xref: /netbsd-src/sys/kern/uipc_syscalls.c (revision d710132b4b8ce7f7cccaaf660cb16aa16b4077a0)
1 /*	$NetBSD: uipc_syscalls.c,v 1.79 2003/04/05 16:54:34 christos 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 <sys/cdefs.h>
39 __KERNEL_RCSID(0, "$NetBSD: uipc_syscalls.c,v 1.79 2003/04/05 16:54:34 christos Exp $");
40 
41 #include "opt_ktrace.h"
42 #include "opt_pipe.h"
43 
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/filedesc.h>
47 #include <sys/proc.h>
48 #include <sys/file.h>
49 #include <sys/buf.h>
50 #include <sys/malloc.h>
51 #include <sys/mbuf.h>
52 #include <sys/protosw.h>
53 #include <sys/socket.h>
54 #include <sys/socketvar.h>
55 #include <sys/signalvar.h>
56 #include <sys/un.h>
57 #ifdef KTRACE
58 #include <sys/ktrace.h>
59 #endif
60 #include <sys/event.h>
61 
62 #include <sys/mount.h>
63 #include <sys/sa.h>
64 #include <sys/syscallargs.h>
65 
66 #include <uvm/uvm_extern.h>
67 
68 /*
69  * System call interface to the socket abstraction.
70  */
71 extern	struct fileops socketops;
72 
73 int
74 sys_socket(struct lwp *l, void *v, register_t *retval)
75 {
76 	struct sys_socket_args /* {
77 		syscallarg(int)	domain;
78 		syscallarg(int)	type;
79 		syscallarg(int)	protocol;
80 	} */ *uap = v;
81 
82 	struct proc	*p;
83 	struct filedesc	*fdp;
84 	struct socket	*so;
85 	struct file	*fp;
86 	int		fd, error;
87 
88 	p = l->l_proc;
89 	fdp = p->p_fd;
90 	/* falloc() will use the desciptor for us */
91 	if ((error = falloc(p, &fp, &fd)) != 0)
92 		return (error);
93 	fp->f_flag = FREAD|FWRITE;
94 	fp->f_type = DTYPE_SOCKET;
95 	fp->f_ops = &socketops;
96 	error = socreate(SCARG(uap, domain), &so, SCARG(uap, type),
97 			 SCARG(uap, protocol));
98 	if (error) {
99 		FILE_UNUSE(fp, p);
100 		fdremove(fdp, fd);
101 		ffree(fp);
102 	} else {
103 		fp->f_data = (caddr_t)so;
104 		FILE_SET_MATURE(fp);
105 		FILE_UNUSE(fp, p);
106 		*retval = fd;
107 	}
108 	return (error);
109 }
110 
111 /* ARGSUSED */
112 int
113 sys_bind(struct lwp *l, void *v, register_t *retval)
114 {
115 	struct sys_bind_args /* {
116 		syscallarg(int)				s;
117 		syscallarg(const struct sockaddr *)	name;
118 		syscallarg(unsigned int)		namelen;
119 	} */ *uap = v;
120 	struct proc	*p;
121 	struct file	*fp;
122 	struct mbuf	*nam;
123 	int		error;
124 
125 	p = l->l_proc;
126 	/* getsock() will use the descriptor for us */
127 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
128 		return (error);
129 	error = sockargs(&nam, SCARG(uap, name), SCARG(uap, namelen),
130 	    MT_SONAME);
131 	if (error) {
132 		FILE_UNUSE(fp, p);
133 		return (error);
134 	}
135 	MCLAIM(nam, ((struct socket *)fp->f_data)->so_mowner);
136 	error = sobind((struct socket *)fp->f_data, nam, p);
137 	m_freem(nam);
138 	FILE_UNUSE(fp, p);
139 	return (error);
140 }
141 
142 /* ARGSUSED */
143 int
144 sys_listen(struct lwp *l, void *v, register_t *retval)
145 {
146 	struct sys_listen_args /* {
147 		syscallarg(int)	s;
148 		syscallarg(int)	backlog;
149 	} */ *uap = v;
150 	struct proc	*p;
151 	struct file	*fp;
152 	int		error;
153 
154 	p = l->l_proc;
155 	/* getsock() will use the descriptor for us */
156 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
157 		return (error);
158 	error = solisten((struct socket *)fp->f_data, SCARG(uap, backlog));
159 	FILE_UNUSE(fp, p);
160 	return (error);
161 }
162 
163 int
164 sys_accept(struct lwp *l, void *v, register_t *retval)
165 {
166 	struct sys_accept_args /* {
167 		syscallarg(int)			s;
168 		syscallarg(struct sockaddr *)	name;
169 		syscallarg(unsigned int *)	anamelen;
170 	} */ *uap = v;
171 	struct proc	*p;
172 	struct filedesc	*fdp;
