xref: /netbsd-src/sys/kern/uipc_syscalls.c (revision e5548b402ae4c44fb816de42c7bba9581ce23ef5)
1 /*	$NetBSD: uipc_syscalls.c,v 1.95 2005/12/11 12:24:30 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 1.95 2005/12/11 12:24:30 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_socket(struct lwp *l, void *v, register_t *retval)
73 {
74 	struct sys_socket_args /* {
75 		syscallarg(int)	domain;
76 		syscallarg(int)	type;
77 		syscallarg(int)	protocol;
78 	} */ *uap = v;
79 
80 	struct proc	*p;
81 	struct filedesc	*fdp;
82 	struct socket	*so;
83 	struct file	*fp;
84 	int		fd, error;
85 
86 	p = l->l_proc;
87 	fdp = p->p_fd;
88 	/* falloc() will use the desciptor for us */
89 	if ((error = falloc(p, &fp, &fd)) != 0)
90 		return (error);
91 	fp->f_flag = FREAD|FWRITE;
92 	fp->f_type = DTYPE_SOCKET;
93 	fp->f_ops = &socketops;
94 	error = socreate(SCARG(uap, domain), &so, SCARG(uap, type),
95 			 SCARG(uap, protocol), l);
96 	if (error) {
97 		FILE_UNUSE(fp, l);
98 		fdremove(fdp, fd);
99 		ffree(fp);
100 	} else {
101 		fp->f_data = (caddr_t)so;
102 		FILE_SET_MATURE(fp);
103 		FILE_UNUSE(fp, l);
104 		*retval = fd;
105 	}
106 	return (error);
107 }
108 
109 /* ARGSUSED */
110 int
111 sys_bind(struct lwp *l, void *v, register_t *retval)
112 {
113 	struct sys_bind_args /* {
114 		syscallarg(int)				s;
115 		syscallarg(const struct sockaddr *)	name;
116 		syscallarg(unsigned int)		namelen;
117 	} */ *uap = v;
118 	struct proc	*p;
119 	struct file	*fp;
120 	struct mbuf	*nam;
121 	int		error;
122 
123 	p = l->l_proc;
124 	/* getsock() will use the descriptor for us */
125 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
126 		return (error);
127 	error = sockargs(&nam, SCARG(uap, name), SCARG(uap, namelen),
128 	    MT_SONAME);
129 	if (error) {
130 		FILE_UNUSE(fp, l);
131 		return (error);
132 	}
133 	MCLAIM(nam, ((struct socket *)fp->f_data)->so_mowner);
134 	error = sobind((struct socket *)fp->f_data, nam, l);
135 	m_freem(nam);
136 	FILE_UNUSE(fp, l);
137 	return (error);
138 }
139 
140 /* ARGSUSED */
141 int
142 sys_listen(struct lwp *l, void *v, register_t *retval)
143 {
144 	struct sys_listen_args /* {
145 		syscallarg(int)	s;
146 		syscallarg(int)	backlog;
147 	} */ *uap = v;
148 	struct proc	*p;
149 	struct file	*fp;
150 	int		error;
151 
152 	p = l->l_proc;
153 	/* getsock() will use the descriptor for us */
154 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
155 		return (error);
156 	error = solisten((struct socket *)fp->f_data, SCARG(uap, backlog));
157 	FILE_UNUSE(fp, l);
158 	return (error);
159 }
160 
161 int
162 sys_accept(struct lwp *l, void *v, register_t *retval)
163 {
164 	struct sys_accept_args /* {
165 		syscallarg(int)			s;
166 		syscallarg(struct sockaddr *)	name;
167 		syscallarg(unsigned int *)	anamelen;
168 	} */ *uap = v;
169 	struct proc	*p;
170 	struct filedesc	*fdp;
171 	struct file	*fp;
172 	struct mbuf	*nam;
173 	unsigned int	namelen;
174 	int		error, s, fd;
175 	struct socket	*so;
176 	int		fflag;
177 
178 	p = l->l_proc;
179 	fdp = p->p_fd;
180 	if (SCARG(uap, name) && (error = copyin(SCARG(uap, anamelen),
181 	    &namelen, sizeof(namelen))))
182 		return (error);
183 
184 	/* getsock() will use the descriptor for us */
185 	if ((error = getsock(fdp, SCARG(uap, s), &fp)) != 0)
186 		return (error);
187 	s = splsoftnet();
188 	so = (struct socket *)fp->f_data;
189 	FILE_UNUSE(fp, l);
190 	if (!(so->so_proto->pr_flags & PR_LISTEN)) {
191 		splx(s);
192 		return (EOPNOTSUPP);
193 	}
194 	if ((so->so_options & SO_ACCEPTCONN) == 0) {
195 		splx(s);
196 		return (EINVAL);
197 	}
198 	if ((so->so_state & SS_NBIO) && so->so_qlen == 0) {
199 		splx(s);
200 		return (EWOULDBLOCK);
201 	}
202 	while (so->so_qlen == 0 && so->so_error == 0) {
203 		if (so->so_state & SS_CANTRCVMORE) {
204 			so->so_error = ECONNABORTED;
205 			break;
206 		}
207 		error = tsleep((caddr_t)&so->so_timeo, PSOCK | PCATCH,
208 			       netcon, 0);
209 		if (error) {
210 			splx(s);
211 			return (error);
212 		}
213 	}
214 	if (so->so_error) {
215 		error = so->so_error;
216 		so->so_error = 0;
217 		splx(s);
218 		return (error);
219 	}
220 	fflag = fp->f_flag;
