xref: /netbsd/sys/kern/uipc_syscalls.c (revision c4a72b64)
1 /*	$NetBSD: uipc_syscalls.c,v 1.74 2002/11/26 18:44:36 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 1.74 2002/11/26 18:44:36 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/syscallargs.h>
64 
65 #include <uvm/uvm_extern.h>
66 
67 /*
68  * System call interface to the socket abstraction.
69  */
70 extern	struct fileops socketops;
71 
72 int
73 sys_socket(struct proc *p, void *v, register_t *retval)
74 {
75 	struct sys_socket_args /* {
76 		syscallarg(int)	domain;
77 		syscallarg(int)	type;
78 		syscallarg(int)	protocol;
79 	} */ *uap = v;
80 	struct filedesc	*fdp;
81 	struct socket	*so;
82 	struct file	*fp;
83 	int		fd, error;
84 
85 	fdp = p->p_fd;
86 	/* falloc() will use the desciptor for us */
87 	if ((error = falloc(p, &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));
94 	if (error) {
95 		FILE_UNUSE(fp, p);
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, p);
102 		*retval = fd;
103 	}
104 	return (error);
105 }
106 
107 /* ARGSUSED */
108 int
109 sys_bind(struct proc *p, void *v, register_t *retval)
110 {
111 	struct sys_bind_args /* {
112 		syscallarg(int)				s;
113 		syscallarg(const struct sockaddr *)	name;
114 		syscallarg(unsigned int)		namelen;
115 	} */ *uap = v;
116 	struct file	*fp;
117 	struct mbuf	*nam;
118 	int		error;
119 
120 	/* getsock() will use the descriptor for us */
121 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
122 		return (error);
123 	error = sockargs(&nam, SCARG(uap, name), SCARG(uap, namelen),
124 	    MT_SONAME);
125 	if (error) {
126 		FILE_UNUSE(fp, p);
127 		return (error);
128 	}
129 	error = sobind((struct socket *)fp->f_data, nam, p);
130 	m_freem(nam);
131 	FILE_UNUSE(fp, p);
132 	return (error);
133 }
134 
135 /* ARGSUSED */
136 int
137 sys_listen(struct proc *p, void *v, register_t *retval)
138 {
139 	struct sys_listen_args /* {
140 		syscallarg(int)	s;
141 		syscallarg(int)	backlog;
142 	} */ *uap = v;
143 	struct file	*fp;
144 	int		error;
145 
146 	/* getsock() will use the descriptor for us */
147 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
148 		return (error);
149 	error = solisten((struct socket *)fp->f_data, SCARG(uap, backlog));
150 	FILE_UNUSE(fp, p);
151 	return (error);
152 }
153 
154 int
155 sys_accept(struct proc *p, void *v, register_t *retval)
156 {
157 	struct sys_accept_args /* {
158 		syscallarg(int)			s;
159 		syscallarg(struct sockaddr *)	name;
160 		syscallarg(unsigned int *)	anamelen;
161 	} */ *uap = v;
162 	struct filedesc	*fdp;
163 	struct file	*fp;
164 	struct mbuf	*nam;
165 	unsigned int	namelen;
166 	int		error, s, fd;
167 	struct socket	*so;
168 
169 	fdp = p->p_fd;
170 	if (SCARG(uap, name) && (error = copyin((caddr_t)SCARG(uap, anamelen),
171 	    (caddr_t)&namelen, sizeof(namelen))))
172 		return (error);
173 	if (SCARG(uap, name) != NULL &&
174 	    uvm_useracc((caddr_t)SCARG(uap, name), sizeof(struct sockaddr),
175 	     B_WRITE) == FALSE)
176 		return (EFAULT);
177 
178 	/* getsock() will use the descriptor for us */
179 	if ((error = getsock(fdp, SCARG(uap, s), &fp)) != 0)
180 		return (error);
181 	s = splsoftnet();
182 	so = (struct socket *)fp->f_data;
183 	FILE_UNUSE(fp, p);
184 	if (!(so->so_proto->pr_flags & PR_LISTEN)) {
185 		splx(s);
186 		return (EOPNOTSUPP);
187 	}
188 	if ((so->so_options & SO_ACCEPTCONN) == 0) {
189 		splx(s);
190 		return (EINVAL);
191 	}
192 	if ((so->so_state & SS_NBIO) && so->so_qlen == 0) {
193 		splx(s);
194 		return (EWOULDBLOCK);
195 	}
196 	while (so->so_qlen == 0 && so->so_error == 0) {
197 		if (so->so_state & SS_CANTRCVMORE) {
198 			so->so_error = ECONNABORTED;
199 			break;
200 		}
201 		error = tsleep((caddr_t)&so->so_timeo, PSOCK | PCATCH,
202 			       netcon, 0);
203 		if (error) {
204 			splx(s);
205 			return (error);
206 		}
207 	}
208 	if (so->so_error) {
209 		error = so->so_error;
210 		so->so_error = 0;
211 		splx(s);
212 		return (error);
213 	}