173 	struct file	*fp;
174 	struct mbuf	*nam;
175 	unsigned int	namelen;
176 	int		error, s, fd;
177 	struct socket	*so;
178 	int		fflag;
179 
180 	p = l->l_proc;
181 	fdp = p->p_fd;
182 	if (SCARG(uap, name) && (error = copyin((caddr_t)SCARG(uap, anamelen),
183 	    (caddr_t)&namelen, sizeof(namelen))))
184 		return (error);
185 	if (SCARG(uap, name) != NULL &&
186 	    uvm_useracc((caddr_t)SCARG(uap, name), sizeof(struct sockaddr),
187 	     B_WRITE) == FALSE)
188 		return (EFAULT);
189 
190 	/* getsock() will use the descriptor for us */
191 	if ((error = getsock(fdp, SCARG(uap, s), &fp)) != 0)
192 		return (error);
193 	s = splsoftnet();
194 	so = (struct socket *)fp->f_data;
195 	FILE_UNUSE(fp, p);
196 	if (!(so->so_proto->pr_flags & PR_LISTEN)) {
197 		splx(s);
198 		return (EOPNOTSUPP);
199 	}
200 	if ((so->so_options & SO_ACCEPTCONN) == 0) {
201 		splx(s);
202 		return (EINVAL);
203 	}
204 	if ((so->so_state & SS_NBIO) && so->so_qlen == 0) {
205 		splx(s);
206 		return (EWOULDBLOCK);
207 	}
208 	while (so->so_qlen == 0 && so->so_error == 0) {
209 		if (so->so_state & SS_CANTRCVMORE) {
210 			so->so_error = ECONNABORTED;
211 			break;
212 		}
213 		error = tsleep((caddr_t)&so->so_timeo, PSOCK | PCATCH,
214 			       netcon, 0);
215 		if (error) {
216 			splx(s);
217 			return (error);
218 		}
219 	}
220 	if (so->so_error) {
221 		error = so->so_error;
222 		so->so_error = 0;
223 		splx(s);
224 		return (error);
225 	}
226 	fflag = fp->f_flag;
227 	/* falloc() will use the descriptor for us */
228 	if ((error = falloc(p, &fp, &fd)) != 0) {
229 		splx(s);
230 		return (error);
231 	}
232 	*retval = fd;
233 
234 	/* connection has been removed from the listen queue */
235 	KNOTE(&so->so_rcv.sb_sel.sel_klist, 0);
236 
237 	{ struct socket *aso = TAILQ_FIRST(&so->so_q);
238 	  if (soqremque(aso, 1) == 0)
239 		panic("accept");
240 	  so = aso;
241 	}
242 	fp->f_type = DTYPE_SOCKET;
243 	fp->f_flag = fflag;
244 	fp->f_ops = &socketops;
245 	fp->f_data = (caddr_t)so;
246 	FILE_UNUSE(fp, p);
247 	nam = m_get(M_WAIT, MT_SONAME);
248 	if ((error = soaccept(so, nam)) == 0 && SCARG(uap, name)) {
249 		if (namelen > nam->m_len)
250 			namelen = nam->m_len;
251 		/* SHOULD COPY OUT A CHAIN HERE */
252 		if ((error = copyout(mtod(nam, caddr_t),
253 		    (caddr_t)SCARG(uap, name), namelen)) == 0)
254 			error = copyout((caddr_t)&namelen,
255 			    (caddr_t)SCARG(uap, anamelen),
256 			    sizeof(*SCARG(uap, anamelen)));
257 	}
258 	/* if an error occurred, free the file descriptor */
259 	if (error) {
260 		fdremove(fdp, fd);
261 		ffree(fp);
262 	}
263 	m_freem(nam);
264 	splx(s);
265 	FILE_SET_MATURE(fp);
266 	return (error);
267 }
268 
269 /* ARGSUSED */
270 int
271 sys_connect(struct lwp *l, void *v, register_t *retval)
272 {
273 	struct sys_connect_args /* {
274 		syscallarg(int)				s;
275 		syscallarg(const struct sockaddr *)	name;
276 		syscallarg(unsigned int)		namelen;
277 	} */ *uap = v;
278 	struct proc	*p;
279 	struct file	*fp;
280 	struct socket	*so;
281 	struct mbuf	*nam;
282 	int		error, s;
283 
284 	p = l->l_proc;
285 	/* getsock() will use the descriptor for us */
286 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
287 		return (error);
288 	so = (struct socket *)fp->f_data;
289 	if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING)) {
290 		error = EALREADY;
291 		goto out;
292 	}
293 	error = sockargs(&nam, SCARG(uap, name), SCARG(uap, namelen),
294 	    MT_SONAME);
295 	if (error)
296 		goto out;
297 	MCLAIM(nam, so->so_mowner);
298 	error = soconnect(so, nam);
299 	if (error)
300 		goto bad;
