xref: /netbsd-src/sys/kern/sys_generic.c (revision 1ffa7b76c40339c17a0fb2a09fac93f287cfc046)
1 /*	$NetBSD: sys_generic.c,v 1.72 2003/03/26 17:50:16 jdolecek Exp $	*/
2 
3 /*
4  * Copyright (c) 1982, 1986, 1989, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  * (c) UNIX System Laboratories, Inc.
7  * All or some portions of this file are derived from material licensed
8  * to the University of California by American Telephone and Telegraph
9  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10  * the permission of UNIX System Laboratories, Inc.
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  * 1. Redistributions of source code must retain the above copyright
16  *    notice, this list of conditions and the following disclaimer.
17  * 2. Redistributions in binary form must reproduce the above copyright
18  *    notice, this list of conditions and the following disclaimer in the
19  *    documentation and/or other materials provided with the distribution.
20  * 3. All advertising materials mentioning features or use of this software
21  *    must display the following acknowledgement:
22  *	This product includes software developed by the University of
23  *	California, Berkeley and its contributors.
24  * 4. Neither the name of the University nor the names of its contributors
25  *    may be used to endorse or promote products derived from this software
26  *    without specific prior written permission.
27  *
28  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
29  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
30  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
31  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
32  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
33  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
34  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
35  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
36  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
37  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
38  * SUCH DAMAGE.
39  *
40  *	@(#)sys_generic.c	8.9 (Berkeley) 2/14/95
41  */
42 
43 #include <sys/cdefs.h>
44 __KERNEL_RCSID(0, "$NetBSD: sys_generic.c,v 1.72 2003/03/26 17:50:16 jdolecek Exp $");
45 
46 #include "opt_ktrace.h"
47 
48 #include <sys/param.h>
49 #include <sys/systm.h>
50 #include <sys/filedesc.h>
51 #include <sys/ioctl.h>
52 #include <sys/file.h>
53 #include <sys/proc.h>
54 #include <sys/socketvar.h>
55 #include <sys/signalvar.h>
56 #include <sys/uio.h>
57 #include <sys/kernel.h>
58 #include <sys/stat.h>
59 #include <sys/malloc.h>
60 #include <sys/poll.h>
61 #ifdef KTRACE
62 #include <sys/ktrace.h>
63 #endif
64 
65 #include <sys/mount.h>
66 #include <sys/sa.h>
67 #include <sys/syscallargs.h>
68 
69 int selscan __P((struct proc *, fd_mask *, fd_mask *, int, register_t *));
70 int pollscan __P((struct proc *, struct pollfd *, int, register_t *));
71 
72 /*
73  * Read system call.
74  */
75 /* ARGSUSED */
76 int
77 sys_read(struct lwp *l, void *v, register_t *retval)
78 {
79 	struct sys_read_args /* {
80 		syscallarg(int)		fd;
81 		syscallarg(void *)	buf;
82 		syscallarg(size_t)	nbyte;
83 	} */ *uap = v;
84 	int		fd;
85 	struct file	*fp;
86 	struct proc	*p;
87 	struct filedesc	*fdp;
88 
89 	fd = SCARG(uap, fd);
90 	p = l->l_proc;
91 	fdp = p->p_fd;
92 
93 	if ((fp = fd_getfile(fdp, fd)) == NULL)
94 		return (EBADF);
95 
96 	if ((fp->f_flag & FREAD) == 0) {
97 		simple_unlock(&fp->f_slock);
98 		return (EBADF);
99 	}
100 
101 	FILE_USE(fp);
102 
103 	/* dofileread() will unuse the descriptor for us */
104 	return (dofileread(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
105 	    &fp->f_offset, FOF_UPDATE_OFFSET, retval));
106 }
107 
108 int
109 dofileread(struct proc *p, int fd, struct file *fp, void *buf, size_t nbyte,
110 	off_t *offset, int flags, register_t *retval)
111 {
112 	struct uio	auio;
113 	struct iovec	aiov;
114 	size_t		cnt;
115 	int		error;
116 #ifdef KTRACE
117 	struct iovec	ktriov;
118 #endif
119 	error = 0;
120 
121 	aiov.iov_base = (caddr_t)buf;
122 	aiov.iov_len = nbyte;
123 	auio.uio_iov = &aiov;
124 	auio.uio_iovcnt = 1;
125 	auio.uio_resid = nbyte;
126 	auio.uio_rw = UIO_READ;
127 	auio.uio_segflg = UIO_USERSPACE;
128 	auio.uio_procp = p;
129 
130 	/*
131 	 * Reads return ssize_t because -1 is returned on error.  Therefore
132 	 * we must restrict the length to SSIZE_MAX to avoid garbage return
133 	 * values.