214 	/* falloc() will use the descriptor for us */
215 	if ((error = falloc(p, &fp, &fd)) != 0) {
216 		splx(s);
217 		return (error);
218 	}
219 	*retval = fd;
220 
221 	/* connection has been removed from the listen queue */
222 	KNOTE(&so->so_rcv.sb_sel.sel_klist, 0);
223 
224 	{ struct socket *aso = TAILQ_FIRST(&so->so_q);
225 	  if (soqremque(aso, 1) == 0)
226 		panic("accept");
227 	  so = aso;
228 	}
229 	fp->f_type = DTYPE_SOCKET;
230 	fp->f_flag = FREAD|FWRITE;
231 	fp->f_ops = &socketops;
232 	fp->f_data = (caddr_t)so;
233 	FILE_UNUSE(fp, p);
234 	nam = m_get(M_WAIT, MT_SONAME);
235 	if ((error = soaccept(so, nam)) == 0 && SCARG(uap, name)) {
236 		if (namelen > nam->m_len)
237 			namelen = nam->m_len;
238 		/* SHOULD COPY OUT A CHAIN HERE */
239 		if ((error = copyout(mtod(nam, caddr_t),
240 		    (caddr_t)SCARG(uap, name), namelen)) == 0)
241 			error = copyout((caddr_t)&namelen,
242 			    (caddr_t)SCARG(uap, anamelen),
243 			    sizeof(*SCARG(uap, anamelen)));
244 	}
245 	/* if an error occurred, free the file descriptor */
246 	if (error) {
247 		fdremove(fdp, fd);
248 		ffree(fp);
249 	}
250 	m_freem(nam);
251 	splx(s);
252 	FILE_SET_MATURE(fp);
253 	return (error);
254 }
255 
256 /* ARGSUSED */
257 int
258 sys_connect(struct proc *p, void *v, register_t *retval)
259 {
260 	struct sys_connect_args /* {
261 		syscallarg(int)				s;
262 		syscallarg(const struct sockaddr *)	name;
263 		syscallarg(unsigned int)		namelen;
264 	} */ *uap = v;
265 	struct file	*fp;
266 	struct socket	*so;
267 	struct mbuf	*nam;
268 	int		error, s;
269 
270 	/* getsock() will use the descriptor for us */
271 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
272 		return (error);
273 	so = (struct socket *)fp->f_data;
274 	if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING)) {
275 		error = EALREADY;
276 		goto out;
277 	}
278 	error = sockargs(&nam, SCARG(uap, name), SCARG(uap, namelen),
279 	    MT_SONAME);
280 	if (error)
281 		goto out;
282 	error = soconnect(so, nam);
283 	if (error)
284 		goto bad;
285 	if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING)) {
286 		m_freem(nam);
287 		error = EINPROGRESS;
288 		goto out;
289 	}
290 	s = splsoftnet();
291 	while ((so->so_state & SS_ISCONNECTING) && so->so_error == 0) {
292 		error = tsleep((caddr_t)&so->so_timeo, PSOCK | PCATCH,
293 			       netcon, 0);
294 		if (error)
295 			break;
296 	}
297 	if (error == 0) {
298 		error = so->so_error;
299 		so->so_error = 0;
300 	}
301 	splx(s);
302  bad:
303 	so->so_state &= ~SS_ISCONNECTING;
304 	m_freem(nam);
305 	if (error == ERESTART)
306 		error = EINTR;
307  out:
308 	FILE_UNUSE(fp, p);
309 	return (error);
310 }
311 
312 int
313 sys_socketpair(struct proc *p, void *v, register_t *retval)
314 {
315 	struct sys_socketpair_args /* {
316 		syscallarg(int)		domain;
317 		syscallarg(int)		type;
318 		syscallarg(int)		protocol;
319 		syscallarg(int *)	rsv;
320 	} */ *uap = v;
321 	struct filedesc	*fdp;
322 	struct file	*fp1, *fp2;
323 	struct socket	*so1, *so2;
324 	int		fd, error, sv[2];
325 
326 	fdp = p->p_fd;
327 	error = socreate(SCARG(uap, domain), &so1, SCARG(uap, type),
328 			 SCARG(uap, protocol));
329 	if (error)
330 		return (error);
331 	error = socreate(SCARG(uap, domain), &so2, SCARG(uap, type),
332 			 SCARG(uap, protocol));
333 	if (error)
334 		goto free1;
335 	/* falloc() will use the descriptor for us */
336 	if ((error = falloc(p, &fp1, &fd)) != 0)
337 		goto free2;
338 	sv[0] = fd;
339 	fp1->f_flag = FREAD|FWRITE;
340 	fp1->f_type = DTYPE_SOCKET;
341 	fp1->f_ops = &socketops;
342 	fp1->f_data = (caddr_t)so1;
343 	if ((error = falloc(p, &fp2, &fd)) != 0)
344 		goto free3;
345 	fp2->f_flag = FREAD|FWRITE;
346 	fp2->f_type = DTYPE_SOCKET;
347 	fp2->f_ops = &socketops;
348 	fp2->f_data = (caddr_t)so2;
349 	sv[1] = fd;
350 	if ((error = soconnect2(so1, so2)) != 0)
351 		goto free4;
352 	if (SCARG(uap, type) == SOCK_DGRAM) {
353 		/*
354 		 * Datagram socket connection is asymmetric.