301 	if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING)) {
302 		m_freem(nam);
303 		error = EINPROGRESS;
304 		goto out;
305 	}
306 	s = splsoftnet();
307 	while ((so->so_state & SS_ISCONNECTING) && so->so_error == 0) {
308 		error = tsleep((caddr_t)&so->so_timeo, PSOCK | PCATCH,
309 			       netcon, 0);
310 		if (error)
311 			break;
312 	}
313 	if (error == 0) {
314 		error = so->so_error;
315 		so->so_error = 0;
316 	}
317 	splx(s);
318  bad:
319 	so->so_state &= ~SS_ISCONNECTING;
320 	m_freem(nam);
321 	if (error == ERESTART)
322 		error = EINTR;
323  out:
324 	FILE_UNUSE(fp, p);
325 	return (error);
326 }
327 
328 int
329 sys_socketpair(struct lwp *l, void *v, register_t *retval)
330 {
331 	struct sys_socketpair_args /* {
332 		syscallarg(int)		domain;
333 		syscallarg(int)		type;
334 		syscallarg(int)		protocol;
335 		syscallarg(int *)	rsv;
336 	} */ *uap = v;
337 	struct proc *p;
338 	struct filedesc	*fdp;
339 	struct file	*fp1, *fp2;
340 	struct socket	*so1, *so2;
341 	int		fd, error, sv[2];
342 
343 	p = l->l_proc;
344 	fdp = p->p_fd;
345 	error = socreate(SCARG(uap, domain), &so1, SCARG(uap, type),
346 			 SCARG(uap, protocol));
347 	if (error)
348 		return (error);
349 	error = socreate(SCARG(uap, domain), &so2, SCARG(uap, type),
350 			 SCARG(uap, protocol));
351 	if (error)
352 		goto free1;
353 	/* falloc() will use the descriptor for us */
354 	if ((error = falloc(p, &fp1, &fd)) != 0)
355 		goto free2;
356 	sv[0] = fd;
357 	fp1->f_flag = FREAD|FWRITE;
358 	fp1->f_type = DTYPE_SOCKET;
359 	fp1->f_ops = &socketops;
360 	fp1->f_data = (caddr_t)so1;
361 	if ((error = falloc(p, &fp2, &fd)) != 0)
362 		goto free3;
363 	fp2->f_flag = FREAD|FWRITE;
364 	fp2->f_type = DTYPE_SOCKET;
365 	fp2->f_ops = &socketops;
366 	fp2->f_data = (caddr_t)so2;
367 	sv[1] = fd;
368 	if ((error = soconnect2(so1, so2)) != 0)
369 		goto free4;
370 	if (SCARG(uap, type) == SOCK_DGRAM) {
371 		/*
372 		 * Datagram socket connection is asymmetric.
373 		 */
374 		 if ((error = soconnect2(so2, so1)) != 0)
375 			goto free4;
376 	}
377 	error = copyout((caddr_t)sv, (caddr_t)SCARG(uap, rsv),
378 	    2 * sizeof(int));
379 	FILE_SET_MATURE(fp1);
380 	FILE_SET_MATURE(fp2);
381 	FILE_UNUSE(fp1, p);
382 	FILE_UNUSE(fp2, p);
383 	return (error);
384  free4:
385 	FILE_UNUSE(fp2, p);
386 	ffree(fp2);
387 	fdremove(fdp, sv[1]);
388  free3:
389 	FILE_UNUSE(fp1, p);
390 	ffree(fp1);
391 	fdremove(fdp, sv[0]);
392  free2:
393 	(void)soclose(so2);
394  free1:
395 	(void)soclose(so1);
396 	return (error);
397 }
398 
399 int
400 sys_sendto(struct lwp *l, void *v, register_t *retval)
401 {
402 	struct sys_sendto_args /* {
403 		syscallarg(int)				s;
404 		syscallarg(const void *)		buf;
405 		syscallarg(size_t)			len;
406 		syscallarg(int)				flags;
407 		syscallarg(const struct sockaddr *)	to;
408 		syscallarg(unsigned int)		tolen;
409 	} */ *uap = v;
410 	struct proc	*p;
411 	struct msghdr	msg;
412 	struct iovec	aiov;
413 
414 	p = l->l_proc;
415 	msg.msg_name = (caddr_t)SCARG(uap, to);		/* XXX kills const */
416 	msg.msg_namelen = SCARG(uap, tolen);
417 	msg.msg_iov = &aiov;
418 	msg.msg_iovlen = 1;
419 	msg.msg_control = 0;
420 	msg.msg_flags = 0;
421 	aiov.iov_base = (char *)SCARG(uap, buf);	/* XXX kills const */
422 	aiov.iov_len = SCARG(uap, len);
423 	return (sendit(p, SCARG(uap, s), &msg, SCARG(uap, flags), retval));
424 }
425 
426 int
427 sys_sendmsg(struct lwp *l, void *v, register_t *retval)
428 {
429 	struct sys_sendmsg_args /* {
430 		syscallarg(int)				s;
431 		syscallarg(const struct msghdr *)	msg;
432 		syscallarg(int)				flags;