134 	 */
135 	if (auio.uio_resid > SSIZE_MAX) {
136 		error = EINVAL;
137 		goto out;
138 	}
139 
140 #ifdef KTRACE
141 	/*
142 	 * if tracing, save a copy of iovec
143 	 */
144 	if (KTRPOINT(p, KTR_GENIO))
145 		ktriov = aiov;
146 #endif
147 	cnt = auio.uio_resid;
148 	error = (*fp->f_ops->fo_read)(fp, offset, &auio, fp->f_cred, flags);
149 	if (error)
150 		if (auio.uio_resid != cnt && (error == ERESTART ||
151 		    error == EINTR || error == EWOULDBLOCK))
152 			error = 0;
153 	cnt -= auio.uio_resid;
154 #ifdef KTRACE
155 	if (KTRPOINT(p, KTR_GENIO) && error == 0)
156 		ktrgenio(p, fd, UIO_READ, &ktriov, cnt, error);
157 #endif
158 	*retval = cnt;
159  out:
160 	FILE_UNUSE(fp, p);
161 	return (error);
162 }
163 
164 /*
165  * Scatter read system call.
166  */
167 int
168 sys_readv(struct lwp *l, void *v, register_t *retval)
169 {
170 	struct sys_readv_args /* {
171 		syscallarg(int)				fd;
172 		syscallarg(const struct iovec *)	iovp;
173 		syscallarg(int)				iovcnt;
174 	} */ *uap = v;
175 	int		fd;
176 	struct file	*fp;
177 	struct proc	*p;
178 	struct filedesc	*fdp;
179 
180 	fd = SCARG(uap, fd);
181 	p = l->l_proc;
182 	fdp = p->p_fd;
183 
184 	if ((fp = fd_getfile(fdp, fd)) == NULL)
185 		return (EBADF);
186 
187 	if ((fp->f_flag & FREAD) == 0) {
188 		simple_unlock(&fp->f_slock);
189 		return (EBADF);
190 	}
191 
192 	FILE_USE(fp);
193 
194 	/* dofilereadv() will unuse the descriptor for us */
195 	return (dofilereadv(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt),
196 	    &fp->f_offset, FOF_UPDATE_OFFSET, retval));
197 }
198 
199 int
200 dofilereadv(struct proc *p, int fd, struct file *fp, const struct iovec *iovp,
201 	int iovcnt, off_t *offset, int flags, register_t *retval)
202 {
203 	struct uio	auio;
204 	struct iovec	*iov, *needfree, aiov[UIO_SMALLIOV];
205 	int		i, error;
206 	size_t		cnt;
207 	u_int		iovlen;
208 #ifdef KTRACE
209 	struct iovec	*ktriov;
210 #endif
211 
212 	error = 0;
213 #ifdef KTRACE
214 	ktriov = NULL;
215 #endif
216 	/* note: can't use iovlen until iovcnt is validated */
217 	iovlen = iovcnt * sizeof(struct iovec);
218 	if ((u_int)iovcnt > UIO_SMALLIOV) {
219 		if ((u_int)iovcnt > IOV_MAX) {
220 			error = EINVAL;
221 			goto out;
222 		}
223 		iov = malloc(iovlen, M_IOV, M_WAITOK);
224 		needfree = iov;
225 	} else if ((u_int)iovcnt > 0) {
226 		iov = aiov;
227 		needfree = NULL;
228 	} else {
229 		error = EINVAL;
230 		goto out;
231 	}
232 
233 	auio.uio_iov = iov;
234 	auio.uio_iovcnt = iovcnt;
235 	auio.uio_rw = UIO_READ;
236 	auio.uio_segflg = UIO_USERSPACE;
237 	auio.uio_procp = p;
238 	error = copyin(iovp, iov, iovlen);
239 	if (error)
240 		goto done;
241 	auio.uio_resid = 0;
242 	for (i = 0; i < iovcnt; i++) {
243 		auio.uio_resid += iov->iov_len;
244 		/*
245 		 * Reads return ssize_t because -1 is returned on error.
