
420 	if ((unsigned int)msg.msg_iovlen > 0) {
421 		error = copyin((caddr_t)msg.msg_iov, (caddr_t)iov,
422 		    (size_t)(msg.msg_iovlen * sizeof(struct iovec)));
423 		if (error)
424 			goto done;
425 	}
426 	msg.msg_iov = iov;
427 	msg.msg_flags = 0;
428 	error = sendit(p, SCARG(uap, s), &msg, SCARG(uap, flags), retval);
429 done:
430 	if (iov != aiov)
431 		free(iov, M_IOV);
432 	return (error);
433 }
434 
435 int
436 sendit(struct proc *p, int s, struct msghdr *mp, int flags, register_t *retsize)
437 {
438 	struct file	*fp;
439 	struct uio	auio;
440 	struct iovec	*iov;
441 	int		i, len, error;
442 	struct mbuf	*to, *control;
443 	struct socket	*so;
444 #ifdef KTRACE
445 	struct iovec	*ktriov;
446 #endif
447 
448 #ifdef KTRACE
449 	ktriov = NULL;
450 #endif
451 	/* getsock() will use the descriptor for us */
452 	if ((error = getsock(p->p_fd, s, &fp)) != 0)
453 		return (error);
454 	auio.uio_iov = mp->msg_iov;
455 	auio.uio_iovcnt = mp->msg_iovlen;
456 	auio.uio_segflg = UIO_USERSPACE;
457 	auio.uio_rw = UIO_WRITE;
458 	auio.uio_procp = p;
459 	auio.uio_offset = 0;			/* XXX */
460 	auio.uio_resid = 0;
461 	iov = mp->msg_iov;
462 	for (i = 0; i < mp->msg_iovlen; i++, iov++) {
463 #if 0
464 		/* cannot happen; iov_len is unsigned */
465 		if (iov->iov_len < 0) {
466 			error = EINVAL;
467 			goto out;
468 		}
469 #endif
470 		/*
471 		 * Writes return ssize_t because -1 is returned on error.
472 		 * Therefore, we must restrict the length to SSIZE_MAX to
473 		 * avoid garbage return values.
474 		 */
475 		auio.uio_resid += iov->iov_len;
476 		if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
477 			error = EINVAL;
478 			goto out;
479 		}
480 	}
481 	if (mp->msg_name) {
482 		error = sockargs(&to, mp->msg_name, mp->msg_namelen,
483 				 MT_SONAME);
484 		if (error)
485 			goto out;
486 	} else
487 		to = 0;
488 	if (mp->msg_control) {
489 		if (mp->msg_controllen < sizeof(struct cmsghdr)) {
490 			error = EINVAL;
491 			goto bad;
492 		}
493 		error = sockargs(&control, mp->msg_control,
494 				 mp->msg_controllen, MT_CONTROL);
495 		if (error)
496 			goto bad;
497 	} else
498 		control = 0;
499 #ifdef KTRACE
500 	if (KTRPOINT(p, KTR_GENIO)) {
501 		int iovlen = auio.uio_iovcnt * sizeof(struct iovec);
502 
503 		ktriov = malloc(iovlen, M_TEMP, M_WAITOK);
504 		memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen);
505 	}
506 #endif
507 	len = auio.uio_resid;
508 	so = (struct socket *)fp->f_data;
509 	error = (*so->so_send)(so, to, &auio, NULL, control, flags);
510 	if (error) {
511 		if (auio.uio_resid != len && (error == ERESTART ||
512 		    error == EINTR || error == EWOULDBLOCK))
513 			error = 0;
514 		if (error == EPIPE)
515 			psignal(p, SIGPIPE);
516 	}
517 	if (error == 0)
518 		*retsize = len - auio.uio_resid;
519 #ifdef KTRACE
520 	if (ktriov != NULL) {
521 		if (error == 0)
522 			ktrgenio(p, s, UIO_WRITE, ktriov, *retsize, error);
523 		free(ktriov, M_TEMP);
524 	}
525 #endif
526  bad:
527 	if (to)
528 		m_freem(to);
529  out:
530 	FILE_UNUSE(fp, p);
531 	return (error);
532 }
533 
534 int
535 sys_recvfrom(struct proc *p, void *v, register_t *retval)
536 {
537 	struct sys_recvfrom_args /* {
538 		syscallarg(int)			s;