433 	} */ *uap = v;
434 	struct proc	*p;
435 	struct msghdr	msg;
436 	struct iovec	aiov[UIO_SMALLIOV], *iov;
437 	int		error;
438 
439 	error = copyin(SCARG(uap, msg), (caddr_t)&msg, sizeof(msg));
440 	if (error)
441 		return (error);
442 	if ((unsigned int)msg.msg_iovlen > UIO_SMALLIOV) {
443 		if ((unsigned int)msg.msg_iovlen > IOV_MAX)
444 			return (EMSGSIZE);
445 		iov = malloc(sizeof(struct iovec) * msg.msg_iovlen,
446 		    M_IOV, M_WAITOK);
447 	} else
448 		iov = aiov;
449 	if ((unsigned int)msg.msg_iovlen > 0) {
450 		error = copyin((caddr_t)msg.msg_iov, (caddr_t)iov,
451 		    (size_t)(msg.msg_iovlen * sizeof(struct iovec)));
452 		if (error)
453 			goto done;
454 	}
455 	msg.msg_iov = iov;
456 	msg.msg_flags = 0;
457 	p = l->l_proc;
458 	error = sendit(p, SCARG(uap, s), &msg, SCARG(uap, flags), retval);
459 done:
460 	if (iov != aiov)
461 		free(iov, M_IOV);
462 	return (error);
463 }
464 
465 int
466 sendit(struct proc *p, int s, struct msghdr *mp, int flags, register_t *retsize)
467 {
468 	struct file	*fp;
469 	struct uio	auio;
470 	struct iovec	*iov;
471 	int		i, len, error;
472 	struct mbuf	*to, *control;
473 	struct socket	*so;
474 #ifdef KTRACE
475 	struct iovec	*ktriov;
476 #endif
477 
478 #ifdef KTRACE
479 	ktriov = NULL;
480 #endif
481 	/* getsock() will use the descriptor for us */
482 	if ((error = getsock(p->p_fd, s, &fp)) != 0)
483 		return (error);
484 	so = (struct socket *)fp->f_data;
485 	auio.uio_iov = mp->msg_iov;
486 	auio.uio_iovcnt = mp->msg_iovlen;
487 	auio.uio_segflg = UIO_USERSPACE;
488 	auio.uio_rw = UIO_WRITE;
489 	auio.uio_procp = p;
490 	auio.uio_offset = 0;			/* XXX */
491 	auio.uio_resid = 0;
492 	iov = mp->msg_iov;
493 	for (i = 0; i < mp->msg_iovlen; i++, iov++) {
494 #if 0
495 		/* cannot happen; iov_len is unsigned */
496 		if (iov->iov_len < 0) {
497 			error = EINVAL;
498 			goto out;
499 		}
500 #endif
501 		/*
502 		 * Writes return ssize_t because -1 is returned on error.
503 		 * Therefore, we must restrict the length to SSIZE_MAX to
504 		 * avoid garbage return values.
505 		 */
506 		auio.uio_resid += iov->iov_len;
507 		if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
508 			error = EINVAL;
509 			goto out;
510 		}
511 	}
512 	if (mp->msg_name) {
513 		error = sockargs(&to, mp->msg_name, mp->msg_namelen,
514 				 MT_SONAME);
515 		if (error)
516 			goto out;
517 		MCLAIM(to, so->so_mowner);
518 	} else
519 		to = 0;
520 	if (mp->msg_control) {
521 		if (mp->msg_controllen < sizeof(struct cmsghdr)) {
522 			error = EINVAL;
523 			goto bad;
524 		}
525 		error = sockargs(&control, mp->msg_control,
526 				 mp->msg_controllen, MT_CONTROL);
527 		if (error)
528 			goto bad;
529 		MCLAIM(control, so->so_mowner);
530 	} else
531 		control = 0;
532 #ifdef KTRACE
533 	if (KTRPOINT(p, KTR_GENIO)) {
534 		int iovlen = auio.uio_iovcnt * sizeof(struct iovec);
535 
536 		ktriov = malloc(iovlen, M_TEMP, M_WAITOK);
537 		memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen);
538 	}
539 #endif
540 	len = auio.uio_resid;
541 	error = (*so->so_send)(so, to, &auio, NULL, control, flags);
542 	if (error) {
543 		if (auio.uio_resid != len && (error == ERESTART ||
544 		    error == EINTR || error == EWOULDBLOCK))
545 			error = 0;
546 		if (error == EPIPE)
547 			psignal(p, SIGPIPE);
548 	}
549 	if (error == 0)
550 		*retsize = len - auio.uio_resid;
551 #ifdef KTRACE
552 	if (ktriov != NULL) {
553 		if (error == 0)
554 			ktrgenio(p, s, UIO_WRITE, ktriov, *retsize, error);
555 		free(ktriov, M_TEMP);
556 	}
557 #endif
558  bad:
559 	if (to)
560 		m_freem(to);
561  out:
562 	FILE_UNUSE(fp, p);