246 		 * Therefore we must restrict the length to SSIZE_MAX to
247 		 * avoid garbage return values.
248 		 */
249 		if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
250 			error = EINVAL;
251 			goto done;
252 		}
253 		iov++;
254 	}
255 #ifdef KTRACE
256 	/*
257 	 * if tracing, save a copy of iovec
258 	 */
259 	if (KTRPOINT(p, KTR_GENIO))  {
260 		ktriov = malloc(iovlen, M_TEMP, M_WAITOK);
261 		memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen);
262 	}
263 #endif
264 	cnt = auio.uio_resid;
265 	error = (*fp->f_ops->fo_read)(fp, offset, &auio, fp->f_cred, flags);
266 	if (error)
267 		if (auio.uio_resid != cnt && (error == ERESTART ||
268 		    error == EINTR || error == EWOULDBLOCK))
269 			error = 0;
270 	cnt -= auio.uio_resid;
271 #ifdef KTRACE
272 	if (ktriov != NULL) {
273 		if (error == 0)
274 			ktrgenio(p, fd, UIO_READ, ktriov, cnt, error);
275 		free(ktriov, M_TEMP);
276 	}
277 #endif
278 	*retval = cnt;
279  done:
280 	if (needfree)
281 		free(needfree, M_IOV);
282  out:
283 	FILE_UNUSE(fp, p);
284 	return (error);
285 }
286 
287 /*
288  * Write system call
289  */
290 int
291 sys_write(struct lwp *l, void *v, register_t *retval)
292 {
293 	struct sys_write_args /* {
294 		syscallarg(int)			fd;
295 		syscallarg(const void *)	buf;
296 		syscallarg(size_t)		nbyte;
297 	} */ *uap = v;
298 	int		fd;
299 	struct file	*fp;
300 	struct proc	*p;
301 	struct filedesc	*fdp;
302 
303 	fd = SCARG(uap, fd);
304 	p = l->l_proc;
305 	fdp = p->p_fd;
306 
307 	if ((fp = fd_getfile(fdp, fd)) == NULL)
308 		return (EBADF);
309 
310 	if ((fp->f_flag & FWRITE) == 0) {
311 		simple_unlock(&fp->f_slock);
312 		return (EBADF);
313 	}
314 
315 	FILE_USE(fp);
316 
317 	/* dofilewrite() will unuse the descriptor for us */
318 	return (dofilewrite(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
319 	    &fp->f_offset, FOF_UPDATE_OFFSET, retval));
320 }
321 
322 int
323 dofilewrite(struct proc *p, int fd, struct file *fp, const void *buf,
324 	size_t nbyte, off_t *offset, int flags, register_t *retval)
325 {
326 	struct uio	auio;
327 	struct iovec	aiov;
328 	size_t		cnt;
329 	int		error;
330 #ifdef KTRACE
331 	struct iovec	ktriov;
332 #endif
333 
334 	error = 0;
335 	aiov.iov_base = (caddr_t)buf;		/* XXX kills const */
336 	aiov.iov_len = nbyte;
337 	auio.uio_iov = &aiov;
338 	auio.uio_iovcnt = 1;
339 	auio.uio_resid = nbyte;
340 	auio.uio_rw = UIO_WRITE;
341 	auio.uio_segflg = UIO_USERSPACE;
342 	auio.uio_procp = p;
343 
344 	/*
345 	 * Writes return ssize_t because -1 is returned on error.  Therefore
346 	 * we must restrict the length to SSIZE_MAX to avoid garbage return
347 	 * values.