539 		syscallarg(void *)		buf;
540 		syscallarg(size_t)		len;
541 		syscallarg(int)			flags;
542 		syscallarg(struct sockaddr *)	from;
543 		syscallarg(unsigned int *)	fromlenaddr;
544 	} */ *uap = v;
545 	struct msghdr	msg;
546 	struct iovec	aiov;
547 	int		error;
548 
549 	if (SCARG(uap, fromlenaddr)) {
550 		error = copyin((caddr_t)SCARG(uap, fromlenaddr),
551 			       (caddr_t)&msg.msg_namelen,
552 			       sizeof(msg.msg_namelen));
553 		if (error)
554 			return (error);
555 	} else
556 		msg.msg_namelen = 0;
557 	msg.msg_name = (caddr_t)SCARG(uap, from);
558 	msg.msg_iov = &aiov;
559 	msg.msg_iovlen = 1;
560 	aiov.iov_base = SCARG(uap, buf);
561 	aiov.iov_len = SCARG(uap, len);
562 	msg.msg_control = 0;
563 	msg.msg_flags = SCARG(uap, flags);
564 	return (recvit(p, SCARG(uap, s), &msg,
565 		       (caddr_t)SCARG(uap, fromlenaddr), retval));
566 }
567 
568 int
569 sys_recvmsg(struct proc *p, void *v, register_t *retval)
570 {
571 	struct sys_recvmsg_args /* {
572 		syscallarg(int)			s;
573 		syscallarg(struct msghdr *)	msg;
574 		syscallarg(int)			flags;
575 	} */ *uap = v;
576 	struct msghdr	msg;
577 	struct iovec	aiov[UIO_SMALLIOV], *uiov, *iov;
578 	int		error;
579 
580 	error = copyin((caddr_t)SCARG(uap, msg), (caddr_t)&msg,
581 		       sizeof(msg));
582 	if (error)
583 		return (error);
584 	if ((unsigned int)msg.msg_iovlen > UIO_SMALLIOV) {
585 		if ((unsigned int)msg.msg_iovlen > IOV_MAX)
586 			return (EMSGSIZE);
587 		iov = malloc(sizeof(struct iovec) * msg.msg_iovlen,
588 		    M_IOV, M_WAITOK);
589 	} else
590 		iov = aiov;
591 	if ((unsigned int)msg.msg_iovlen > 0) {
592 		error = copyin((caddr_t)msg.msg_iov, (caddr_t)iov,
593 		    (size_t)(msg.msg_iovlen * sizeof(struct iovec)));
594 		if (error)
595 			goto done;
596 	}
597 	uiov = msg.msg_iov;
598 	msg.msg_iov = iov;
599 	msg.msg_flags = SCARG(uap, flags);
600 	if ((error = recvit(p, SCARG(uap, s), &msg, (caddr_t)0, retval)) == 0) {
601 		msg.msg_iov = uiov;
602 		error = copyout((caddr_t)&msg, (caddr_t)SCARG(uap, msg),
603 		    sizeof(msg));
604 	}
605 done:
606 	if (iov != aiov)
607 		free(iov, M_IOV);
608 	return (error);
609 }
610 
611 int
612 recvit(struct proc *p, int s, struct msghdr *mp, caddr_t namelenp,
613 	register_t *retsize)
614 {
615 	struct file	*fp;
616 	struct uio	auio;
617 	struct iovec	*iov;
618 	int		i, len, error;
619 	struct mbuf	*from, *control;
620 	struct socket	*so;
621 #ifdef KTRACE
622 	struct iovec	*ktriov;
623 #endif
624 
625 	from = 0;
626 	control = 0;
627 #ifdef KTRACE
628 	ktriov = NULL;
629 #endif
630 
631 	/* getsock() will use the descriptor for us */
632 	if ((error = getsock(p->p_fd, s, &fp)) != 0)
633 		return (error);
634 	auio.uio_iov = mp->msg_iov;
635 	auio.uio_iovcnt = mp->msg_iovlen;
636 	auio.uio_segflg = UIO_USERSPACE;
637 	auio.uio_rw = UIO_READ;
638 	auio.uio_procp = p;
639 	auio.uio_offset = 0;			/* XXX */
640 	auio.uio_resid = 0;
641 	iov = mp->msg_iov;
642 	for (i = 0; i < mp->msg_iovlen; i++, iov++) {
643 #if 0
644 		/* cannot happen iov_len is unsigned */
645 		if (iov->iov_len < 0) {
646 			error = EINVAL;
647 			goto out1;
648 		}
649 #endif
650 		/*
651 		 * Reads return ssize_t because -1 is returned on error.