221 	/* falloc() will use the descriptor for us */
222 	if ((error = falloc(p, &fp, &fd)) != 0) {
223 		splx(s);
224 		return (error);
225 	}
226 	*retval = fd;
227 
228 	/* connection has been removed from the listen queue */
229 	KNOTE(&so->so_rcv.sb_sel.sel_klist, 0);
230 
231 	{ struct socket *aso = TAILQ_FIRST(&so->so_q);
232 	  if (soqremque(aso, 1) == 0)
233 		panic("accept");
234 	  so = aso;
235 	}
236 	fp->f_type = DTYPE_SOCKET;
237 	fp->f_flag = fflag;
238 	fp->f_ops = &socketops;
239 	fp->f_data = (caddr_t)so;
240 	FILE_UNUSE(fp, l);
241 	nam = m_get(M_WAIT, MT_SONAME);
242 	if ((error = soaccept(so, nam)) == 0 && SCARG(uap, name)) {
243 		if (namelen > nam->m_len)
244 			namelen = nam->m_len;
245 		/* SHOULD COPY OUT A CHAIN HERE */
246 		if ((error = copyout(mtod(nam, caddr_t),
247 		    (caddr_t)SCARG(uap, name), namelen)) == 0)
248 			error = copyout((caddr_t)&namelen,
249 			    (caddr_t)SCARG(uap, anamelen),
250 			    sizeof(*SCARG(uap, anamelen)));
251 	}
252 	/* if an error occurred, free the file descriptor */
253 	if (error) {
254 		fdremove(fdp, fd);
255 		ffree(fp);
256 	}
257 	m_freem(nam);
258 	splx(s);
259 	FILE_SET_MATURE(fp);
260 	return (error);
261 }
262 
263 /* ARGSUSED */
264 int
265 sys_connect(struct lwp *l, void *v, register_t *retval)
266 {
267 	struct sys_connect_args /* {
268 		syscallarg(int)				s;
269 		syscallarg(const struct sockaddr *)	name;
270 		syscallarg(unsigned int)		namelen;
271 	} */ *uap = v;
272 	struct proc	*p;
273 	struct file	*fp;
274 	struct socket	*so;
275 	struct mbuf	*nam;
276 	int		error, s;
277 	int		interrupted = 0;
278 
279 	p = l->l_proc;
280 	/* getsock() will use the descriptor for us */
281 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
282 		return (error);
283 	so = (struct socket *)fp->f_data;
284 	if (so->so_state & SS_ISCONNECTING) {
285 		error = EALREADY;
286 		goto out;
287 	}
288 	error = sockargs(&nam, SCARG(uap, name), SCARG(uap, namelen),
289 	    MT_SONAME);
290 	if (error)
291 		goto out;
292 	MCLAIM(nam, so->so_mowner);
293 	error = soconnect(so, nam, l);
294 	if (error)
295 		goto bad;
296 	if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING)) {
297 		m_freem(nam);
298 		error = EINPROGRESS;
299 		goto out;
300 	}
301 	s = splsoftnet();
302 	while ((so->so_state & SS_ISCONNECTING) && so->so_error == 0) {
303 		error = tsleep((caddr_t)&so->so_timeo, PSOCK | PCATCH,
304 			       netcon, 0);
305 		if (error) {
306 			if (error == EINTR || error == ERESTART)
307 				interrupted = 1;
308 			break;
309 		}
310 	}
311 	if (error == 0) {
312 		error = so->so_error;
313 		so->so_error = 0;
314 	}
315 	splx(s);
316  bad:
317 	if (!interrupted)
318 		so->so_state &= ~SS_ISCONNECTING;
319 	m_freem(nam);
320 	if (error == ERESTART)
321 		error = EINTR;
322  out:
323 	FILE_UNUSE(fp, l);
324 	return (error);
325 }
326 
327 int
328 sys_socketpair(struct lwp *l, void *v, register_t *retval)
329 {
330 	struct sys_socketpair_args /* {
331 		syscallarg(int)		domain;
332 		syscallarg(int)		type;
333 		syscallarg(int)		protocol;
334 		syscallarg(int *)	rsv;
335 	} */ *uap = v;
336 	struct proc *p;
337 	struct filedesc	*fdp;
338 	struct file	*fp1, *fp2;
339 	struct socket	*so1, *so2;
340 	int		fd, error, sv[2];
341 
342 	p = l->l_proc;
343 	fdp = p->p_fd;
344 	error = socreate(SCARG(uap, domain), &so1, SCARG(uap, type),
345 			 SCARG(uap, protocol), l);
346 	if (error)
347 		return (error);
348 	error = socreate(SCARG(uap, domain), &so2, SCARG(uap, type),
349 			 SCARG(uap, protocol), l);
350 	if (error)
351 		goto free1;
352 	/* falloc() will use the descriptor for us */
353 	if ((error = falloc(p, &fp1, &fd)) != 0)
354 		goto free2;
355 	sv[0] = fd;
356 	fp1->f_flag = FREAD|FWRITE;
357 	fp1->f_type = DTYPE_SOCKET;
358 	fp1->f_ops = &socketops;
359 	fp1->f_data = (caddr_t)so1;
360 	if ((error = falloc(p, &fp2, &fd)) != 0)
361 		goto free3;
362 	fp2->f_flag = FREAD|FWRITE;
363 	fp2->f_type = DTYPE_SOCKET;
364 	fp2->f_ops = &socketops;
365 	fp2->f_data = (caddr_t)so2;
366 	sv[1] = fd;
367 	if ((error = soconnect2(so1, so2)) != 0)
368 		goto free4;
369 	if (SCARG(uap, type) == SOCK_DGRAM) {
370 		/*
371 		 * Datagram socket connection is asymmetric.