348 	 */
349 	if (auio.uio_resid > SSIZE_MAX) {
350 		error = EINVAL;
351 		goto out;
352 	}
353 
354 #ifdef KTRACE
355 	/*
356 	 * if tracing, save a copy of iovec
357 	 */
358 	if (KTRPOINT(p, KTR_GENIO))
359 		ktriov = aiov;
360 #endif
361 	cnt = auio.uio_resid;
362 	error = (*fp->f_ops->fo_write)(fp, offset, &auio, fp->f_cred, flags);
363 	if (error) {
364 		if (auio.uio_resid != cnt && (error == ERESTART ||
365 		    error == EINTR || error == EWOULDBLOCK))
366 			error = 0;
367 		if (error == EPIPE)
368 			psignal(p, SIGPIPE);
369 	}
370 	cnt -= auio.uio_resid;
371 #ifdef KTRACE
372 	if (KTRPOINT(p, KTR_GENIO) && error == 0)
373 		ktrgenio(p, fd, UIO_WRITE, &ktriov, cnt, error);
374 #endif
375 	*retval = cnt;
376  out:
377 	FILE_UNUSE(fp, p);
378 	return (error);
379 }
380 
381 /*
382  * Gather write system call
383  */
384 int
385 sys_writev(struct lwp *l, void *v, register_t *retval)
386 {
387 	struct sys_writev_args /* {
388 		syscallarg(int)				fd;
389 		syscallarg(const struct iovec *)	iovp;
390 		syscallarg(int)				iovcnt;
391 	} */ *uap = v;
392 	int		fd;
393 	struct file	*fp;
394 	struct proc	*p;
395 	struct filedesc	*fdp;
396 
397 	fd = SCARG(uap, fd);
398 	p = l->l_proc;
399 	fdp = p->p_fd;
400 
401 	if ((fp = fd_getfile(fdp, fd)) == NULL)
402 		return (EBADF);
403 
404 	if ((fp->f_flag & FWRITE) == 0) {
405 		simple_unlock(&fp->f_slock);
406 		return (EBADF);
407 	}
408 
409 	FILE_USE(fp);
410 
411 	/* dofilewritev() will unuse the descriptor for us */
412 	return (dofilewritev(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt),
413 	    &fp->f_offset, FOF_UPDATE_OFFSET, retval));
414 }
415 
416 int
417 dofilewritev(struct proc *p, int fd, struct file *fp, const struct iovec *iovp,
418 	int iovcnt, off_t *offset, int flags, register_t *retval)
419 {
420 	struct uio	auio;
421 	struct iovec	*iov, *needfree, aiov[UIO_SMALLIOV];
422 	int		i, error;
423 	size_t		cnt;
424 	u_int		iovlen;
425 #ifdef KTRACE
426 	struct iovec	*ktriov;
427 #endif
428 
429 	error = 0;
430 #ifdef KTRACE
431 	ktriov = NULL;
432 #endif
433 	/* note: can't use iovlen until iovcnt is validated */
434 	iovlen = iovcnt * sizeof(struct iovec);
435 	if ((u_int)iovcnt > UIO_SMALLIOV) {
436 		if ((u_int)iovcnt > IOV_MAX) {
437 			error = EINVAL;
438 			goto out;
439 		}
440 		iov = malloc(iovlen, M_IOV, M_WAITOK);
441 		needfree = iov;
442 	} else if ((u_int)iovcnt > 0) {
443 		iov = aiov;
444 		needfree = NULL;
445 	} else {
446 		error = EINVAL;
447 		goto out;
448 	}
449 
450 	auio.uio_iov = iov;
451 	auio.uio_iovcnt = iovcnt;
452 	auio.uio_rw = UIO_WRITE;
453 	auio.uio_segflg = UIO_USERSPACE;
454 	auio.uio_procp = p;
455 	error = copyin(iovp, iov, iovlen);
456 	if (error)
457 		goto done;
458 	auio.uio_resid = 0;
459 	for (i = 0; i < iovcnt; i++) {
460 		auio.uio_resid += iov->iov_len;
461 		/*
462 		 * Writes return ssize_t because -1 is returned on error.
463 		 * Therefore we must restrict the length to SSIZE_MAX to
464 		 * avoid garbage return values.