652 		 * Therefore we must restrict the length to SSIZE_MAX to
653 		 * avoid garbage return values.
654 		 */
655 		auio.uio_resid += iov->iov_len;
656 		if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
657 			error = EINVAL;
658 			goto out1;
659 		}
660 	}
661 #ifdef KTRACE
662 	if (KTRPOINT(p, KTR_GENIO)) {
663 		int iovlen = auio.uio_iovcnt * sizeof(struct iovec);
664 
665 		ktriov = malloc(iovlen, M_TEMP, M_WAITOK);
666 		memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen);
667 	}
668 #endif
669 	len = auio.uio_resid;
670 	so = (struct socket *)fp->f_data;
671 	error = (*so->so_receive)(so, &from, &auio, NULL,
672 			  mp->msg_control ? &control : NULL, &mp->msg_flags);
673 	if (error) {
674 		if (auio.uio_resid != len && (error == ERESTART ||
675 		    error == EINTR || error == EWOULDBLOCK))
676 			error = 0;
677 	}
678 #ifdef KTRACE
679 	if (ktriov != NULL) {
680 		if (error == 0)
681 			ktrgenio(p, s, UIO_READ, ktriov,
682 			    len - auio.uio_resid, error);
683 		free(ktriov, M_TEMP);
684 	}
685 #endif
686 	if (error)
687 		goto out;
688 	*retsize = len - auio.uio_resid;
689 	if (mp->msg_name) {
690 		len = mp->msg_namelen;
691 		if (len <= 0 || from == 0)
692 			len = 0;
693 		else {
694 			if (len > from->m_len)
695 				len = from->m_len;
696 			/* else if len < from->m_len ??? */
697 			error = copyout(mtod(from, caddr_t),
698 					(caddr_t)mp->msg_name, (unsigned)len);
699 			if (error)
700 				goto out;
701 		}
702 		mp->msg_namelen = len;
703 		if (namelenp &&
704 		    (error = copyout((caddr_t)&len, namelenp, sizeof(int))))
705 			goto out;
706 	}
707 	if (mp->msg_control) {
708 		len = mp->msg_controllen;
709 		if (len <= 0 || control == 0)
710 			len = 0;
711 		else {
712 			struct mbuf *m = control;
713 			caddr_t p = (caddr_t)mp->msg_control;
714 
715 			do {
716 				i = m->m_len;
717 				if (len < i) {
718 					mp->msg_flags |= MSG_CTRUNC;
719 					i = len;
720 				}
721 				error = copyout(mtod(m, caddr_t), p,
722 				    (unsigned)i);
723 				if (m->m_next)
724 					i = ALIGN(i);
725 				p += i;
726 				len -= i;
727 				if (error != 0 || len <= 0)
728 					break;
729 			} while ((m = m->m_next) != NULL);
730 			len = p - (caddr_t)mp->msg_control;
731 		}
732 		mp->msg_controllen = len;
733 	}
734  out:
735 	if (from)
736 		m_freem(from);
737 	if (control)
738 		m_freem(control);
739  out1:
740 	FILE_UNUSE(fp, p);
741 	return (error);
742 }
743 
744 /* ARGSUSED */
745 int
746 sys_shutdown(struct proc *p, void *v, register_t *retval)
747 {
748 	struct sys_shutdown_args /* {
749 		syscallarg(int)	s;
750 		syscallarg(int)	how;
751 	} */ *uap = v;
752 	struct file	*fp;
753 	int		error;
754 
755 	/* getsock() will use the descriptor for us */