372 		 */
373 		 if ((error = soconnect2(so2, so1)) != 0)
374 			goto free4;
375 	}
376 	error = copyout((caddr_t)sv, (caddr_t)SCARG(uap, rsv),
377 	    2 * sizeof(int));
378 	FILE_SET_MATURE(fp1);
379 	FILE_SET_MATURE(fp2);
380 	FILE_UNUSE(fp1, l);
381 	FILE_UNUSE(fp2, l);
382 	return (error);
383  free4:
384 	FILE_UNUSE(fp2, l);
385 	ffree(fp2);
386 	fdremove(fdp, sv[1]);
387  free3:
388 	FILE_UNUSE(fp1, l);
389 	ffree(fp1);
390 	fdremove(fdp, sv[0]);
391  free2:
392 	(void)soclose(so2);
393  free1:
394 	(void)soclose(so1);
395 	return (error);
396 }
397 
398 int
399 sys_sendto(struct lwp *l, void *v, register_t *retval)
400 {
401 	struct sys_sendto_args /* {
402 		syscallarg(int)				s;
403 		syscallarg(const void *)		buf;
404 		syscallarg(size_t)			len;
405 		syscallarg(int)				flags;
406 		syscallarg(const struct sockaddr *)	to;
407 		syscallarg(unsigned int)		tolen;
408 	} */ *uap = v;
409 	struct msghdr	msg;
410 	struct iovec	aiov;
411 
412 	msg.msg_name = __UNCONST(SCARG(uap, to)); /* XXXUNCONST kills const */
413 	msg.msg_namelen = SCARG(uap, tolen);
414 	msg.msg_iov = &aiov;
415 	msg.msg_iovlen = 1;
416 	msg.msg_control = 0;
417 	msg.msg_flags = 0;
418 	aiov.iov_base = __UNCONST(SCARG(uap, buf)); /* XXXUNCONST kills const */
419 	aiov.iov_len = SCARG(uap, len);
420 	return (sendit(l, SCARG(uap, s), &msg, SCARG(uap, flags), retval));
421 }
422 
423 int
424 sys_sendmsg(struct lwp *l, void *v, register_t *retval)
425 {
426 	struct sys_sendmsg_args /* {
427 		syscallarg(int)				s;
428 		syscallarg(const struct msghdr *)	msg;
429 		syscallarg(int)				flags;
430 	} */ *uap = v;
431 	struct msghdr	msg;
432 	struct iovec	aiov[UIO_SMALLIOV], *iov;
433 	int		error;
434 
435 	error = copyin(SCARG(uap, msg), (caddr_t)&msg, sizeof(msg));
436 	if (error)
437 		return (error);
438 	if ((unsigned int)msg.msg_iovlen > UIO_SMALLIOV) {
439 		if ((unsigned int)msg.msg_iovlen > IOV_MAX)
440 			return (EMSGSIZE);
441 		iov = malloc(sizeof(struct iovec) * msg.msg_iovlen,
442 		    M_IOV, M_WAITOK);
443 	} else
444 		iov = aiov;
445 	if ((unsigned int)msg.msg_iovlen > 0) {
446 		error = copyin((caddr_t)msg.msg_iov, (caddr_t)iov,
447 		    (size_t)(msg.msg_iovlen * sizeof(struct iovec)));
448 		if (error)
449 			goto done;
450 	}
451 	msg.msg_iov = iov;
452 	msg.msg_flags = 0;
453 	error = sendit(l, SCARG(uap, s), &msg, SCARG(uap, flags), retval);
454 done:
455 	if (iov != aiov)
456 		free(iov, M_IOV);
457 	return (error);
458 }
459 
460 int
461 sendit(struct lwp *l, int s, struct msghdr *mp, int flags, register_t *retsize)
462 {
463 	struct proc	*p;
464 	struct file	*fp;
465 	struct uio	auio;
466 	struct iovec	*iov;
467 	int		i, len, error;
468 	struct mbuf	*to, *control;
469 	struct socket	*so;
470 #ifdef KTRACE
471 	struct iovec	*ktriov;
472 #endif
473 
474 #ifdef KTRACE
475 	ktriov = NULL;
476 #endif
477 	p = l->l_proc;
478 	/* getsock() will use the descriptor for us */
479 	if ((error = getsock(p->p_fd, s, &fp)) != 0)
480 		return (error);
481 	so = (struct socket *)fp->f_data;
482 	auio.uio_iov = mp->msg_iov;
483 	auio.uio_iovcnt = mp->msg_iovlen;
484 	auio.uio_segflg = UIO_USERSPACE;
485 	auio.uio_rw = UIO_WRITE;
486 	auio.uio_lwp = l;
487 	auio.uio_offset = 0;			/* XXX */
488 	auio.uio_resid = 0;
489 	iov = mp->msg_iov;
490 	for (i = 0; i < mp->msg_iovlen; i++, iov++) {
491 #if 0
492 		/* cannot happen; iov_len is unsigned */
493 		if (iov->iov_len < 0) {
494 			error = EINVAL;
495 			goto out;
496 		}
497 #endif
498 		/*
499 		 * Writes return ssize_t because -1 is returned on error.
