xref: /netbsd-src/sys/kern/uipc_syscalls.c (revision ce2c90c7c172d95d2402a5b3d96d8f8e6d138a21)
1 /*	$NetBSD: uipc_syscalls.c,v 1.103 2006/10/12 01:32:19 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. 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 1.103 2006/10/12 01:32:19 christos 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 __unused)
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 __unused)
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 __unused)
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 __unused)
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 			psignal(p, SIGPIPE);
542 	}
543 	if (error == 0)
544 		*retsize = len - auio.uio_resid;
545 #ifdef KTRACE
546 	if (ktriov != NULL) {
547 		if (error == 0)
548 			ktrgenio(l, s, UIO_WRITE, ktriov, *retsize, error);
549 		free(ktriov, M_TEMP);
550 	}
551 #endif
552  bad:
553 	if (to)
554 		m_freem(to);
555  out:
556 	FILE_UNUSE(fp, l);
557 	return (error);
558 }
559 
560 int
561 sys_recvfrom(struct lwp *l, void *v, register_t *retval)
562 {
563 	struct sys_recvfrom_args /* {
564 		syscallarg(int)			s;
565 		syscallarg(void *)		buf;
566 		syscallarg(size_t)		len;
567 		syscallarg(int)			flags;
568 		syscallarg(struct sockaddr *)	from;
569 		syscallarg(unsigned int *)	fromlenaddr;
570 	} */ *uap = v;
571 	struct msghdr	msg;
572 	struct iovec	aiov;
573 	int		error;
574 
575 	if (SCARG(uap, fromlenaddr)) {
576 		error = copyin((caddr_t)SCARG(uap, fromlenaddr),
577 			       (caddr_t)&msg.msg_namelen,
578 			       sizeof(msg.msg_namelen));
579 		if (error)
580 			return (error);
581 	} else
582 		msg.msg_namelen = 0;
583 	msg.msg_name = (caddr_t)SCARG(uap, from);
584 	msg.msg_iov = &aiov;
585 	msg.msg_iovlen = 1;
586 	aiov.iov_base = SCARG(uap, buf);
587 	aiov.iov_len = SCARG(uap, len);
588 	msg.msg_control = 0;
589 	msg.msg_flags = SCARG(uap, flags);
590 	return (recvit(l, SCARG(uap, s), &msg,
591 		       (caddr_t)SCARG(uap, fromlenaddr), retval));
592 }
593 
594 int
595 sys_recvmsg(struct lwp *l, void *v, register_t *retval)
596 {
597 	struct sys_recvmsg_args /* {
598 		syscallarg(int)			s;
599 		syscallarg(struct msghdr *)	msg;
600 		syscallarg(int)			flags;
601 	} */ *uap = v;
602 	struct iovec	aiov[UIO_SMALLIOV], *uiov, *iov;
603 	struct msghdr	msg;
604 	int		error;
605 
606 	error = copyin((caddr_t)SCARG(uap, msg), (caddr_t)&msg,
607 		       sizeof(msg));
608 	if (error)
609 		return (error);
610 	if ((unsigned int)msg.msg_iovlen > UIO_SMALLIOV) {
611 		if ((unsigned int)msg.msg_iovlen > IOV_MAX)
612 			return (EMSGSIZE);
613 		iov = malloc(sizeof(struct iovec) * msg.msg_iovlen,
614 		    M_IOV, M_WAITOK);
615 	} else
616 		iov = aiov;
617 	if ((unsigned int)msg.msg_iovlen > 0) {
618 		error = copyin((caddr_t)msg.msg_iov, (caddr_t)iov,
619 		    (size_t)(msg.msg_iovlen * sizeof(struct iovec)));
620 		if (error)
621 			goto done;
622 	}
623 	uiov = msg.msg_iov;
624 	msg.msg_iov = iov;
625 	msg.msg_flags = SCARG(uap, flags);
626 	if ((error = recvit(l, SCARG(uap, s), &msg, (caddr_t)0, retval)) == 0) {
627 		msg.msg_iov = uiov;
628 		error = copyout((caddr_t)&msg, (caddr_t)SCARG(uap, msg),
629 		    sizeof(msg));
630 	}
631 done:
632 	if (iov != aiov)
633 		free(iov, M_IOV);
634 	return (error);
635 }
636 
637 /*
638  * Adjust for a truncated SCM_RIGHTS control message.  This means
639  *  closing any file descriptors that aren't entirely present in the
640  *  returned buffer.  m is the mbuf holding the (already externalized)
641  *  SCM_RIGHTS message; len is the length it is being truncated to.  p
642  *  is the affected process.