465 		 */
466 		if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
467 			error = EINVAL;
468 			goto done;
469 		}
470 		iov++;
471 	}
472 #ifdef KTRACE
473 	/*
474 	 * if tracing, save a copy of iovec
475 	 */
476 	if (KTRPOINT(p, KTR_GENIO))  {
477 		ktriov = malloc(iovlen, M_TEMP, M_WAITOK);
478 		memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen);
479 	}
480 #endif
481 	cnt = auio.uio_resid;
482 	error = (*fp->f_ops->fo_write)(fp, offset, &auio, fp->f_cred, flags);
483 	if (error) {
484 		if (auio.uio_resid != cnt && (error == ERESTART ||
485 		    error == EINTR || error == EWOULDBLOCK))
486 			error = 0;
487 		if (error == EPIPE)
488 			psignal(p, SIGPIPE);
489 	}
490 	cnt -= auio.uio_resid;
491 #ifdef KTRACE
492 	if (KTRPOINT(p, KTR_GENIO))
493 		if (error == 0) {
494 			ktrgenio(p, fd, UIO_WRITE, ktriov, cnt, error);
495 		free(ktriov, M_TEMP);
496 	}
497 #endif
498 	*retval = cnt;
499  done:
500 	if (needfree)
501 		free(needfree, M_IOV);
502  out:
503 	FILE_UNUSE(fp, p);
504 	return (error);
505 }
506 
507 /*
508  * Ioctl system call
509  */
510 /* ARGSUSED */
511 int
512 sys_ioctl(struct lwp *l, void *v, register_t *retval)
513 {
514 	struct sys_ioctl_args /* {
515 		syscallarg(int)		fd;
516 		syscallarg(u_long)	com;
517 		syscallarg(caddr_t)	data;
518 	} */ *uap = v;
519 	struct file	*fp;
520 	struct proc	*p;
521 	struct filedesc	*fdp;
522 	u_long		com;
523 	int		error;
524 	u_int		size;
525 	caddr_t		data, memp;
526 	int		tmp;
527 #define	STK_PARAMS	128
528 	u_long		stkbuf[STK_PARAMS/sizeof(u_long)];
529 
530 	error = 0;
531 	p = l->l_proc;
532 	fdp = p->p_fd;
533 
534 	if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL)
535 		return (EBADF);
536 
537 	FILE_USE(fp);
538 
539 	if ((fp->f_flag & (FREAD | FWRITE)) == 0) {
540 		error = EBADF;
541 		com = 0;
542 		goto out;
543 	}
544 
545 	switch (com = SCARG(uap, com)) {
546 	case FIONCLEX:
547 		fdp->fd_ofileflags[SCARG(uap, fd)] &= ~UF_EXCLOSE;
548 		goto out;
549 
550 	case FIOCLEX:
551 		fdp->fd_ofileflags[SCARG(uap, fd)] |= UF_EXCLOSE;
552 		goto out;
553 	}
554 
555 	/*
556 	 * Interpret high order word to find amount of data to be
557 	 * copied to/from the user's address space.
558 	 */
559 	size = IOCPARM_LEN(com);
560 	if (size > IOCPARM_MAX) {
561 		error = ENOTTY;
562 		goto out;
563 	}
564 	memp = NULL;
565 	if (size > sizeof(stkbuf)) {
566 		memp = (caddr_t)malloc((u_long)size, M_IOCTLOPS, M_WAITOK);
567 		data = memp;
568 	} else
569 		data = (caddr_t)stkbuf;
570 	if (com&IOC_IN) {
571 		if (size) {
572 			error = copyin(SCARG(uap, data), data, size);
573 			if (error) {
574 				if (memp)
575 					free(memp, M_IOCTLOPS);
576 				goto out;
577 			}
578 		} else
579 			*(caddr_t *)data = SCARG(uap, data);
580 	} else if ((com&IOC_OUT) && size)
581 		/*
582 		 * Zero the buffer so the user always
583 		 * gets back something deterministic.