563 	return (error);
564 }
565 
566 int
567 sys_recvfrom(struct lwp *l, void *v, register_t *retval)
568 {
569 	struct sys_recvfrom_args /* {
570 		syscallarg(int)			s;
571 		syscallarg(void *)		buf;
572 		syscallarg(size_t)		len;
573 		syscallarg(int)			flags;
574 		syscallarg(struct sockaddr *)	from;
575 		syscallarg(unsigned int *)	fromlenaddr;
576 	} */ *uap = v;
577 	struct proc	*p;
578 	struct msghdr	msg;
579 	struct iovec	aiov;
580 	int		error;
581 
582 	if (SCARG(uap, fromlenaddr)) {
583 		error = copyin((caddr_t)SCARG(uap, fromlenaddr),
584 			       (caddr_t)&msg.msg_namelen,
585 			       sizeof(msg.msg_namelen));
586 		if (error)
587 			return (error);
588 	} else
589 		msg.msg_namelen = 0;
590 	msg.msg_name = (caddr_t)SCARG(uap, from);
591 	msg.msg_iov = &aiov;
592 	msg.msg_iovlen = 1;
593 	aiov.iov_base = SCARG(uap, buf);
594 	aiov.iov_len = SCARG(uap, len);
595 	msg.msg_control = 0;
596 	msg.msg_flags = SCARG(uap, flags);
597 	p = l->l_proc;
598 	return (recvit(p, SCARG(uap, s), &msg,
599 		       (caddr_t)SCARG(uap, fromlenaddr), retval));
600 }
601 
602 int
603 sys_recvmsg(struct lwp *l, void *v, register_t *retval)
604 {
605 	struct sys_recvmsg_args /* {
606 		syscallarg(int)			s;
607 		syscallarg(struct msghdr *)	msg;
608 		syscallarg(int)			flags;
609 	} */ *uap = v;
610 	struct proc	*p;
611 	struct msghdr	msg;
612 	struct iovec	aiov[UIO_SMALLIOV], *uiov, *iov;
613 	int		error;
614 
615 	error = copyin((caddr_t)SCARG(uap, msg), (caddr_t)&msg,
616 		       sizeof(msg));
617 	if (error)
618 		return (error);
619 	if ((unsigned int)msg.msg_iovlen > UIO_SMALLIOV) {
620 		if ((unsigned int)msg.msg_iovlen > IOV_MAX)
621 			return (EMSGSIZE);
622 		iov = malloc(sizeof(struct iovec) * msg.msg_iovlen,
623 		    M_IOV, M_WAITOK);
624 	} else
625 		iov = aiov;
626 	if ((unsigned int)msg.msg_iovlen > 0) {
627 		error = copyin((caddr_t)msg.msg_iov, (caddr_t)iov,
628 		    (size_t)(msg.msg_iovlen * sizeof(struct iovec)));
629 		if (error)
630 			goto done;
631 	}
632 	uiov = msg.msg_iov;
633 	msg.msg_iov = iov;
634 	msg.msg_flags = SCARG(uap, flags);
635 	p = l->l_proc;
636 	if ((error = recvit(p, SCARG(uap, s), &msg, (caddr_t)0, retval)) == 0) {
637 		msg.msg_iov = uiov;
638 		error = copyout((caddr_t)&msg, (caddr_t)SCARG(uap, msg),
639 		    sizeof(msg));
640 	}
641 done:
642 	if (iov != aiov)
643 		free(iov, M_IOV);
644 	return (error);
645 }
646 
647 int
648 recvit(struct proc *p, int s, struct msghdr *mp, caddr_t namelenp,
649 	register_t *retsize)
650 {
651 	struct file	*fp;
652 	struct uio	auio;
653 	struct iovec	*iov;
654 	int		i, len, error;
655 	struct mbuf	*from, *control;
656 	struct socket	*so;
657 #ifdef KTRACE
658 	struct iovec	*ktriov;
659 #endif
660 
661 	from = 0;
662 	control = 0;
663 #ifdef KTRACE
664 	ktriov = NULL;
665 #endif
666 
667 	/* getsock() will use the descriptor for us */
668 	if ((error = getsock(p->p_fd, s, &fp)) != 0)
669 		return (error);
670 	so = (struct socket *)fp->f_data;
671 	auio.uio_iov = mp->msg_iov;
672 	auio.uio_iovcnt = mp->msg_iovlen;
673 	auio.uio_segflg = UIO_USERSPACE;
674 	auio.uio_rw = UIO_READ;
675 	auio.uio_procp = p;
676 	auio.uio_offset = 0;			/* XXX */
677 	auio.uio_resid = 0;
678 	iov = mp->msg_iov;
679 	for (i = 0; i < mp->msg_iovlen; i++, iov++) {
680 #if 0
681 		/* cannot happen iov_len is unsigned */
682 		if (iov->iov_len < 0) {
683 			error = EINVAL;
684 			goto out1;
685 		}
686 #endif
687 		/*
688 		 * Reads return ssize_t because -1 is returned on error.
689 		 * Therefore we must restrict the length to SSIZE_MAX to
690 		 * avoid garbage return values.