355 		 */
356 		 if ((error = soconnect2(so2, so1)) != 0)
357 			goto free4;
358 	}
359 	error = copyout((caddr_t)sv, (caddr_t)SCARG(uap, rsv),
360 	    2 * sizeof(int));
361 	FILE_SET_MATURE(fp1);
362 	FILE_SET_MATURE(fp2);
363 	FILE_UNUSE(fp1, p);
364 	FILE_UNUSE(fp2, p);
365 	return (error);
366  free4:
367 	FILE_UNUSE(fp2, p);
368 	ffree(fp2);
369 	fdremove(fdp, sv[1]);
370  free3:
371 	FILE_UNUSE(fp1, p);
372 	ffree(fp1);
373 	fdremove(fdp, sv[0]);
374  free2:
375 	(void)soclose(so2);
376  free1:
377 	(void)soclose(so1);
378 	return (error);
379 }
380 
381 int
382 sys_sendto(struct proc *p, void *v, register_t *retval)
383 {
384 	struct sys_sendto_args /* {
385 		syscallarg(int)				s;
386 		syscallarg(const void *)		buf;
387 		syscallarg(size_t)			len;
388 		syscallarg(int)				flags;
389 		syscallarg(const struct sockaddr *)	to;
390 		syscallarg(unsigned int)		tolen;
391 	} */ *uap = v;
392 	struct msghdr	msg;
393 	struct iovec	aiov;
394 
395 	msg.msg_name = (caddr_t)SCARG(uap, to);		/* XXX kills const */
396 	msg.msg_namelen = SCARG(uap, tolen);
397 	msg.msg_iov = &aiov;
398 	msg.msg_iovlen = 1;
399 	msg.msg_control = 0;
400 	msg.msg_flags = 0;
401 	aiov.iov_base = (char *)SCARG(uap, buf);	/* XXX kills const */
402 	aiov.iov_len = SCARG(uap, len);
403 	return (sendit(p, SCARG(uap, s), &msg, SCARG(uap, flags), retval));
404 }
405 
406 int
407 sys_sendmsg(struct proc *p, void *v, register_t *retval)
408 {
409 	struct sys_sendmsg_args /* {
410 		syscallarg(int)				s;
411 		syscallarg(const struct msghdr *)	msg;
412 		syscallarg(int)				flags;
413 	} */ *uap = v;
414 	struct msghdr	msg;
415 	struct iovec	aiov[UIO_SMALLIOV], *iov;
416 	int		error;
417 
418 	error = copyin(SCARG(uap, msg), (caddr_t)&msg, sizeof(msg));
419 	if (error)
420 		return (error);
421 	if ((unsigned int)msg.msg_iovlen > UIO_SMALLIOV) {
422 		if ((unsigned int)msg.msg_iovlen > IOV_MAX)
423 			return (EMSGSIZE);
424 		iov = malloc(sizeof(struct iovec) * msg.msg_iovlen,
425 		    M_IOV, M_WAITOK);
426 	} else
427 		iov = aiov;
428 	if ((unsigned int)msg.msg_iovlen > 0) {
429 		error = copyin((caddr_t)msg.msg_iov, (caddr_t)iov,
430 		    (size_t)(msg.msg_iovlen * sizeof(struct iovec)));
431 		if (error)
432 			goto done;
433 	}
434 	msg.msg_iov = iov;
435 	msg.msg_flags = 0;
436 	error = sendit(p, SCARG(uap, s), &msg, SCARG(uap, flags), retval);
437 done:
438 	if (iov != aiov)
439 		free(iov, M_IOV);
440 	return (error);
441 }
442 
443 int
444 sendit(struct proc *p, int s, struct msghdr *mp, int flags, register_t *retsize)
445 {
446 	struct file	*fp;
447 	struct uio	auio;
448 	struct iovec	*iov;
449 	int		i, len, error;
450 	struct mbuf	*to, *control;
451 	struct socket	*so;
452 #ifdef KTRACE
453 	struct iovec	*ktriov;
454 #endif
455 
456 #ifdef KTRACE
457 	ktriov = NULL;
458 #endif
459 	/* getsock() will use the descriptor for us */
460 	if ((error = getsock(p->p_fd, s, &fp)) != 0)
461 		return (error);
462 	auio.uio_iov = mp->msg_iov;
463 	auio.uio_iovcnt = mp->msg_iovlen;
464 	auio.uio_segflg = UIO_USERSPACE;
465 	auio.uio_rw = UIO_WRITE;
466 	auio.uio_procp = p;
467 	auio.uio_offset = 0;			/* XXX */
468 	auio.uio_resid = 0;
469 	iov = mp->msg_iov;
470 	for (i = 0; i < mp->msg_iovlen; i++, iov++) {
471 #if 0
472 		/* cannot happen; iov_len is unsigned */
473 		if (iov->iov_len < 0) {
474 			error = EINVAL;
475 			goto out;
476 		}
477 #endif
478 		/*
479 		 * Writes return ssize_t because -1 is returned on error.