584 		 */
585 		memset(data, 0, size);
586 	else if (com&IOC_VOID)
587 		*(caddr_t *)data = SCARG(uap, data);
588 
589 	switch (com) {
590 
591 	case FIONBIO:
592 		if ((tmp = *(int *)data) != 0)
593 			fp->f_flag |= FNONBLOCK;
594 		else
595 			fp->f_flag &= ~FNONBLOCK;
596 		error = (*fp->f_ops->fo_ioctl)(fp, FIONBIO, (caddr_t)&tmp, p);
597 		break;
598 
599 	case FIOASYNC:
600 		if ((tmp = *(int *)data) != 0)
601 			fp->f_flag |= FASYNC;
602 		else
603 			fp->f_flag &= ~FASYNC;
604 		error = (*fp->f_ops->fo_ioctl)(fp, FIOASYNC, (caddr_t)&tmp, p);
605 		break;
606 
607 	case FIOSETOWN:
608 		tmp = *(int *)data;
609 		if (fp->f_type == DTYPE_SOCKET) {
610 			((struct socket *)fp->f_data)->so_pgid = tmp;
611 			error = 0;
612 			break;
613 		}
614 		if (tmp <= 0) {
615 			tmp = -tmp;
616 		} else {
617 			struct proc *p1 = pfind(tmp);
618 			if (p1 == 0) {
619 				error = ESRCH;
620 				break;
621 			}
622 			tmp = p1->p_pgrp->pg_id;
623 		}
624 		error = (*fp->f_ops->fo_ioctl)
625 			(fp, TIOCSPGRP, (caddr_t)&tmp, p);
626 		break;
627 
628 	case FIOGETOWN:
629 		if (fp->f_type == DTYPE_SOCKET) {
630 			error = 0;
631 			*(int *)data = ((struct socket *)fp->f_data)->so_pgid;
632 			break;
633 		}
634 		error = (*fp->f_ops->fo_ioctl)(fp, TIOCGPGRP, data, p);
635 		if (error == 0)
636 			*(int *)data = -*(int *)data;
637 		break;
638 
639 	default:
640 		error = (*fp->f_ops->fo_ioctl)(fp, com, data, p);
641 		/*
642 		 * Copy any data to user, size was
643 		 * already set and checked above.
644 		 */
645 		if (error == 0 && (com&IOC_OUT) && size)
646 			error = copyout(data, SCARG(uap, data), size);
647 		break;
648 	}
649 	if (memp)
650 		free(memp, M_IOCTLOPS);
651  out:
652 	FILE_UNUSE(fp, p);
653 	switch (error) {
654 	case -1:
655 		printf("sys_ioctl: _IO%s%s('%c', %lu, %lu) returned -1: "
656 		    "pid=%d comm=%s\n",
657 		    (com & IOC_IN) ? "W" : "", (com & IOC_OUT) ? "R" : "",
658 		    (char)IOCGROUP(com), (com & 0xff), IOCPARM_LEN(com),
659 		    p->p_pid, p->p_comm);
660 		/* FALLTHROUGH */
661 	case EPASSTHROUGH:
662 		error = ENOTTY;
663 		/* FALLTHROUGH */
664 	default:
665 		return (error);
666 	}
667 }
668 
669 int	selwait, nselcoll;
670 
671 /*
672  * Select system call.
673  */
674 int
675 sys_select(struct lwp *l, void *v, register_t *retval)
676 {
677 	struct sys_select_args /* {
678 		syscallarg(int)			nd;
679 		syscallarg(fd_set *)		in;
680 		syscallarg(fd_set *)		ou;
681 		syscallarg(fd_set *)		ex;
682 		syscallarg(struct timeval *)	tv;
683 	} */ *uap = v;
684 	struct proc	*p;
685 	caddr_t		bits;
686 	char		smallbits[howmany(FD_SETSIZE, NFDBITS) *
687 			    sizeof(fd_mask) * 6];
688 	struct		timeval atv;
689 	int		s, ncoll, error, timo;
690 	size_t		ni;
691 
692 	error = 0;
693 	p = l->l_proc;
694 	if (SCARG(uap, nd) < 0)
695 		return (EINVAL);
696 	if (SCARG(uap, nd) > p->p_fd->fd_nfiles) {
697 		/* forgiving; slightly wrong */
698 		SCARG(uap, nd) = p->p_fd->fd_nfiles;
699 	}
700 	ni = howmany(SCARG(uap, nd), NFDBITS) * sizeof(fd_mask);
701 	if (ni * 6 > sizeof(smallbits))
702 		bits = malloc(ni * 6, M_TEMP, M_WAITOK);
703 	else
704 		bits = smallbits;
705 
706 #define	getbits(name, x)						\
707 	if (SCARG(uap, name)) {						\
708 		error = copyin(SCARG(uap, name), bits + ni * x, ni);	\
709 		if (error)						\
710 			goto done;					\
711 	} else								\
712 		memset(bits + ni * x, 0, ni);
713 	getbits(in, 0);
714 	getbits(ou, 1);
715 	getbits(ex, 2);
716 #undef	getbits
717 
718 	timo = 0;
719 	if (SCARG(uap, tv)) {
720 		error = copyin(SCARG(uap, tv), (caddr_t)&atv,
721 			sizeof(atv));
722 		if (error)
723 			goto done;
724 		if (itimerfix(&atv)) {
725 			error = EINVAL;
726 			goto done;
727 		}
728 		s = splclock();
729 		timeradd(&atv, &time, &atv);
730 		splx(s);
731 	}
732 
733  retry:
734 	ncoll = nselcoll;
735 	l->l_flag |= L_SELECT;
736 	error = selscan(p, (fd_mask *)(bits + ni * 0),
737 			   (fd_mask *)(bits + ni * 3), SCARG(uap, nd), retval);
738 	if (error || *retval)
739 		goto done;
740 	if (SCARG(uap, tv)) {
741 		/*
742 		 * We have to recalculate the timeout on every retry.