500 		 * Therefore, we must restrict the length to SSIZE_MAX to
501 		 * avoid garbage return values.
502 		 */
503 		auio.uio_resid += iov->iov_len;
504 		if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
505 			error = EINVAL;
506 			goto out;
507 		}
508 	}
509 	if (mp->msg_name) {
510 		error = sockargs(&to, mp->msg_name, mp->msg_namelen,
511 				 MT_SONAME);
512 		if (error)
513 			goto out;
514 		MCLAIM(to, so->so_mowner);
515 	} else
516 		to = 0;
517 	if (mp->msg_control) {
518 		if (mp->msg_controllen < sizeof(struct cmsghdr)) {
519 			error = EINVAL;
520 			goto bad;
521 		}
522 		error = sockargs(&control, mp->msg_control,
523 				 mp->msg_controllen, MT_CONTROL);
524 		if (error)
525 			goto bad;
526 		MCLAIM(control, so->so_mowner);
527 	} else
528 		control = 0;
529 #ifdef KTRACE
530 	if (KTRPOINT(p, KTR_GENIO)) {
531 		int iovlen = auio.uio_iovcnt * sizeof(struct iovec);
532 
533 		ktriov = malloc(iovlen, M_TEMP, M_WAITOK);
534 		memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen);
535 	}
536 #endif
537 	len = auio.uio_resid;
538 	error = (*so->so_send)(so, to, &auio, NULL, control, flags, l);
539 	if (error) {
540 		if (auio.uio_resid != len && (error == ERESTART ||
541 		    error == EINTR || error == EWOULDBLOCK))
542 			error = 0;
543 		if (error == EPIPE)
544 			psignal(p, SIGPIPE);
545 	}
546 	if (error == 0)
547 		*retsize = len - auio.uio_resid;
548 #ifdef KTRACE
549 	if (ktriov != NULL) {
550 		if (error == 0)
551 			ktrgenio(l, s, UIO_WRITE, ktriov, *retsize, error);
552 		free(ktriov, M_TEMP);
553 	}
554 #endif
555  bad:
556 	if (to)
557 		m_freem(to);
558  out:
559 	FILE_UNUSE(fp, l);
560 	return (error);
561 }
562 
563 int
564 sys_recvfrom(struct lwp *l, void *v, register_t *retval)
565 {
566 	struct sys_recvfrom_args /* {
567 		syscallarg(int)			s;
568 		syscallarg(void *)		buf;
569 		syscallarg(size_t)		len;
570 		syscallarg(int)			flags;
571 		syscallarg(struct sockaddr *)	from;
572 		syscallarg(unsigned int *)	fromlenaddr;
573 	} */ *uap = v;
574 	struct msghdr	msg;
575 	struct iovec	aiov;
576 	int		error;
577 
578 	if (SCARG(uap, fromlenaddr)) {
579 		error = copyin((caddr_t)SCARG(uap, fromlenaddr),
580 			       (caddr_t)&msg.msg_namelen,
581 			       sizeof(msg.msg_namelen));
582 		if (error)
583 			return (error);
584 	} else
585 		msg.msg_namelen = 0;
586 	msg.msg_name = (caddr_t)SCARG(uap, from);
587 	msg.msg_iov = &aiov;
588 	msg.msg_iovlen = 1;
589 	aiov.iov_base = SCARG(uap, buf);
590 	aiov.iov_len = SCARG(uap, len);
591 	msg.msg_control = 0;
592 	msg.msg_flags = SCARG(uap, flags);
593 	return (recvit(l, SCARG(uap, s), &msg,
594 		       (caddr_t)SCARG(uap, fromlenaddr), retval));
595 }
596 
597 int
598 sys_recvmsg(struct lwp *l, void *v, register_t *retval)
599 {
600 	struct sys_recvmsg_args /* {
601 		syscallarg(int)			s;
602 		syscallarg(struct msghdr *)	msg;
603 		syscallarg(int)			flags;
604 	} */ *uap = v;
605 	struct iovec	aiov[UIO_SMALLIOV], *uiov, *iov;
606 	struct msghdr	msg;