756 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
757 		return (error);
758 	error = soshutdown((struct socket *)fp->f_data, SCARG(uap, how));
759 	FILE_UNUSE(fp, p);
760 	return (error);
761 }
762 
763 /* ARGSUSED */
764 int
765 sys_setsockopt(struct proc *p, void *v, register_t *retval)
766 {
767 	struct sys_setsockopt_args /* {
768 		syscallarg(int)			s;
769 		syscallarg(int)			level;
770 		syscallarg(int)			name;
771 		syscallarg(const void *)	val;
772 		syscallarg(unsigned int)	valsize;
773 	} */ *uap = v;
774 	struct file	*fp;
775 	struct mbuf	*m;
776 	int		error;
777 
778 	m = NULL;
779 	/* getsock() will use the descriptor for us */
780 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
781 		return (error);
782 	if (SCARG(uap, valsize) > MLEN) {
783 		error = EINVAL;
784 		goto out;
785 	}
786 	if (SCARG(uap, val)) {
787 		m = m_get(M_WAIT, MT_SOOPTS);
788 		error = copyin(SCARG(uap, val), mtod(m, caddr_t),
789 			       SCARG(uap, valsize));
790 		if (error) {
791 			(void) m_free(m);
792 			goto out;
793 		}
794 		m->m_len = SCARG(uap, valsize);
795 	}
796 	error = sosetopt((struct socket *)fp->f_data, SCARG(uap, level),
797 			 SCARG(uap, name), m);
798  out:
799 	FILE_UNUSE(fp, p);
800 	return (error);
801 }
802 
803 /* ARGSUSED */
804 int
805 sys_getsockopt(struct proc *p, void *v, register_t *retval)
806 {
807 	struct sys_getsockopt_args /* {
808 		syscallarg(int)			s;
809 		syscallarg(int)			level;
810 		syscallarg(int)			name;
811 		syscallarg(void *)		val;
812 		syscallarg(unsigned int *)	avalsize;
813 	} */ *uap = v;
814 	struct file	*fp;
815 	struct mbuf	*m, *m0;
816 	unsigned int	op, i, valsize;
817 	int		error;
818 
819 	m = NULL;
820 	/* getsock() will use the descriptor for us */
821 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
822 		return (error);
823 	if (SCARG(uap, val)) {
824 		error = copyin((caddr_t)SCARG(uap, avalsize),
825 			       (caddr_t)&valsize, sizeof(valsize));
826 		if (error)
827 			goto out;
828 	} else
829 		valsize = 0;
830 	if ((error = sogetopt((struct socket *)fp->f_data, SCARG(uap, level),
831 	    SCARG(uap, name), &m)) == 0 && SCARG(uap, val) && valsize &&
832 	    m != NULL) {
833 		op = 0;
834 		while (m && !error && op < valsize) {
835 			i = min(m->m_len, (valsize - op));
836 			error = copyout(mtod(m, caddr_t), SCARG(uap, val), i);
837 			op += i;
838 			SCARG(uap, val) = ((u_int8_t *)SCARG(uap, val)) + i;
839 			m0 = m;
840 			MFREE(m0, m);
841 		}
842 		valsize = op;
843 		if (error == 0)
844 			error = copyout(&valsize,
845 					SCARG(uap, avalsize), sizeof(valsize));
846 	}
847 	if (m != NULL)
848 		(void) m_free(m);
849  out:
850 	FILE_UNUSE(fp, p);
851 	return (error);
852 }
853 
854 #ifndef NEW_PIPE