480 		 * Therefore, we must restrict the length to SSIZE_MAX to
481 		 * avoid garbage return values.
482 		 */
483 		auio.uio_resid += iov->iov_len;
484 		if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
485 			error = EINVAL;
486 			goto out;
487 		}
488 	}
489 	if (mp->msg_name) {
490 		error = sockargs(&to, mp->msg_name, mp->msg_namelen,
491 				 MT_SONAME);
492 		if (error)
493 			goto out;
494 	} else
495 		to = 0;
496 	if (mp->msg_control) {
497 		if (mp->msg_controllen < sizeof(struct cmsghdr)) {
498 			error = EINVAL;
499 			goto bad;
500 		}
501 		error = sockargs(&control, mp->msg_control,
502 				 mp->msg_controllen, MT_CONTROL);
503 		if (error)
504 			goto bad;
505 	} else
506 		control = 0;
507 #ifdef KTRACE
508 	if (KTRPOINT(p, KTR_GENIO)) {
509 		int iovlen = auio.uio_iovcnt * sizeof(struct iovec);
510 
511 		ktriov = malloc(iovlen, M_TEMP, M_WAITOK);
512 		memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen);
513 	}
514 #endif
515 	len = auio.uio_resid;
516 	so = (struct socket *)fp->f_data;
517 	error = (*so->so_send)(so, to, &auio, NULL, control, flags);
518 	if (error) {
519 		if (auio.uio_resid != len && (error == ERESTART ||
520 		    error == EINTR || error == EWOULDBLOCK))
521 			error = 0;
522 		if (error == EPIPE)
523 			psignal(p, SIGPIPE);
524 	}
525 	if (error == 0)
526 		*retsize = len - auio.uio_resid;
527 #ifdef KTRACE
528 	if (ktriov != NULL) {
529 		if (error == 0)
530 			ktrgenio(p, s, UIO_WRITE, ktriov, *retsize, error);
531 		free(ktriov, M_TEMP);
532 	}
533 #endif
534  bad:
535 	if (to)
536 		m_freem(to);
537  out:
538 	FILE_UNUSE(fp, p);
539 	return (error);
540 }
541 
542 int
543 sys_recvfrom(struct proc *p, void *v, register_t *retval)
544 {
545 	struct sys_recvfrom_args /* {
546 		syscallarg(int)			s;
547 		syscallarg(void *)		buf;
548 		syscallarg(size_t)		len;
549 		syscallarg(int)			flags;
550 		syscallarg(struct sockaddr *)	from;
551 		syscallarg(unsigned int *)	fromlenaddr;
552 	} */ *uap = v;
553 	struct msghdr	msg;
554 	struct iovec	aiov;
555 	int		error;
556 
557 	if (SCARG(uap, fromlenaddr)) {
558 		error = copyin((caddr_t)SCARG(uap, fromlenaddr),
559 			       (caddr_t)&msg.msg_namelen,
560 			       sizeof(msg.msg_namelen));
561 		if (error)
562 			return (error);
563 	} else
564 		msg.msg_namelen = 0;
565 	msg.msg_name = (caddr_t)SCARG(uap, from);
566 	msg.msg_iov = &aiov;
567 	msg.msg_iovlen = 1;
568 	aiov.iov_base = SCARG(uap, buf);
569 	aiov.iov_len = SCARG(uap, len);
570 	msg.msg_control = 0;
571 	msg.msg_flags = SCARG(uap, flags);