607 	int		error;
608 
609 	error = copyin((caddr_t)SCARG(uap, msg), (caddr_t)&msg,
610 		       sizeof(msg));
611 	if (error)
612 		return (error);
613 	if ((unsigned int)msg.msg_iovlen > UIO_SMALLIOV) {
614 		if ((unsigned int)msg.msg_iovlen > IOV_MAX)
615 			return (EMSGSIZE);
616 		iov = malloc(sizeof(struct iovec) * msg.msg_iovlen,
617 		    M_IOV, M_WAITOK);
618 	} else
619 		iov = aiov;
620 	if ((unsigned int)msg.msg_iovlen > 0) {
621 		error = copyin((caddr_t)msg.msg_iov, (caddr_t)iov,
622 		    (size_t)(msg.msg_iovlen * sizeof(struct iovec)));
623 		if (error)
624 			goto done;
625 	}
626 	uiov = msg.msg_iov;
627 	msg.msg_iov = iov;
628 	msg.msg_flags = SCARG(uap, flags);
629 	if ((error = recvit(l, SCARG(uap, s), &msg, (caddr_t)0, retval)) == 0) {
630 		msg.msg_iov = uiov;
631 		error = copyout((caddr_t)&msg, (caddr_t)SCARG(uap, msg),
632 		    sizeof(msg));
633 	}
634 done:
635 	if (iov != aiov)
636 		free(iov, M_IOV);
637 	return (error);
638 }
639 
640 /*
641  * Adjust for a truncated SCM_RIGHTS control message.  This means
642  *  closing any file descriptors that aren't entirely present in the
643  *  returned buffer.  m is the mbuf holding the (already externalized)
644  *  SCM_RIGHTS message; len is the length it is being truncated to.  p
645  *  is the affected process.
646  */
647 static void
648 adjust_rights(struct mbuf *m, int len, struct lwp *l)
649 {
650 	int nfd;
651 	int i;
652 	int nok;
653 	int *fdv;
654 
655 	nfd = m->m_len < CMSG_SPACE(sizeof(int)) ? 0
656 	    : (m->m_len - CMSG_SPACE(sizeof(int))) / sizeof(int) + 1;
657 	nok = (len < CMSG_LEN(0)) ? 0 : ((len - CMSG_LEN(0)) / sizeof(int));
658 	fdv = (int *) CMSG_DATA(mtod(m,struct cmsghdr *));
659 	for (i = nok; i < nfd; i++)
660 		fdrelease(l,fdv[i]);
661 }
662 
663 int
664 recvit(struct lwp *l, int s, struct msghdr *mp, caddr_t namelenp,
665 	register_t *retsize)
666 {
667 	struct proc	*p;
668 	struct file	*fp;
669 	struct uio	auio;
670 	struct iovec	*iov;
671 	int		i, len, error;
672 	struct mbuf	*from, *control;
673 	struct socket	*so;
674 #ifdef KTRACE
675 	struct iovec	*ktriov;
676 #endif
677 
678 	p = l->l_proc;
679 	from = 0;
680 	control = 0;
681 #ifdef KTRACE
682 	ktriov = NULL;
683 #endif
684 
685 	/* getsock() will use the descriptor for us */
686 	if ((error = getsock(p->p_fd, s, &fp)) != 0)
687 		return (error);
688 	so = (struct socket *)fp->f_data;
689 	auio.uio_iov = mp->msg_iov;
690 	auio.uio_iovcnt = mp->msg_iovlen;
691 	auio.uio_segflg = UIO_USERSPACE;
692 	auio.uio_rw = UIO_READ;
693 	auio.uio_lwp = l;
694 	auio.uio_offset = 0;			/* XXX */
695 	auio.uio_resid = 0;
696 	iov = mp->msg_iov;
697 	for (i = 0; i < mp->msg_iovlen; i++, iov++) {
698 #if 0
699 		/* cannot happen iov_len is unsigned */
700 		if (iov->iov_len < 0) {
701 			error = EINVAL;
702 			goto out1;
703 		}
704 #endif
705 		/*
706 		 * Reads return ssize_t because -1 is returned on error.
707 		 * Therefore we must restrict the length to SSIZE_MAX to
708 		 * avoid garbage return values.