855 /* ARGSUSED */
856 int
857 sys_pipe(struct proc *p, void *v, register_t *retval)
858 {
859 	struct filedesc	*fdp;
860 	struct file	*rf, *wf;
861 	struct socket	*rso, *wso;
862 	int		fd, error;
863 
864 	fdp = p->p_fd;
865 	if ((error = socreate(AF_LOCAL, &rso, SOCK_STREAM, 0)) != 0)
866 		return (error);
867 	if ((error = socreate(AF_LOCAL, &wso, SOCK_STREAM, 0)) != 0)
868 		goto free1;
869 	/* remember this socket pair implements a pipe */
870 	wso->so_state |= SS_ISAPIPE;
871 	rso->so_state |= SS_ISAPIPE;
872 	/* falloc() will use the descriptor for us */
873 	if ((error = falloc(p, &rf, &fd)) != 0)
874 		goto free2;
875 	retval[0] = fd;
876 	rf->f_flag = FREAD;
877 	rf->f_type = DTYPE_SOCKET;
878 	rf->f_ops = &socketops;
879 	rf->f_data = (caddr_t)rso;
880 	if ((error = falloc(p, &wf, &fd)) != 0)
881 		goto free3;
882 	wf->f_flag = FWRITE;
883 	wf->f_type = DTYPE_SOCKET;
884 	wf->f_ops = &socketops;
885 	wf->f_data = (caddr_t)wso;
886 	retval[1] = fd;
887 	if ((error = unp_connect2(wso, rso)) != 0)
888 		goto free4;
889 	FILE_SET_MATURE(rf);
890 	FILE_SET_MATURE(wf);
891 	FILE_UNUSE(rf, p);
892 	FILE_UNUSE(wf, p);
893 	return (0);
894  free4:
895 	FILE_UNUSE(wf, p);
896 	ffree(wf);
897 	fdremove(fdp, retval[1]);
898  free3:
899 	FILE_UNUSE(rf, p);
900 	ffree(rf);
901 	fdremove(fdp, retval[0]);
902  free2:
903 	(void)soclose(wso);
904  free1:
905 	(void)soclose(rso);
906 	return (error);
907 }
908 #endif /* !NEW_PIPE */
909 
910 /*
911  * Get socket name.
912  */
913 /* ARGSUSED */
914 int
915 sys_getsockname(struct proc *p, void *v, register_t *retval)
916 {
917 	struct sys_getsockname_args /* {
918 		syscallarg(int)			fdes;
919 		syscallarg(struct sockaddr *)	asa;
920 		syscallarg(unsigned int *)	alen;
921 	} */ *uap = v;
922 	struct file	*fp;
923 	struct socket	*so;
924 	struct mbuf	*m;
925 	unsigned int	len;
926 	int		error;
927 
928 	/* getsock() will use the descriptor for us */
929 	if ((error = getsock(p->p_fd, SCARG(uap, fdes), &fp)) != 0)
930 		return (error);
931 	error = copyin((caddr_t)SCARG(uap, alen), (caddr_t)&len, sizeof(len));
932 	if (error)
933 		goto out;
934 	so = (struct socket *)fp->f_data;
935 	m = m_getclr(M_WAIT, MT_SONAME);
936 	error = (*so->so_proto->pr_usrreq)(so, PRU_SOCKADDR, (struct mbuf *)0,
937 	    m, (struct mbuf *)0, (struct proc *)0);
938 	if (error)
939 		goto bad;
940 	if (len > m->m_len)
941 		len = m->m_len;
942 	error = copyout(mtod(m, caddr_t), (caddr_t)SCARG(uap, asa), len);
943 	if (error == 0)
944 		error = copyout((caddr_t)&len, (caddr_t)SCARG(uap, alen),
945 		    sizeof(len));
946  bad:
947 	m_freem(m);
948  out:
949 	FILE_UNUSE(fp, p);
950 	return (error);
951 }
952 
953 /*
954  * Get name of peer for connected socket.