643  */
644 static void
645 adjust_rights(struct mbuf *m, int len, struct lwp *l)
646 {
647 	int nfd;
648 	int i;
649 	int nok;
650 	int *fdv;
651 
652 	nfd = m->m_len < CMSG_SPACE(sizeof(int)) ? 0
653 	    : (m->m_len - CMSG_SPACE(sizeof(int))) / sizeof(int) + 1;
654 	nok = (len < CMSG_LEN(0)) ? 0 : ((len - CMSG_LEN(0)) / sizeof(int));
655 	fdv = (int *) CMSG_DATA(mtod(m,struct cmsghdr *));
656 	for (i = nok; i < nfd; i++)
657 		fdrelease(l,fdv[i]);
658 }
659 
660 int
661 recvit(struct lwp *l, int s, struct msghdr *mp, caddr_t namelenp,
662 	register_t *retsize)
663 {
664 	struct proc	*p;
665 	struct file	*fp;
666 	struct uio	auio;
667 	struct iovec	*iov;
668 	int		i, len, error;
669 	struct mbuf	*from, *control;
670 	struct socket	*so;
671 #ifdef KTRACE
672 	struct iovec	*ktriov;
673 #endif
674 
675 	p = l->l_proc;
676 	from = 0;
677 	control = 0;
678 #ifdef KTRACE
679 	ktriov = NULL;
680 #endif
681 
682 	/* getsock() will use the descriptor for us */
683 	if ((error = getsock(p->p_fd, s, &fp)) != 0)
684 		return (error);
685 	so = (struct socket *)fp->f_data;
686 	auio.uio_iov = mp->msg_iov;
687 	auio.uio_iovcnt = mp->msg_iovlen;
688 	auio.uio_rw = UIO_READ;
689 	auio.uio_offset = 0;			/* XXX */
690 	auio.uio_resid = 0;
691 	KASSERT(l == curlwp);
692 	auio.uio_vmspace = l->l_proc->p_vmspace;
693 	iov = mp->msg_iov;
694 	for (i = 0; i < mp->msg_iovlen; i++, iov++) {
695 #if 0
696 		/* cannot happen iov_len is unsigned */
697 		if (iov->iov_len < 0) {
698 			error = EINVAL;
699 			goto out1;
700 		}
701 #endif
702 		/*
703 		 * Reads return ssize_t because -1 is returned on error.
704 		 * Therefore we must restrict the length to SSIZE_MAX to
705 		 * avoid garbage return values.