691 		 */
692 		auio.uio_resid += iov->iov_len;
693 		if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
694 			error = EINVAL;
695 			goto out1;
696 		}
697 	}
698 #ifdef KTRACE
699 	if (KTRPOINT(p, KTR_GENIO)) {
700 		int iovlen = auio.uio_iovcnt * sizeof(struct iovec);
701 
702 		ktriov = malloc(iovlen, M_TEMP, M_WAITOK);
703 		memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen);
704 	}
705 #endif
706 	len = auio.uio_resid;
707 	error = (*so->so_receive)(so, &from, &auio, NULL,
708 			  mp->msg_control ? &control : NULL, &mp->msg_flags);
709 	if (error) {
710 		if (auio.uio_resid != len && (error == ERESTART ||
711 		    error == EINTR || error == EWOULDBLOCK))
712 			error = 0;
713 	}
714 #ifdef KTRACE
715 	if (ktriov != NULL) {
716 		if (error == 0)
717 			ktrgenio(p, s, UIO_READ, ktriov,
718 			    len - auio.uio_resid, error);
719 		free(ktriov, M_TEMP);
720 	}
721 #endif
722 	if (error)
723 		goto out;
724 	*retsize = len - auio.uio_resid;
725 	if (mp->msg_name) {
726 		len = mp->msg_namelen;
727 		if (len <= 0 || from == 0)
728 			len = 0;
729 		else {
730 			if (len > from->m_len)
731 				len = from->m_len;
732 			/* else if len < from->m_len ??? */
733 			error = copyout(mtod(from, caddr_t),
734 					(caddr_t)mp->msg_name, (unsigned)len);
735 			if (error)
736 				goto out;
737 		}
738 		mp->msg_namelen = len;
739 		if (namelenp &&
740 		    (error = copyout((caddr_t)&len, namelenp, sizeof(int))))
741 			goto out;
742 	}
743 	if (mp->msg_control) {
744 		len = mp->msg_controllen;
745 		if (len <= 0 || control == 0)
746 			len = 0;
747 		else {
748 			struct mbuf *m = control;
749 			caddr_t p = (caddr_t)mp->msg_control;
750 
751 			do {
752 				i = m->m_len;
753 				if (len < i) {
754 					mp->msg_flags |= MSG_CTRUNC;
755 					i = len;
756 				}
757 				error = copyout(mtod(m, caddr_t), p,
758 				    (unsigned)i);
759 				if (m->m_next)
760 					i = ALIGN(i);
761 				p += i;
762 				len -= i;
763 				if (error != 0 || len <= 0)
764 					break;
765 			} while ((m = m->m_next) != NULL);
766 			len = p - (caddr_t)mp->msg_control;
767 		}
768 		mp->msg_controllen = len;
769 	}
770  out:
771 	if (from)
772 		m_freem(from);
773 	if (control)
774 		m_freem(control);
775  out1:
776 	FILE_UNUSE(fp, p);
777 	return (error);
778 }
779 
780 /* ARGSUSED */
781 int
782 sys_shutdown(struct lwp *l, void *v, register_t *retval)
783 {
784 	struct sys_shutdown_args /* {
785 		syscallarg(int)	s;
786 		syscallarg(int)	how;
787 	} */ *uap = v;
788 	struct proc	*p;
789 	struct file	*fp;
790 	int		error;
791 
792 	p = l->l_proc;
793 	/* getsock() will use the descriptor for us */
794 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
795 		return (error);
796 	error = soshutdown((struct socket *)fp->f_data, SCARG(uap, how));
797 	FILE_UNUSE(fp, p);
798 	return (error);
799 }
800 
801 /* ARGSUSED */
802 int
803 sys_setsockopt(struct lwp *l, void *v, register_t *retval)
804 {
805 	struct sys_setsockopt_args /* {
806 		syscallarg(int)			s;
807 		syscallarg(int)			level;
808 		syscallarg(int)			name;
809 		syscallarg(const void *)	val;
810 		syscallarg(unsigned int)	valsize;
811 	} */ *uap = v;
812 	struct proc	*p;
813 	struct file	*fp;
814 	struct mbuf	*m;
815 	struct socket	*so;
816 	int		error;
817 	unsigned int	len;
818 
819 	p = l->l_proc;
820 	m = NULL;
821 	/* getsock() will use the descriptor for us */
822 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
823 		return (error);
824 	so = (struct socket *)fp->f_data;
825 	len = SCARG(uap, valsize);
826 	if (len > MCLBYTES) {