743 		 */
744 		timo = hzto(&atv);
745 		if (timo <= 0)
746 			goto done;
747 	}
748 	s = splsched();
749 	if ((l->l_flag & L_SELECT) == 0 || nselcoll != ncoll) {
750 		splx(s);
751 		goto retry;
752 	}
753 	l->l_flag &= ~L_SELECT;
754 	error = tsleep((caddr_t)&selwait, PSOCK | PCATCH, "select", timo);
755 	splx(s);
756 	if (error == 0)
757 		goto retry;
758  done:
759 	l->l_flag &= ~L_SELECT;
760 	/* select is not restarted after signals... */
761 	if (error == ERESTART)
762 		error = EINTR;
763 	if (error == EWOULDBLOCK)
764 		error = 0;
765 	if (error == 0) {
766 
767 #define	putbits(name, x)						\
768 		if (SCARG(uap, name)) {					\
769 			error = copyout(bits + ni * x, SCARG(uap, name), ni); \
770 			if (error)					\
771 				goto out;				\
772 		}
773 		putbits(in, 3);
774 		putbits(ou, 4);
775 		putbits(ex, 5);
776 #undef putbits
777 	}
778  out:
779 	if (ni * 6 > sizeof(smallbits))
780 		free(bits, M_TEMP);
781 	return (error);
782 }
783 
784 int
785 selscan(struct proc *p, fd_mask *ibitp, fd_mask *obitp, int nfd,
786 	register_t *retval)
787 {
788 	struct filedesc	*fdp;
789 	int		msk, i, j, fd, n;
790 	fd_mask		ibits, obits;
791 	struct file	*fp;
792 	static const int flag[3] = { POLLRDNORM | POLLHUP | POLLERR,
793 			       POLLWRNORM | POLLHUP | POLLERR,
794 			       POLLRDBAND };
795 
796 	fdp = p->p_fd;
797 	n = 0;
798 	for (msk = 0; msk < 3; msk++) {
799 		for (i = 0; i < nfd; i += NFDBITS) {
800 			ibits = *ibitp++;
801 			obits = 0;
802 			while ((j = ffs(ibits)) && (fd = i + --j) < nfd) {
803 				ibits &= ~(1 << j);
804 				if ((fp = fd_getfile(fdp, fd)) == NULL)
805 					return (EBADF);
806 				FILE_USE(fp);
807 				if ((*fp->f_ops->fo_poll)(fp, flag[msk], p)) {
808 					obits |= (1 << j);
809 					n++;
810 				}
811 				FILE_UNUSE(fp, p);
812 			}
813 			*obitp++ = obits;
814 		}
815 	}
816 	*retval = n;
817 	return (0);
818 }
819 
820 /*
821  * Poll system call.