709 		 */
710 		auio.uio_resid += iov->iov_len;
711 		if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
712 			error = EINVAL;
713 			goto out1;
714 		}
715 	}
716 #ifdef KTRACE
717 	if (KTRPOINT(p, KTR_GENIO)) {
718 		int iovlen = auio.uio_iovcnt * sizeof(struct iovec);
719 
720 		ktriov = malloc(iovlen, M_TEMP, M_WAITOK);
721 		memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen);
722 	}
723 #endif
724 	len = auio.uio_resid;
725 	error = (*so->so_receive)(so, &from, &auio, NULL,
726 			  mp->msg_control ? &control : NULL, &mp->msg_flags);
727 	if (error) {
728 		if (auio.uio_resid != len && (error == ERESTART ||
729 		    error == EINTR || error == EWOULDBLOCK))
730 			error = 0;
731 	}
732 #ifdef KTRACE
733 	if (ktriov != NULL) {
734 		if (error == 0)
735 			ktrgenio(l, s, UIO_READ, ktriov,
736 			    len - auio.uio_resid, error);
737 		free(ktriov, M_TEMP);
738 	}
739 #endif
740 	if (error)
741 		goto out;
742 	*retsize = len - auio.uio_resid;
743 	if (mp->msg_name) {
744 		len = mp->msg_namelen;
745 		if (len <= 0 || from == 0)
746 			len = 0;
747 		else {
748 			if (len > from->m_len)
749 				len = from->m_len;
750 			/* else if len < from->m_len ??? */
751 			error = copyout(mtod(from, caddr_t),
752 					(caddr_t)mp->msg_name, (unsigned)len);
753 			if (error)
754 				goto out;
755 		}
756 		mp->msg_namelen = len;
757 		if (namelenp &&
758 		    (error = copyout((caddr_t)&len, namelenp, sizeof(int))))
759 			goto out;
760 	}
761 	if (mp->msg_control) {
762 		len = mp->msg_controllen;
763 		if (len <= 0 || control == 0)
764 			len = 0;
765 		else {
766 			struct mbuf *m = control;
767 			caddr_t q = (caddr_t)mp->msg_control;
768 
769 			do {
770 				i = m->m_len;
771 				if (len < i) {
772 					mp->msg_flags |= MSG_CTRUNC;
773 					i = len;
774 					if (mtod(m, struct cmsghdr *)->
775 					    cmsg_type == SCM_RIGHTS)
776 						adjust_rights(m, len, l);
777 				}
778 				error = copyout(mtod(m, caddr_t), q,
779 				    (unsigned)i);
780 				m = m->m_next;
781 				if (m)
782 					i = ALIGN(i);
783 				q += i;
784 				len -= i;
785 				if (error != 0 || len <= 0)
786 					break;
787 			} while (m != NULL);
788 			while (m) {
789 				if (mtod(m, struct cmsghdr *)->
790 				    cmsg_type == SCM_RIGHTS)
791 					adjust_rights(m, 0, l);
792 				m = m->m_next;
793 			}
794 			len = q - (caddr_t)mp->msg_control;
795 		}
796 		mp->msg_controllen = len;
797 	}
798  out:
799 	if (from)
800 		m_freem(from);
801 	if (control)
802 		m_freem(control);
803  out1:
804 	FILE_UNUSE(fp, l);
805 	return (error);
806 }
807 
808 /* ARGSUSED */
809 int
810 sys_shutdown(struct lwp *l, void *v, register_t *retval)
811 {
812 	struct sys_shutdown_args /* {
813 		syscallarg(int)	s;
814 		syscallarg(int)	how;
815 	} */ *uap = v;
816 	struct proc	*p;
817 	struct file	*fp;
818 	int		error;
819 
820 	p = l->l_proc;
821 	/* getsock() will use the descriptor for us */
822 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
823 		return (error);
824 	error = soshutdown((struct socket *)fp->f_data, SCARG(uap, how));
825 	FILE_UNUSE(fp, l);
826 	return (error);
827 }
828 
829 /* ARGSUSED */
830 int
831 sys_setsockopt(struct lwp *l, void *v, register_t *retval)
832 {
833 	struct sys_setsockopt_args /* {
834 		syscallarg(int)			s;
835 		syscallarg(int)			level;
836 		syscallarg(int)			name;
837 		syscallarg(const void *)	val;
838 		syscallarg(unsigned int)	valsize;
839 	} */ *uap = v;
840 	struct proc	*p;
841 	struct file	*fp;
842 	struct mbuf	*m;
843 	struct socket	*so;
844 	int		error;
845 	unsigned int	len;
846 
847 	p = l->l_proc;
848 	m = NULL;
849 	/* getsock() will use the descriptor for us */
850 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
851 		return (error);
852 	so = (struct socket *)fp->f_data;
853 	len = SCARG(uap, valsize);
854 	if (len > MCLBYTES) {
855 		error = EINVAL;
856 		goto out;
857 	}
858 	if (SCARG(uap, val)) {
859 		m = m_get(M_WAIT, MT_SOOPTS);
860 		MCLAIM(m, so->so_mowner);
861 		if (len > MLEN)
862 			m_clget(m, M_WAIT);
863 		error = copyin(SCARG(uap, val), mtod(m, caddr_t), len);
864 		if (error) {
865 			(void) m_free(m);
866 			goto out;
867 		}
868 		m->m_len = SCARG(uap, valsize);
869 	}
870 	error = sosetopt(so, SCARG(uap, level), SCARG(uap, name), m);
871  out:
872 	FILE_UNUSE(fp, l);
873 	return (error);
874 }
875 
876 /* ARGSUSED */
877 int