827 		error = EINVAL;
828 		goto out;
829 	}
830 	if (SCARG(uap, val)) {
831 		m = m_get(M_WAIT, MT_SOOPTS);
832 		MCLAIM(m, so->so_mowner);
833 		if (len > MLEN)
834 			m_clget(m, M_WAIT);
835 		error = copyin(SCARG(uap, val), mtod(m, caddr_t), len);
836 		if (error) {
837 			(void) m_free(m);
838 			goto out;
839 		}
840 		m->m_len = SCARG(uap, valsize);
841 	}
842 	error = sosetopt(so, SCARG(uap, level), SCARG(uap, name), m);
843  out:
844 	FILE_UNUSE(fp, p);
845 	return (error);
846 }
847 
848 /* ARGSUSED */
849 int
850 sys_getsockopt(struct lwp *l, void *v, register_t *retval)
851 {
852 	struct sys_getsockopt_args /* {
853 		syscallarg(int)			s;
854 		syscallarg(int)			level;
855 		syscallarg(int)			name;
856 		syscallarg(void *)		val;
857 		syscallarg(unsigned int *)	avalsize;
858 	} */ *uap = v;
859 	struct proc	*p;
860 	struct file	*fp;
861 	struct mbuf	*m;
862 	unsigned int	op, i, valsize;
863 	int		error;
864 
865 	p = l->l_proc;
866 	m = NULL;
867 	/* getsock() will use the descriptor for us */
868 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
869 		return (error);
870 	if (SCARG(uap, val)) {
871 		error = copyin((caddr_t)SCARG(uap, avalsize),
872 			       (caddr_t)&valsize, sizeof(valsize));
873 		if (error)
874 			goto out;
875 	} else
876 		valsize = 0;
877 	if ((error = sogetopt((struct socket *)fp->f_data, SCARG(uap, level),
878 	    SCARG(uap, name), &m)) == 0 && SCARG(uap, val) && valsize &&
879 	    m != NULL) {
880 		op = 0;
881 		while (m && !error && op < valsize) {
882 			i = min(m->m_len, (valsize - op));
883 			error = copyout(mtod(m, caddr_t), SCARG(uap, val), i);
884 			op += i;
885 			SCARG(uap, val) = ((u_int8_t *)SCARG(uap, val)) + i;
886 			m = m_free(m);
887 		}
888 		valsize = op;
889 		if (error == 0)
890 			error = copyout(&valsize,
891 					SCARG(uap, avalsize), sizeof(valsize));
892 	}
893 	if (m != NULL)
894 		(void) m_freem(m);
895  out:
896 	FILE_UNUSE(fp, p);
897 	return (error);
898 }
899 
900 #ifdef PIPE_SOCKETPAIR
901 /* ARGSUSED */
902 int
903 sys_pipe(struct lwp *l, void *v, register_t *retval)
904 {
905 	struct proc	*p;
906 	struct filedesc	*fdp;
907 	struct file	*rf, *wf;
908 	struct socket	*rso, *wso;
909 	int		fd, error;
910 
911 	p = l->l_proc;
912 	fdp = p->p_fd;
913 	if ((error = socreate(AF_LOCAL, &rso, SOCK_STREAM, 0)) != 0)
914 		return (error);
915 	if ((error = socreate(AF_LOCAL, &wso, SOCK_STREAM, 0)) != 0)
916 		goto free1;
917 	/* remember this socket pair implements a pipe */
918 	wso->so_state |= SS_ISAPIPE;
919 	rso->so_state |= SS_ISAPIPE;
920 	/* falloc() will use the descriptor for us */
921 	if ((error = falloc(p, &rf, &fd)) != 0)
922 		goto free2;
923 	retval[0] = fd;
924 	rf->f_flag = FREAD;
925 	rf->f_type = DTYPE_SOCKET;
926 	rf->f_ops = &socketops;
927 	rf->f_data = (caddr_t)rso;
928 	if ((error = falloc(p, &wf, &fd)) != 0)
929 		goto free3;
930 	wf->f_flag = FWRITE;
931 	wf->f_type = DTYPE_SOCKET;
932 	wf->f_ops = &socketops;
933 	wf->f_data = (caddr_t)wso;
934 	retval[1] = fd;
935 	if ((error = unp_connect2(wso, rso)) != 0)
936 		goto free4;
937 	FILE_SET_MATURE(rf);
938 	FILE_SET_MATURE(wf);
939 	FILE_UNUSE(rf, p);
940 	FILE_UNUSE(wf, p);
941 	return (0);
942  free4:
943 	FILE_UNUSE(wf, p);
944 	ffree(wf);
945 	fdremove(fdp, retval[1]);
946  free3:
947 	FILE_UNUSE(rf, p);
948 	ffree(rf);
949 	fdremove(fdp, retval[0]);
950  free2:
951 	(void)soclose(wso);
952  free1:
953 	(void)soclose(rso);
954 	return (error);
955 }
956 #endif /* PIPE_SOCKETPAIR */
957 
958 /*
959  * Get socket name.