955  */
956 /* ARGSUSED */
957 int
958 sys_getpeername(struct proc *p, void *v, register_t *retval)
959 {
960 	struct sys_getpeername_args /* {
961 		syscallarg(int)			fdes;
962 		syscallarg(struct sockaddr *)	asa;
963 		syscallarg(unsigned int *)	alen;
964 	} */ *uap = v;
965 	struct file	*fp;
966 	struct socket	*so;
967 	struct mbuf	*m;
968 	unsigned int	len;
969 	int		error;
970 
971 	/* getsock() will use the descriptor for us */
972 	if ((error = getsock(p->p_fd, SCARG(uap, fdes), &fp)) != 0)
973 		return (error);
974 	so = (struct socket *)fp->f_data;
975 	if ((so->so_state & (SS_ISCONNECTED|SS_ISCONFIRMING)) == 0) {
976 		error = ENOTCONN;
977 		goto out;
978 	}
979 	error = copyin((caddr_t)SCARG(uap, alen), (caddr_t)&len, sizeof(len));
980 	if (error)
981 		goto out;
982 	m = m_getclr(M_WAIT, MT_SONAME);
983 	error = (*so->so_proto->pr_usrreq)(so, PRU_PEERADDR, (struct mbuf *)0,
984 	    m, (struct mbuf *)0, (struct proc *)0);
985 	if (error)
986 		goto bad;
987 	if (len > m->m_len)
988 		len = m->m_len;
989 	error = copyout(mtod(m, caddr_t), (caddr_t)SCARG(uap, asa), len);
990 	if (error)
991 		goto bad;
992 	error = copyout((caddr_t)&len, (caddr_t)SCARG(uap, alen), sizeof(len));
993  bad:
994 	m_freem(m);
995  out:
996 	FILE_UNUSE(fp, p);
997 	return (error);
998 }
999 
1000 /*
1001  * XXX In a perfect world, we wouldn't pass around socket control
1002  * XXX arguments in mbufs, and this could go away.
1003  */
1004 int
1005 sockargs(struct mbuf **mp, const void *buf, size_t buflen, int type)
1006 {
1007 	struct sockaddr	*sa;
1008 	struct mbuf	*m;
1009 	int		error;
1010 
1011 	/*
1012 	 * We can't allow socket names > UCHAR_MAX in length, since that
1013 	 * will overflow sa_len.  Control data more than a page size in
1014 	 * length is just too much.
1015 	 */
1016 	if (buflen > (type == MT_SONAME ? UCHAR_MAX : PAGE_SIZE))
1017 		return (EINVAL);
1018 
1019 	/* Allocate an mbuf to hold the arguments. */
1020 	m = m_get(M_WAIT, type);
1021 	if (buflen > MLEN) {
1022 		/*
1023 		 * Won't fit into a regular mbuf, so we allocate just
1024 		 * enough external storage to hold the argument.
1025 		 */
1026 		MEXTMALLOC(m, buflen, M_WAITOK);
1027 	}
1028 	m->m_len = buflen;
1029 	error = copyin(buf, mtod(m, caddr_t), buflen);
1030 	if (error) {
1031 		(void) m_free(m);
1032 		return (error);
1033 	}
1034 	*mp = m;
1035 	if (type == MT_SONAME) {
1036 		sa = mtod(m, struct sockaddr *);
1037 #if BYTE_ORDER != BIG_ENDIAN
1038 		/*
1039 		 * 4.3BSD compat thing - need to stay, since bind(2),
1040 		 * connect(2), sendto(2) were not versioned for COMPAT_43.
1041 		 */
1042 		if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
1043 			sa->sa_family = sa->sa_len;
1044 #endif
1045 		sa->sa_len = buflen;
1046 	}
1047 	return (0);
1048 }
1049 
1050 int
1051 getsock(struct filedesc *fdp, int fdes, struct file **fpp)
1052 {
1053 	struct file	*fp;
1054 
1055 	if ((fp = fd_getfile(fdp, fdes)) == NULL)
1056 		return (EBADF);
1057 
1058 	FILE_USE(fp);
1059 
1060 	if (fp->f_type != DTYPE_SOCKET) {
1061 		FILE_UNUSE(fp, NULL);
1062 		return (ENOTSOCK);
1063 	}
1064 	*fpp = fp;
1065 	return (0);
1066 }
1067