878 sys_getsockopt(struct lwp *l, void *v, register_t *retval)
879 {
880 	struct sys_getsockopt_args /* {
881 		syscallarg(int)			s;
882 		syscallarg(int)			level;
883 		syscallarg(int)			name;
884 		syscallarg(void *)		val;
885 		syscallarg(unsigned int *)	avalsize;
886 	} */ *uap = v;
887 	struct proc	*p;
888 	struct file	*fp;
889 	struct mbuf	*m;
890 	unsigned int	op, i, valsize;
891 	int		error;
892 
893 	p = l->l_proc;
894 	m = NULL;
895 	/* getsock() will use the descriptor for us */
896 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
897 		return (error);
898 	if (SCARG(uap, val)) {
899 		error = copyin((caddr_t)SCARG(uap, avalsize),
900 			       (caddr_t)&valsize, sizeof(valsize));
901 		if (error)
902 			goto out;
903 	} else
904 		valsize = 0;
905 	if ((error = sogetopt((struct socket *)fp->f_data, SCARG(uap, level),
906 	    SCARG(uap, name), &m)) == 0 && SCARG(uap, val) && valsize &&
907 	    m != NULL) {
908 		op = 0;
909 		while (m && !error && op < valsize) {
910 			i = min(m->m_len, (valsize - op));
911 			error = copyout(mtod(m, caddr_t), SCARG(uap, val), i);
912 			op += i;
913 			SCARG(uap, val) = ((u_int8_t *)SCARG(uap, val)) + i;
914 			m = m_free(m);
915 		}
916 		valsize = op;
917 		if (error == 0)
918 			error = copyout(&valsize,
919 					SCARG(uap, avalsize), sizeof(valsize));
920 	}
921 	if (m != NULL)
922 		(void) m_freem(m);
923  out:
924 	FILE_UNUSE(fp, l);
925 	return (error);
926 }
927 
928 #ifdef PIPE_SOCKETPAIR
929 /* ARGSUSED */
930 int
931 sys_pipe(struct lwp *l, void *v, register_t *retval)
932 {
933 	struct proc	*p;
934 	struct filedesc	*fdp;
935 	struct file	*rf, *wf;
936 	struct socket	*rso, *wso;
937 	int		fd, error;
938 
939 	p = l->l_proc;
940 	fdp = p->p_fd;
941 	if ((error = socreate(AF_LOCAL, &rso, SOCK_STREAM, 0, l)) != 0)
942 		return (error);
943 	if ((error = socreate(AF_LOCAL, &wso, SOCK_STREAM, 0, l)) != 0)
944 		goto free1;
945 	/* remember this socket pair implements a pipe */
946 	wso->so_state |= SS_ISAPIPE;
947 	rso->so_state |= SS_ISAPIPE;
948 	/* falloc() will use the descriptor for us */
949 	if ((error = falloc(p, &rf, &fd)) != 0)
950 		goto free2;
951 	retval[0] = fd;
952 	rf->f_flag = FREAD;
953 	rf->f_type = DTYPE_SOCKET;
954 	rf->f_ops = &socketops;
955 	rf->f_data = (caddr_t)rso;
956 	if ((error = falloc(p, &wf, &fd)) != 0)
957 		goto free3;
958 	wf->f_flag = FWRITE;
959 	wf->f_type = DTYPE_SOCKET;
960 	wf->f_ops = &socketops;
961 	wf->f_data = (caddr_t)wso;
962 	retval[1] = fd;
963 	if ((error = unp_connect2(wso, rso, PRU_CONNECT2)) != 0)
964 		goto free4;
965 	FILE_SET_MATURE(rf);
966 	FILE_SET_MATURE(wf);
967 	FILE_UNUSE(rf, l);
968 	FILE_UNUSE(wf, l);
969 	return (0);
970  free4:
971 	FILE_UNUSE(wf, l);
972 	ffree(wf);
973 	fdremove(fdp, retval[1]);
974  free3:
975 	FILE_UNUSE(rf, l);
976 	ffree(rf);
977 	fdremove(fdp, retval[0]);
978  free2:
979 	(void)soclose(wso);
980  free1:
981 	(void)soclose(rso);
982 	return (error);
983 }
984 #endif /* PIPE_SOCKETPAIR */
985 
986 /*
987  * Get socket name.
988  */
989 /* ARGSUSED */
990 int
991 sys_getsockname(struct lwp *l, void *v, register_t *retval)
992 {
993 	struct sys_getsockname_args /* {
994 		syscallarg(int)			fdes;
995 		syscallarg(struct sockaddr *)	asa;
996 		syscallarg(unsigned int *)	alen;
997 	} */ *uap = v;
998 	struct proc	*p;
999 	struct file	*fp;
1000 	struct socket	*so;
1001 	struct mbuf	*m;
1002 	unsigned int	len;
1003 	int		error;
1004 
1005 	p = l->l_proc;
1006 	/* getsock() will use the descriptor for us */
1007 	if ((error = getsock(p->p_fd, SCARG(uap, fdes), &fp)) != 0)
1008 		return (error);
1009 	error = copyin((caddr_t)SCARG(uap, alen), (caddr_t)&len, sizeof(len));
1010 	if (error)
1011 		goto out;
1012 	so = (struct socket *)fp->f_data;
1013 	m = m_getclr(M_WAIT, MT_SONAME);
1014 	MCLAIM(m, so->so_mowner);
1015 	error = (*so->so_proto->pr_usrreq)(so, PRU_SOCKADDR, (struct mbuf *)0,
1016 	    m, (struct mbuf *)0, (struct lwp *)0);
1017 	if (error)
1018 		goto bad;
1019 	if (len > m->m_len)
1020 		len = m->m_len;
1021 	error = copyout(mtod(m, caddr_t), (caddr_t)SCARG(uap, asa), len);
1022 	if (error == 0)
1023 		error = copyout((caddr_t)&len, (caddr_t)SCARG(uap, alen),
1024 		    sizeof(len));
1025  bad:
1026 	m_freem(m);
1027  out:
1028 	FILE_UNUSE(fp, l);
1029 	return (error);
1030 }
1031 
1032 /*
1033  * Get name of peer for connected socket.