572 	return (recvit(p, SCARG(uap, s), &msg,
573 		       (caddr_t)SCARG(uap, fromlenaddr), retval));
574 }
575 
576 int
577 sys_recvmsg(struct proc *p, void *v, register_t *retval)
578 {
579 	struct sys_recvmsg_args /* {
580 		syscallarg(int)			s;
581 		syscallarg(struct msghdr *)	msg;
582 		syscallarg(int)			flags;
583 	} */ *uap = v;
584 	struct msghdr	msg;
585 	struct iovec	aiov[UIO_SMALLIOV], *uiov, *iov;
586 	int		error;
587 
588 	error = copyin((caddr_t)SCARG(uap, msg), (caddr_t)&msg,
589 		       sizeof(msg));
590 	if (error)
591 		return (error);
592 	if ((unsigned int)msg.msg_iovlen > UIO_SMALLIOV) {
593 		if ((unsigned int)msg.msg_iovlen > IOV_MAX)
594 			return (EMSGSIZE);
595 		iov = malloc(sizeof(struct iovec) * msg.msg_iovlen,
596 		    M_IOV, M_WAITOK);
597 	} else
598 		iov = aiov;
599 	if ((unsigned int)msg.msg_iovlen > 0) {
600 		error = copyin((caddr_t)msg.msg_iov, (caddr_t)iov,
601 		    (size_t)(msg.msg_iovlen * sizeof(struct iovec)));
602 		if (error)
603 			goto done;
604 	}
605 	uiov = msg.msg_iov;
606 	msg.msg_iov = iov;
607 	msg.msg_flags = SCARG(uap, flags);
608 	if ((error = recvit(p, SCARG(uap, s), &msg, (caddr_t)0, retval)) == 0) {
609 		msg.msg_iov = uiov;
610 		error = copyout((caddr_t)&msg, (caddr_t)SCARG(uap, msg),
611 		    sizeof(msg));
612 	}
613 done:
614 	if (iov != aiov)
615 		free(iov, M_IOV);
616 	return (error);
617 }
618 
619 int
620 recvit(struct proc *p, int s, struct msghdr *mp, caddr_t namelenp,
621 	register_t *retsize)
622 {
623 	struct file	*fp;
624 	struct uio	auio;
625 	struct iovec	*iov;
626 	int		i, len, error;
627 	struct mbuf	*from, *control;
628 	struct socket	*so;
629 #ifdef KTRACE
630 	struct iovec	*ktriov;
631 #endif
632 
633 	from = 0;
634 	control = 0;
635 #ifdef KTRACE
636 	ktriov = NULL;
637 #endif
638 
639 	/* getsock() will use the descriptor for us */
640 	if ((error = getsock(p->p_fd, s, &fp)) != 0)
641 		return (error);
642 	auio.uio_iov = mp->msg_iov;
643 	auio.uio_iovcnt = mp->msg_iovlen;
644 	auio.uio_segflg = UIO_USERSPACE;
645 	auio.uio_rw = UIO_READ;
646 	auio.uio_procp = p;
647 	auio.uio_offset = 0;			/* XXX */
648 	auio.uio_resid = 0;
649 	iov = mp->msg_iov;
650 	for (i = 0; i < mp->msg_iovlen; i++, iov++) {
651 #if 0
652 		/* cannot happen iov_len is unsigned */
653 		if (iov->iov_len < 0) {
654 			error = EINVAL;
655 			goto out1;
656 		}
657 #endif
658 		/*
659 		 * Reads return ssize_t because -1 is returned on error.
660 		 * Therefore we must restrict the length to SSIZE_MAX to
661 		 * avoid garbage return values.