706 		 */
707 		auio.uio_resid += iov->iov_len;
708 		if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
709 			error = EINVAL;
710 			goto out1;
711 		}
712 	}
713 #ifdef KTRACE
714 	if (KTRPOINT(p, KTR_GENIO)) {
715 		int iovlen = auio.uio_iovcnt * sizeof(struct iovec);
716 
717 		ktriov = malloc(iovlen, M_TEMP, M_WAITOK);
718 		memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen);
719 	}
720 #endif
721 	len = auio.uio_resid;
722 	error = (*so->so_receive)(so, &from, &auio, NULL,
723 			  mp->msg_control ? &control : NULL, &mp->msg_flags);
724 	if (error) {
725 		if (auio.uio_resid != len && (error == ERESTART ||
726 		    error == EINTR || error == EWOULDBLOCK))
727 			error = 0;
728 	}
729 #ifdef KTRACE
730 	if (ktriov != NULL) {
731 		if (error == 0)
732 			ktrgenio(l, s, UIO_READ, ktriov,
733 			    len - auio.uio_resid, error);
734 		free(ktriov, M_TEMP);
735 	}
736 #endif
737 	if (error)
738 		goto out;
739 	*retsize = len - auio.uio_resid;
740 	if (mp->msg_name) {
741 		len = mp->msg_namelen;
742 		if (len <= 0 || from == 0)
743 			len = 0;
744 		else {
745 			if (len > from->m_len)
746 				len = from->m_len;
747 			/* else if len < from->m_len ??? */
748 			error = copyout(mtod(from, caddr_t),
749 					(caddr_t)mp->msg_name, (unsigned)len);
750 			if (error)
751 				goto out;
752 		}
753 		mp->msg_namelen = len;
754 		if (namelenp &&
755 		    (error = copyout((caddr_t)&len, namelenp, sizeof(int))))
756 			goto out;
757 	}
758 	if (mp->msg_control) {
759 		len = mp->msg_controllen;
760 		if (len <= 0 || control == 0)
761 			len = 0;
762 		else {
763 			struct mbuf *m = control;
764 			caddr_t q = (caddr_t)mp->msg_control;
765 
766 			do {
767 				i = m->m_len;
768 				if (len < i) {
769 					mp->msg_flags |= MSG_CTRUNC;
770 					i = len;
771 					if (mtod(m, struct cmsghdr *)->
772 					    cmsg_type == SCM_RIGHTS)
773 						adjust_rights(m, len, l);
774 				}
775 				error = copyout(mtod(m, caddr_t), q,
776 				    (unsigned)i);
777 				m = m->m_next;
778 				if (m)
779 					i = ALIGN(i);
780 				q += i;
781 				len -= i;
782 				if (error != 0 || len <= 0)
783 					break;
784 			} while (m != NULL);
785 			while (m) {
786 				if (mtod(m, struct cmsghdr *)->
787 				    cmsg_type == SCM_RIGHTS)
788 					adjust_rights(m, 0, l);
789 				m = m->m_next;
790 			}
791 			len = q - (caddr_t)mp->msg_control;
792 		}
793 		mp->msg_controllen = len;
794 	}
795  out:
796 	if (from)
797 		m_freem(from);
798 	if (control)
799 		m_freem(control);
800  out1:
801 	FILE_UNUSE(fp, l);
802 	return (error);
803 }
804 
805 /* ARGSUSED */
806 int
807 sys_shutdown(struct lwp *l, void *v, register_t *retval __unused)
808 {
809 	struct sys_shutdown_args /* {
810 		syscallarg(int)	s;
811 		syscallarg(int)	how;
812 	} */ *uap = v;
813 	struct proc	*p;
814 	struct file	*fp;
815 	int		error;
816 
817 	p = l->l_proc;
818 	/* getsock() will use the descriptor for us */