1034  */
1035 /* ARGSUSED */
1036 int
1037 sys_getpeername(struct lwp *l, void *v, register_t *retval)
1038 {
1039 	struct sys_getpeername_args /* {
1040 		syscallarg(int)			fdes;
1041 		syscallarg(struct sockaddr *)	asa;
1042 		syscallarg(unsigned int *)	alen;
1043 	} */ *uap = v;
1044 	struct proc	*p;
1045 	struct file	*fp;
1046 	struct socket	*so;
1047 	struct mbuf	*m;
1048 	unsigned int	len;
1049 	int		error;
1050 
1051 	p = l->l_proc;
1052 	/* getsock() will use the descriptor for us */
1053 	if ((error = getsock(p->p_fd, SCARG(uap, fdes), &fp)) != 0)
1054 		return (error);
1055 	so = (struct socket *)fp->f_data;
1056 	if ((so->so_state & (SS_ISCONNECTED|SS_ISCONFIRMING)) == 0) {
1057 		error = ENOTCONN;
1058 		goto out;
1059 	}
1060 	error = copyin((caddr_t)SCARG(uap, alen), (caddr_t)&len, sizeof(len));
1061 	if (error)
1062 		goto out;
1063 	m = m_getclr(M_WAIT, MT_SONAME);
1064 	MCLAIM(m, so->so_mowner);
1065 	error = (*so->so_proto->pr_usrreq)(so, PRU_PEERADDR, (struct mbuf *)0,
1066 	    m, (struct mbuf *)0, (struct lwp *)0);
1067 	if (error)
1068 		goto bad;
1069 	if (len > m->m_len)
1070 		len = m->m_len;
1071 	error = copyout(mtod(m, caddr_t), (caddr_t)SCARG(uap, asa), len);
1072 	if (error)
1073 		goto bad;
1074 	error = copyout((caddr_t)&len, (caddr_t)SCARG(uap, alen), sizeof(len));
1075  bad:
1076 	m_freem(m);
1077  out:
1078 	FILE_UNUSE(fp, l);
1079 	return (error);
1080 }
1081 
1082 /*
1083  * XXX In a perfect world, we wouldn't pass around socket control
1084  * XXX arguments in mbufs, and this could go away.
1085  */
1086 int
1087 sockargs(struct mbuf **mp, const void *bf, size_t buflen, int type)
1088 {
1089 	struct sockaddr	*sa;
1090 	struct mbuf	*m;
1091 	int		error;
1092 
1093 	/*
1094 	 * We can't allow socket names > UCHAR_MAX in length, since that
1095 	 * will overflow sa_len.  Control data more than a page size in
1096 	 * length is just too much.
1097 	 */
1098 	if (buflen > (type == MT_SONAME ? UCHAR_MAX : PAGE_SIZE))
1099 		return (EINVAL);
1100 
1101 	/* Allocate an mbuf to hold the arguments. */
1102 	m = m_get(M_WAIT, type);
1103 	/* can't claim.  don't who to assign it to. */
1104 	if (buflen > MLEN) {
1105 		/*
1106 		 * Won't fit into a regular mbuf, so we allocate just
1107 		 * enough external storage to hold the argument.
1108 		 */
1109 		MEXTMALLOC(m, buflen, M_WAITOK);
1110 	}
1111 	m->m_len = buflen;
1112 	error = copyin(bf, mtod(m, caddr_t), buflen);
1113 	if (error) {
1114 		(void) m_free(m);
1115 		return (error);
1116 	}
1117 	*mp = m;
1118 	if (type == MT_SONAME) {
1119 		sa = mtod(m, struct sockaddr *);
1120 #if BYTE_ORDER != BIG_ENDIAN
1121 		/*
1122 		 * 4.3BSD compat thing - need to stay, since bind(2),
1123 		 * connect(2), sendto(2) were not versioned for COMPAT_43.
1124 		 */
1125 		if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
1126 			sa->sa_family = sa->sa_len;
1127 #endif
1128 		sa->sa_len = buflen;
1129 	}
1130 	return (0);
1131 }
1132 
1133 int
1134 getsock(struct filedesc *fdp, int fdes, struct file **fpp)
1135 {
1136 	struct file	*fp;
1137 
1138 	if ((fp = fd_getfile(fdp, fdes)) == NULL)
1139 		return (EBADF);
1140 
1141 	FILE_USE(fp);
1142 
1143 	if (fp->f_type != DTYPE_SOCKET) {
1144 		FILE_UNUSE(fp, NULL);
1145 		return (ENOTSOCK);
1146 	}
1147 	*fpp = fp;
1148 	return (0);
1149 }
1150