662 		 */
663 		auio.uio_resid += iov->iov_len;
664 		if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
665 			error = EINVAL;
666 			goto out1;
667 		}
668 	}
669 #ifdef KTRACE
670 	if (KTRPOINT(p, KTR_GENIO)) {
671 		int iovlen = auio.uio_iovcnt * sizeof(struct iovec);
672 
673 		ktriov = malloc(iovlen, M_TEMP, M_WAITOK);
674 		memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen);
675 	}
676 #endif
677 	len = auio.uio_resid;
678 	so = (struct socket *)fp->f_data;
679 	error = (*so->so_receive)(so, &from, &auio, NULL,
680 			  mp->msg_control ? &control : NULL, &mp->msg_flags);
681 	if (error) {
682 		if (auio.uio_resid != len && (error == ERESTART ||
683 		    error == EINTR || error == EWOULDBLOCK))
684 			error = 0;
685 	}
686 #ifdef KTRACE
687 	if (ktriov != NULL) {
688 		if (error == 0)
689 			ktrgenio(p, s, UIO_READ, ktriov,
690 			    len - auio.uio_resid, error);
691 		free(ktriov, M_TEMP);
692 	}
693 #endif
694 	if (error)
695 		goto out;
696 	*retsize = len - auio.uio_resid;
697 	if (mp->msg_name) {
698 		len = mp->msg_namelen;
699 		if (len <= 0 || from == 0)
700 			len = 0;
701 		else {
702 			if (len > from->m_len)
703 				len = from->m_len;
704 			/* else if len < from->m_len ??? */
705 			error = copyout(mtod(from, caddr_t),
706 					(caddr_t)mp->msg_name, (unsigned)len);
707 			if (error)
708 				goto out;
709 		}
710 		mp->msg_namelen = len;
711 		if (namelenp &&
712 		    (error = copyout((caddr_t)&len, namelenp, sizeof(int))))
713 			goto out;
714 	}
715 	if (mp->msg_control) {
716 		len = mp->msg_controllen;
717 		if (len <= 0 || control == 0)
718 			len = 0;
719 		else {
720 			struct mbuf *m = control;
721 			caddr_t p = (caddr_t)mp->msg_control;
722 
723 			do {
724 				i = m->m_len;
725 				if (len < i) {
726 					mp->msg_flags |= MSG_CTRUNC;
727 					i = len;
728 				}
729 				error = copyout(mtod(m, caddr_t), p,
730 				    (unsigned)i);
731 				if (m->m_next)
732 					i = ALIGN(i);
733 				p += i;
734 				len -= i;
735 				if (error != 0 || len <= 0)
736 					break;
737 			} while ((m = m->m_next) != NULL);
738 			len = p - (caddr_t)mp->msg_control;
739 		}
740 		mp->msg_controllen = len;
741 	}
742  out:
743 	if (from)
744 		m_freem(from);
745 	if (control)
746 		m_freem(control);
747  out1:
748 	FILE_UNUSE(fp, p);
749 	return (error);
750 }
751 
752 /* ARGSUSED */
753 int
754 sys_shutdown(struct proc *p, void *v, register_t *retval)
755 {
756 	struct sys_shutdown_args /* {
757 		syscallarg(int)	s;
758 		syscallarg(int)	how;
759 	} */ *uap = v;
760 	struct file	*fp;
761 	int		error;
762 
763 	/* getsock() will use the descriptor for us */
764 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
765 		return (error);
766 	error = soshutdown((struct socket *)fp->f_data, SCARG(uap, how));
767 	FILE_UNUSE(fp, p);
768 	return (error);
769 }
770 
771 /* ARGSUSED */
772 int
773 sys_setsockopt(struct proc *p, void *v, register_t *retval)
774 {
775 	struct sys_setsockopt_args /* {
776 		syscallarg(int)			s;
777 		syscallarg(int)			level;
778 		syscallarg(int)			name;
779 		syscallarg(const void *)	val;
780 		syscallarg(unsigned int)	valsize;
781 	} */ *uap = v;
782 	struct file	*fp;
783 	struct mbuf	*m;
784 	int		error;
785 	unsigned int	l;
786 
787 	m = NULL;
788 	/* getsock() will use the descriptor for us */
789 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
790 		return (error);
791 	l = SCARG(uap, valsize);
792 	if (l > MCLBYTES) {
793 		error = EINVAL;
794 		goto out;
795 	}
796 	if (SCARG(uap, val)) {
797 		m = m_get(M_WAIT, MT_SOOPTS);
798 		if (l > MLEN)
799 			MCLGET(m, M_WAIT);
800 		error = copyin(SCARG(uap, val), mtod(m, caddr_t), l);
801 		if (error) {
802 			(void) m_free(m);
803 			goto out;
804 		}
805 		m->m_len = SCARG(uap, valsize);
806 	}
807 	error = sosetopt((struct socket *)fp->f_data, SCARG(uap, level),
808 			 SCARG(uap, name), m);
809  out:
810 	FILE_UNUSE(fp, p);
811 	return (error);
812 }
813 
814 /* ARGSUSED */
815 int
816 sys_getsockopt(struct proc *p, void *v, register_t *retval)
817 {
818 	struct sys_getsockopt_args /* {