819 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
820 		return (error);
821 	error = soshutdown((struct socket *)fp->f_data, SCARG(uap, how));
822 	FILE_UNUSE(fp, l);
823 	return (error);
824 }
825 
826 /* ARGSUSED */
827 int
828 sys_setsockopt(struct lwp *l, void *v, register_t *retval __unused)
829 {
830 	struct sys_setsockopt_args /* {
831 		syscallarg(int)			s;
832 		syscallarg(int)			level;
833 		syscallarg(int)			name;
834 		syscallarg(const void *)	val;
835 		syscallarg(unsigned int)	valsize;
836 	} */ *uap = v;
837 	struct proc	*p;
838 	struct file	*fp;
839 	struct mbuf	*m;
840 	struct socket	*so;
841 	int		error;
842 	unsigned int	len;
843 
844 	p = l->l_proc;
845 	m = NULL;
846 	/* getsock() will use the descriptor for us */
847 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
848 		return (error);
849 	so = (struct socket *)fp->f_data;
850 	len = SCARG(uap, valsize);
851 	if (len > MCLBYTES) {
852 		error = EINVAL;
853 		goto out;
854 	}
855 	if (SCARG(uap, val)) {
856 		m = m_get(M_WAIT, MT_SOOPTS);
857 		MCLAIM(m, so->so_mowner);
858 		if (len > MLEN)
859 			m_clget(m, M_WAIT);
860 		error = copyin(SCARG(uap, val), mtod(m, caddr_t), len);
861 		if (error) {
862 			(void) m_free(m);
863 			goto out;
864 		}
865 		m->m_len = SCARG(uap, valsize);
866 	}
867 	error = sosetopt(so, SCARG(uap, level), SCARG(uap, name), m);
868  out:
869 	FILE_UNUSE(fp, l);
870 	return (error);
871 }
872 
873 /* ARGSUSED */
874 int
875 sys_getsockopt(struct lwp *l, void *v, register_t *retval __unused)
876 {
877 	struct sys_getsockopt_args /* {
878 		syscallarg(int)			s;
879 		syscallarg(int)			level;
880 		syscallarg(int)			name;
881 		syscallarg(void *)		val;
882 		syscallarg(unsigned int *)	avalsize;
883 	} */ *uap = v;
884 	struct file	*fp;
885 	struct mbuf	*m;
886 	unsigned int	op, i, valsize;
887 	int		error;
888 
889 	m = NULL;
890 	/* getsock() will use the descriptor for us */
891 	if ((error = getsock(l->l_proc->p_fd, SCARG(uap, s), &fp)) != 0)
892 		return (error);
893 	if (SCARG(uap, val)) {
894 		error = copyin((caddr_t)SCARG(uap, avalsize),
895 			       (caddr_t)&valsize, sizeof(valsize));
896 		if (error)
897 			goto out;
898 	} else
899 		valsize = 0;
900 	if ((error = sogetopt((struct socket *)fp->f_data, SCARG(uap, level),
901 	    SCARG(uap, name), &m)) == 0 && SCARG(uap, val) && valsize &&
902 	    m != NULL) {
903 		op = 0;
904 		while (m && !error && op < valsize) {
905 			i = min(m->m_len, (valsize - op));
906 			error = copyout(mtod(m, caddr_t), SCARG(uap, val), i);
907 			op += i;
908 			SCARG(uap, val) = ((uint8_t *)SCARG(uap, val)) + i;
909 			m = m_free(m);
910 		}
911 		valsize = op;
912 		if (error == 0)
913 			error = copyout(&valsize,
914 					SCARG(uap, avalsize), sizeof(valsize));
915 	}
916 	if (m != NULL)
917 		(void) m_freem(m);
918  out:
919 	FILE_UNUSE(fp, l);
920 	return (error);
921 }
922 