960  */
961 /* ARGSUSED */
962 int
963 sys_getsockname(struct lwp *l, void *v, register_t *retval)
964 {
965 	struct sys_getsockname_args /* {
966 		syscallarg(int)			fdes;
967 		syscallarg(struct sockaddr *)	asa;
968 		syscallarg(unsigned int *)	alen;
969 	} */ *uap = v;
970 	struct proc	*p;
971 	struct file	*fp;
972 	struct socket	*so;
973 	struct mbuf	*m;
974 	unsigned int	len;
975 	int		error;
976 
977 	p = l->l_proc;
978 	/* getsock() will use the descriptor for us */
979 	if ((error = getsock(p->p_fd, SCARG(uap, fdes), &fp)) != 0)
980 		return (error);
981 	error = copyin((caddr_t)SCARG(uap, alen), (caddr_t)&len, sizeof(len));
982 	if (error)
983 		goto out;
984 	so = (struct socket *)fp->f_data;
985 	m = m_getclr(M_WAIT, MT_SONAME);
986 	MCLAIM(m, so->so_mowner);
987 	error = (*so->so_proto->pr_usrreq)(so, PRU_SOCKADDR, (struct mbuf *)0,
988 	    m, (struct mbuf *)0, (struct proc *)0);
989 	if (error)
990 		goto bad;
991 	if (len > m->m_len)
992 		len = m->m_len;
993 	error = copyout(mtod(m, caddr_t), (caddr_t)SCARG(uap, asa), len);
994 	if (error == 0)
995 		error = copyout((caddr_t)&len, (caddr_t)SCARG(uap, alen),
996 		    sizeof(len));
997  bad:
998 	m_freem(m);
999  out:
1000 	FILE_UNUSE(fp, p);
1001 	return (error);
1002 }
1003 
1004 /*
1005  * Get name of peer for connected socket.
1006  */
1007 /* ARGSUSED */
1008 int
1009 sys_getpeername(struct lwp *l, void *v, register_t *retval)
1010 {
1011 	struct sys_getpeername_args /* {
1012 		syscallarg(int)			fdes;
1013 		syscallarg(struct sockaddr *)	asa;
1014 		syscallarg(unsigned int *)	alen;
1015 	} */ *uap = v;
1016 	struct proc	*p;
1017 	struct file	*fp;
1018 	struct socket	*so;
1019 	struct mbuf	*m;
1020 	unsigned int	len;
1021 	int		error;
1022 
1023 	p = l->l_proc;
1024 	/* getsock() will use the descriptor for us */
1025 	if ((error = getsock(p->p_fd, SCARG(uap, fdes), &fp)) != 0)
1026 		return (error);
1027 	so = (struct socket *)fp->f_data;
1028 	if ((so->so_state & (SS_ISCONNECTED|SS_ISCONFIRMING)) == 0) {
1029 		error = ENOTCONN;
1030 		goto out;
1031 	}
1032 	error = copyin((caddr_t)SCARG(uap, alen), (caddr_t)&len, sizeof(len));
1033 	if (error)
1034 		goto out;
1035 	m = m_getclr(M_WAIT, MT_SONAME);
1036 	MCLAIM(m, so->so_mowner);
1037 	error = (*so->so_proto->pr_usrreq)(so, PRU_PEERADDR, (struct mbuf *)0,
1038 	    m, (struct mbuf *)0, (struct proc *)0);
1039 	if (error)
1040 		goto bad;
1041 	if (len > m->m_len)
1042 		len = m->m_len;
1043 	error = copyout(mtod(m, caddr_t), (caddr_t)SCARG(uap, asa), len);
1044 	if (error)
1045 		goto bad;
1046 	error = copyout((caddr_t)&len, (caddr_t)SCARG(uap, alen), sizeof(len));
1047  bad:
1048 	m_freem(m);
1049  out:
1050 	FILE_UNUSE(fp, p);
1051 	return (error);
1052 }
1053 
1054 /*
1055  * XXX In a perfect world, we wouldn't pass around socket control
1056  * XXX arguments in mbufs, and this could go away.
1057  */
1058 int
1059 sockargs(struct mbuf **mp, const void *buf, size_t buflen, int type)
1060 {
1061 	struct sockaddr	*sa;
1062 	struct mbuf	*m;
1063 	int		error;
1064 
1065 	/*
1066 	 * We can't allow socket names > UCHAR_MAX in length, since that
1067 	 * will overflow sa_len.  Control data more than a page size in
1068 	 * length is just too much.
1069 	 */
1070 	if (buflen > (type == MT_SONAME ? UCHAR_MAX : PAGE_SIZE))
1071 		return (EINVAL);
1072 
1073 	/* Allocate an mbuf to hold the arguments. */
1074 	m = m_get(M_WAIT, type);
1075 	/* can't claim.  don't who to assign it to. */
1076 	if (buflen > MLEN) {
1077 		/*
1078 		 * Won't fit into a regular mbuf, so we allocate just
1079 		 * enough external storage to hold the argument.
1080 		 */
1081 		MEXTMALLOC(m, buflen, M_WAITOK);
1082 	}
1083 	m->m_len = buflen;
1084 	error = copyin(buf, mtod(m, caddr_t), buflen);
1085 	if (error) {
1086 		(void) m_free(m);
1087 		return (error);
1088 	}
1089 	*mp = m;
1090 	if (type == MT_SONAME) {
1091 		sa = mtod(m, struct sockaddr *);
1092 #if BYTE_ORDER != BIG_ENDIAN
1093 		/*
1094 		 * 4.3BSD compat thing - need to stay, since bind(2),
1095 		 * connect(2), sendto(2) were not versioned for COMPAT_43.
1096 		 */
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