819 		syscallarg(int)			s;
820 		syscallarg(int)			level;
821 		syscallarg(int)			name;
822 		syscallarg(void *)		val;
823 		syscallarg(unsigned int *)	avalsize;
824 	} */ *uap = v;
825 	struct file	*fp;
826 	struct mbuf	*m, *m0;
827 	unsigned int	op, i, valsize;
828 	int		error;
829 
830 	m = NULL;
831 	/* getsock() will use the descriptor for us */
832 	if ((error = getsock(p->p_fd, SCARG(uap, s), &fp)) != 0)
833 		return (error);
834 	if (SCARG(uap, val)) {
835 		error = copyin((caddr_t)SCARG(uap, avalsize),
836 			       (caddr_t)&valsize, sizeof(valsize));
837 		if (error)
838 			goto out;
839 	} else
840 		valsize = 0;
841 	if ((error = sogetopt((struct socket *)fp->f_data, SCARG(uap, level),
842 	    SCARG(uap, name), &m)) == 0 && SCARG(uap, val) && valsize &&
843 	    m != NULL) {
844 		op = 0;
845 		while (m && !error && op < valsize) {
846 			i = min(m->m_len, (valsize - op));
847 			error = copyout(mtod(m, caddr_t), SCARG(uap, val), i);
848 			op += i;
849 			SCARG(uap, val) = ((u_int8_t *)SCARG(uap, val)) + i;
850 			m0 = m;
851 			MFREE(m0, m);
852 		}
853 		valsize = op;
854 		if (error == 0)
855 			error = copyout(&valsize,
856 					SCARG(uap, avalsize), sizeof(valsize));
857 	}
858 	if (m != NULL)
859 		(void) m_free(m);
860  out:
861 	FILE_UNUSE(fp, p);
862 	return (error);
863 }
864 
865 #ifdef PIPE_SOCKETPAIR
866 /* ARGSUSED */
867 int
868 sys_pipe(struct proc *p, void *v, register_t *retval)
869 {
870 	struct filedesc	*fdp;
871 	struct file	*rf, *wf;
872 	struct socket	*rso, *wso;
873 	int		fd, error;
874 
875 	fdp = p->p_fd;
876 	if ((error = socreate(AF_LOCAL, &rso, SOCK_STREAM, 0)) != 0)
877 		return (error);
878 	if ((error = socreate(AF_LOCAL, &wso, SOCK_STREAM, 0)) != 0)
879 		goto free1;
880 	/* remember this socket pair implements a pipe */
881 	wso->so_state |= SS_ISAPIPE;
882 	rso->so_state |= SS_ISAPIPE;
883 	/* falloc() will use the descriptor for us */
884 	if ((error = falloc(p, &rf, &fd)) != 0)
885 		goto free2;
886 	retval[0] = fd;
887 	rf->f_flag = FREAD;
888 	rf->f_type = DTYPE_SOCKET;
889 	rf->f_ops = &socketops;
890 	rf->f_data = (caddr_t)rso;
891 	if ((error = falloc(p, &wf, &fd)) != 0)
892 		goto free3;
893 	wf->f_flag = FWRITE;
894 	wf->f_type = DTYPE_SOCKET;
895 	wf->f_ops = &socketops;
896 	wf->f_data = (caddr_t)wso;
897 	retval[1] = fd;
898 	if ((error = unp_connect2(wso, rso)) != 0)
899 		goto free4;
900 	FILE_SET_MATURE(rf);
901 	FILE_SET_MATURE(wf);
902 	FILE_UNUSE(rf, p);
903 	FILE_UNUSE(wf, p);
904 	return (0);
905  free4:
906 	FILE_UNUSE(wf, p);
907 	ffree(wf);
908 	fdremove(fdp, retval[1]);
909  free3:
910 	FILE_UNUSE(rf, p);
911 	ffree(rf);
912 	fdremove(fdp, retval[0]);
913  free2:
914 	(void)soclose(wso);
915  free1:
916 	(void)soclose(rso);
917 	return (error);
918 }
919 #endif /* PIPE_SOCKETPAIR */
920 
921 /*
922  * Get socket name.
923  */
924 /* ARGSUSED */
925 int
926 sys_getsockname(struct proc *p, void *v, register_t *retval)
927 {
928 	struct sys_getsockname_args /* {
929 		syscallarg(int)			fdes;
930 		syscallarg(struct sockaddr *)	asa;
931 		syscallarg(unsigned int *)	alen;
932 	} */ *uap = v;
933 	struct file	*fp;
934 	struct socket	*so;
935 	struct mbuf	*m;
936 	unsigned int	len;
937 	int		error;
938 
939 	/* getsock() will use the descriptor for us */
940 	if ((error = getsock(p->p_fd, SCARG(uap, fdes), &fp)) != 0)
941 		return (error);
942 	error = copyin((caddr_t)SCARG(uap, alen), (caddr_t)&len, sizeof(len));
943 	if (error)
944 		goto out;
945 	so = (struct socket *)fp->f_data;
946 	m = m_getclr(M_WAIT, MT_SONAME);
947 	error = (*so->so_proto->pr_usrreq)(so, PRU_SOCKADDR, (struct mbuf *)0,
948 	    m, (struct mbuf *)0, (struct proc *)0);
949 	if (error)
950 		goto bad;
951 	if (len > m->m_len)
952 		len = m->m_len;
953 	error = copyout(mtod(m, caddr_t), (caddr_t)SCARG(uap, asa), len);
954 	if (error == 0)
955 		error = copyout((caddr_t)&len, (caddr_t)SCARG(uap, alen),
956 		    sizeof(len));
957  bad:
958 	m_freem(m);
959  out:
960 	FILE_UNUSE(fp, p);
961 	return (error);
962 }
963 
964 /*
965  * Get name of peer for connected socket.