923 #ifdef PIPE_SOCKETPAIR
924 /* ARGSUSED */
925 int
926 sys_pipe(struct lwp *l, void *v __unused, register_t *retval)
927 {
928 	struct filedesc	*fdp;
929 	struct file	*rf, *wf;
930 	struct socket	*rso, *wso;
931 	int		fd, error;
932 
933 	fdp = l->l_proc->p_fd;
934 	if ((error = socreate(AF_LOCAL, &rso, SOCK_STREAM, 0, l)) != 0)
935 		return (error);
936 	if ((error = socreate(AF_LOCAL, &wso, SOCK_STREAM, 0, l)) != 0)
937 		goto free1;
938 	/* remember this socket pair implements a pipe */
939 	wso->so_state |= SS_ISAPIPE;
940 	rso->so_state |= SS_ISAPIPE;
941 	/* falloc() will use the descriptor for us */
942 	if ((error = falloc(l, &rf, &fd)) != 0)
943 		goto free2;
944 	retval[0] = fd;
945 	rf->f_flag = FREAD;
946 	rf->f_type = DTYPE_SOCKET;
947 	rf->f_ops = &socketops;
948 	rf->f_data = (caddr_t)rso;
949 	if ((error = falloc(l, &wf, &fd)) != 0)
950 		goto free3;
951 	wf->f_flag = FWRITE;
952 	wf->f_type = DTYPE_SOCKET;
953 	wf->f_ops = &socketops;
954 	wf->f_data = (caddr_t)wso;
955 	retval[1] = fd;
956 	if ((error = unp_connect2(wso, rso, PRU_CONNECT2)) != 0)
957 		goto free4;
958 	FILE_SET_MATURE(rf);
959 	FILE_SET_MATURE(wf);
960 	FILE_UNUSE(rf, l);
961 	FILE_UNUSE(wf, l);
962 	return (0);
963  free4:
964 	FILE_UNUSE(wf, l);
965 	ffree(wf);
966 	fdremove(fdp, retval[1]);
967  free3:
968 	FILE_UNUSE(rf, l);
969 	ffree(rf);
970 	fdremove(fdp, retval[0]);
971  free2:
972 	(void)soclose(wso);
973  free1:
974 	(void)soclose(rso);
975 	return (error);
976 }
977 #endif /* PIPE_SOCKETPAIR */
978 
979 /*
980  * Get socket name.
981  */
982 /* ARGSUSED */
983 int
984 sys_getsockname(struct lwp *l, void *v, register_t *retval __unused)
985 {
986 	struct sys_getsockname_args /* {
987 		syscallarg(int)			fdes;
988 		syscallarg(struct sockaddr *)	asa;
989 		syscallarg(unsigned int *)	alen;
990 	} */ *uap = v;
991 	struct proc	*p;
992 	struct file	*fp;
993 	struct socket	*so;
994 	struct mbuf	*m;
995 	unsigned int	len;
996 	int		error;
997 
998 	p = l->l_proc;
999 	/* getsock() will use the descriptor for us */
1000 	if ((error = getsock(p->p_fd, SCARG(uap, fdes), &fp)) != 0)
1001 		return (error);
1002 	error = copyin((caddr_t)SCARG(uap, alen), (caddr_t)&len, sizeof(len));
1003 	if (error)
1004 		goto out;
1005 	so = (struct socket *)fp->f_data;
1006 	m = m_getclr(M_WAIT, MT_SONAME);
1007 	MCLAIM(m, so->so_mowner);
1008 	error = (*so->so_proto->pr_usrreq)(so, PRU_SOCKADDR, (struct mbuf *)0,
1009 	    m, (struct mbuf *)0, (struct lwp *)0);
1010 	if (error)
1011 		goto bad;
1012 	if (len > m->m_len)
1013 		len = m->m_len;
1014 	error = copyout(mtod(m, caddr_t), (caddr_t)SCARG(uap, asa), len);
1015 	if (error == 0)
1016 		error = copyout((caddr_t)&len, (caddr_t)SCARG(uap, alen),
1017 		    sizeof(len));
1018  bad:
1019 	m_freem(m);
1020  out:
1021 	FILE_UNUSE(fp, l);
1022 	return (error);
1023 }
1024 
1025 /*
1026  * Get name of peer for connected socket.