966  */
967 /* ARGSUSED */
968 int
969 sys_getpeername(struct proc *p, void *v, register_t *retval)
970 {
971 	struct sys_getpeername_args /* {
972 		syscallarg(int)			fdes;
973 		syscallarg(struct sockaddr *)	asa;
974 		syscallarg(unsigned int *)	alen;
975 	} */ *uap = v;
976 	struct file	*fp;
977 	struct socket	*so;
978 	struct mbuf	*m;
979 	unsigned int	len;
980 	int		error;
981 
982 	/* getsock() will use the descriptor for us */
983 	if ((error = getsock(p->p_fd, SCARG(uap, fdes), &fp)) != 0)
984 		return (error);
985 	so = (struct socket *)fp->f_data;
986 	if ((so->so_state & (SS_ISCONNECTED|SS_ISCONFIRMING)) == 0) {
987 		error = ENOTCONN;
988 		goto out;
989 	}
990 	error = copyin((caddr_t)SCARG(uap, alen), (caddr_t)&len, sizeof(len));
991 	if (error)
992 		goto out;
993 	m = m_getclr(M_WAIT, MT_SONAME);
994 	error = (*so->so_proto->pr_usrreq)(so, PRU_PEERADDR, (struct mbuf *)0,
995 	    m, (struct mbuf *)0, (struct proc *)0);
996 	if (error)
997 		goto bad;
998 	if (len > m->m_len)
999 		len = m->m_len;
1000 	error = copyout(mtod(m, caddr_t), (caddr_t)SCARG(uap, asa), len);
1001 	if (error)
1002 		goto bad;
1003 	error = copyout((caddr_t)&len, (caddr_t)SCARG(uap, alen), sizeof(len));
1004  bad:
1005 	m_freem(m);
1006  out:
1007 	FILE_UNUSE(fp, p);
1008 	return (error);
1009 }
1010 
1011 /*
1012  * XXX In a perfect world, we wouldn't pass around socket control
1013  * XXX arguments in mbufs, and this could go away.
1014  */
1015 int
1016 sockargs(struct mbuf **mp, const void *buf, size_t buflen, int type)
1017 {
1018 	struct sockaddr	*sa;
1019 	struct mbuf	*m;
1020 	int		error;
1021 
1022 	/*
1023 	 * We can't allow socket names > UCHAR_MAX in length, since that
1024 	 * will overflow sa_len.  Control data more than a page size in
1025 	 * length is just too much.
1026 	 */
1027 	if (buflen > (type == MT_SONAME ? UCHAR_MAX : PAGE_SIZE))
1028 		return (EINVAL);
1029 
1030 	/* Allocate an mbuf to hold the arguments. */
1031 	m = m_get(M_WAIT, type);
1032 	if (buflen > MLEN) {
1033 		/*
1034 		 * Won't fit into a regular mbuf, so we allocate just
1035 		 * enough external storage to hold the argument.
1036 		 */
1037 		MEXTMALLOC(m, buflen, M_WAITOK);
1038 	}
1039 	m->m_len = buflen;
1040 	error = copyin(buf, mtod(m, caddr_t), buflen);
1041 	if (error) {
1042 		(void) m_free(m);
1043 		return (error);
1044 	}
1045 	*mp = m;
1046 	if (type == MT_SONAME) {
1047 		sa = mtod(m, struct sockaddr *);
1048 #if BYTE_ORDER != BIG_ENDIAN
1049 		/*
1050 		 * 4.3BSD compat thing - need to stay, since bind(2),
1051 		 * connect(2), sendto(2) were not versioned for COMPAT_43.
1052 		 */
1053 		if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
1054 			sa->sa_family = sa->sa_len;
1055 #endif
1056 		sa->sa_len = buflen;
1057 	}
1058 	return (0);
1059 }
1060 
1061 int
1062 getsock(struct filedesc *fdp, int fdes, struct file **fpp)
1063 {
1064 	struct file	*fp;
1065 
1066 	if ((fp = fd_getfile(fdp, fdes)) == NULL)
1067 		return (EBADF);
1068 
1069 	FILE_USE(fp);
1070 
1071 	if (fp->f_type != DTYPE_SOCKET) {
1072 		FILE_UNUSE(fp, NULL);
1073 		return (ENOTSOCK);
1074 	}
1075 	*fpp = fp;
1076 	return (0);
1077 }
1078