1027  */
1028 /* ARGSUSED */
1029 int
1030 sys_getpeername(struct lwp *l, void *v, register_t *retval __unused)
1031 {
1032 	struct sys_getpeername_args /* {
1033 		syscallarg(int)			fdes;
1034 		syscallarg(struct sockaddr *)	asa;
1035 		syscallarg(unsigned int *)	alen;
1036 	} */ *uap = v;
1037 	struct proc	*p;
1038 	struct file	*fp;
1039 	struct socket	*so;
1040 	struct mbuf	*m;
1041 	unsigned int	len;
1042 	int		error;
1043 
1044 	p = l->l_proc;
1045 	/* getsock() will use the descriptor for us */
1046 	if ((error = getsock(p->p_fd, SCARG(uap, fdes), &fp)) != 0)
1047 		return (error);
1048 	so = (struct socket *)fp->f_data;
1049 	if ((so->so_state & (SS_ISCONNECTED|SS_ISCONFIRMING)) == 0) {
1050 		error = ENOTCONN;
1051 		goto out;
1052 	}
1053 	error = copyin((caddr_t)SCARG(uap, alen), (caddr_t)&len, sizeof(len));
1054 	if (error)
1055 		goto out;
1056 	m = m_getclr(M_WAIT, MT_SONAME);
1057 	MCLAIM(m, so->so_mowner);
1058 	error = (*so->so_proto->pr_usrreq)(so, PRU_PEERADDR, (struct mbuf *)0,
1059 	    m, (struct mbuf *)0, (struct lwp *)0);
1060 	if (error)
1061 		goto bad;
1062 	if (len > m->m_len)
1063 		len = m->m_len;
1064 	error = copyout(mtod(m, caddr_t), (caddr_t)SCARG(uap, asa), len);
1065 	if (error)
1066 		goto bad;
1067 	error = copyout((caddr_t)&len, (caddr_t)SCARG(uap, alen), sizeof(len));
1068  bad:
1069 	m_freem(m);
1070  out:
1071 	FILE_UNUSE(fp, l);
1072 	return (error);
1073 }
1074 
1075 /*
1076  * XXX In a perfect world, we wouldn't pass around socket control
1077  * XXX arguments in mbufs, and this could go away.
1078  */
1079 int
1080 sockargs(struct mbuf **mp, const void *bf, size_t buflen, int type)
1081 {
1082 	struct sockaddr	*sa;
1083 	struct mbuf	*m;
1084 	int		error;
1085 
1086 	/*
1087 	 * We can't allow socket names > UCHAR_MAX in length, since that
1088 	 * will overflow sa_len.  Control data more than a page size in
1089 	 * length is just too much.
1090 	 */
1091 	if (buflen > (type == MT_SONAME ? UCHAR_MAX : PAGE_SIZE))
1092 		return (EINVAL);
1093 
1094 	/* Allocate an mbuf to hold the arguments. */
1095 	m = m_get(M_WAIT, type);
1096 	/* can't claim.  don't who to assign it to. */
1097 	if (buflen > MLEN) {
1098 		/*
1099 		 * Won't fit into a regular mbuf, so we allocate just
1100 		 * enough external storage to hold the argument.
1101 		 */
1102 		MEXTMALLOC(m, buflen, M_WAITOK);
1103 	}
1104 	m->m_len = buflen;
1105 	error = copyin(bf, mtod(m, caddr_t), buflen);
1106 	if (error) {
1107 		(void) m_free(m);
1108 		return (error);
1109 	}
1110 	*mp = m;
1111 	if (type == MT_SONAME) {
1112 		sa = mtod(m, struct sockaddr *);
1113 #if BYTE_ORDER != BIG_ENDIAN
1114 		/*
1115 		 * 4.3BSD compat thing - need to stay, since bind(2),
1116 		 * connect(2), sendto(2) were not versioned for COMPAT_43.
1117 		 */
1118 		if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
1119 			sa->sa_family = sa->sa_len;
1120 #endif
1121 		sa->sa_len = buflen;
1122 	}
1123 	return (0);
1124 }
1125 
1126 int
1127 getsock(struct filedesc *fdp, int fdes, struct file **fpp)
1128 {
1129 	struct file	*fp;
1130 
1131 	if ((fp = fd_getfile(fdp, fdes)) == NULL)
1132 		return (EBADF);
1133 
1134 	FILE_USE(fp);
1135 
1136 	if (fp->f_type != DTYPE_SOCKET) {
1137 		FILE_UNUSE(fp, NULL);
1138 		return (ENOTSOCK);
1139 	}
1140 	*fpp = fp;
1141 	return (0);
1142 }
1143