822  */
823 int
824 sys_poll(struct lwp *l, void *v, register_t *retval)
825 {
826 	struct sys_poll_args /* {
827 		syscallarg(struct pollfd *)	fds;
828 		syscallarg(u_int)		nfds;
829 		syscallarg(int)			timeout;
830 	} */ *uap = v;
831 	struct proc	*p;
832 	caddr_t		bits;
833 	char		smallbits[32 * sizeof(struct pollfd)];
834 	struct timeval	atv;
835 	int		s, ncoll, error, timo;
836 	size_t		ni;
837 
838 	error = 0;
839 	p = l->l_proc;
840 	if (SCARG(uap, nfds) > p->p_fd->fd_nfiles) {
841 		/* forgiving; slightly wrong */
842 		SCARG(uap, nfds) = p->p_fd->fd_nfiles;
843 	}
844 	ni = SCARG(uap, nfds) * sizeof(struct pollfd);
845 	if (ni > sizeof(smallbits))
846 		bits = malloc(ni, M_TEMP, M_WAITOK);
847 	else
848 		bits = smallbits;
849 
850 	error = copyin(SCARG(uap, fds), bits, ni);
851 	if (error)
852 		goto done;
853 
854 	timo = 0;
855 	if (SCARG(uap, timeout) != INFTIM) {
856 		atv.tv_sec = SCARG(uap, timeout) / 1000;
857 		atv.tv_usec = (SCARG(uap, timeout) % 1000) * 1000;
858 		if (itimerfix(&atv)) {
859 			error = EINVAL;
860 			goto done;
861 		}
862 		s = splclock();
863 		timeradd(&atv, &time, &atv);
864 		splx(s);
865 	}
866 
867  retry:
868 	ncoll = nselcoll;
869 	l->l_flag |= L_SELECT;
870 	error = pollscan(p, (struct pollfd *)bits, SCARG(uap, nfds), retval);
871 	if (error || *retval)
872 		goto done;
873 	if (SCARG(uap, timeout) != INFTIM) {
874 		/*
875 		 * We have to recalculate the timeout on every retry.
876 		 */
877 		timo = hzto(&atv);
878 		if (timo <= 0)
879 			goto done;
880 	}
881 	s = splsched();
882 	if ((l->l_flag & L_SELECT) == 0 || nselcoll != ncoll) {
883 		splx(s);
884 		goto retry;
885 	}
886 	l->l_flag &= ~L_SELECT;
887 	error = tsleep((caddr_t)&selwait, PSOCK | PCATCH, "select", timo);
888 	splx(s);
889 	if (error == 0)
890 		goto retry;
891  done:
892 	l->l_flag &= ~L_SELECT;
893 	/* poll is not restarted after signals... */
894 	if (error == ERESTART)
895 		error = EINTR;
896 	if (error == EWOULDBLOCK)
897 		error = 0;
898 	if (error == 0) {
899 		error = copyout(bits, SCARG(uap, fds), ni);
900 		if (error)
901 			goto out;
902 	}
903  out:
904 	if (ni > sizeof(smallbits))
905 		free(bits, M_TEMP);
906 	return (error);
907 }
908 
909 int
910 pollscan(struct proc *p, struct pollfd *fds, int nfd, register_t *retval)
911 {
912 	struct filedesc	*fdp;
913 	int		i, n;
914 	struct file	*fp;
915 
916 	fdp = p->p_fd;
917 	n = 0;
918 	for (i = 0; i < nfd; i++, fds++) {
919 		if (fds->fd >= fdp->fd_nfiles) {
920 			fds->revents = POLLNVAL;
921 			n++;
922 		} else if (fds->fd < 0) {
923 			fds->revents = 0;
924 		} else {
925 			if ((fp = fd_getfile(fdp, fds->fd)) == NULL) {
926 				fds->revents = POLLNVAL;
927 				n++;
928 			} else {
929 				FILE_USE(fp);
930 				fds->revents = (*fp->f_ops->fo_poll)(fp,
931 				    fds->events | POLLERR | POLLHUP, p);
932 				if (fds->revents != 0)
933 					n++;
934 				FILE_UNUSE(fp, p);
935 			}
936 		}
937 	}
938 	*retval = n;
939 	return (0);
940 }
941 
942 /*ARGSUSED*/
943 int
944 seltrue(dev_t dev, int events, struct proc *p)
945 {
946 
947 	return (events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
948 }
949 
950 /*
951  * Record a select request.
952  */
953 void
954 selrecord(struct proc *selector, struct selinfo *sip)
955 {
956 	struct lwp	*l;
957 	struct proc	*p;
958 	pid_t		mypid;
959 	int		collision;
960 
961 	mypid = selector->p_pid;
962 	if (sip->sel_pid == mypid)
963 		return;
964 	collision = 0;
965 	if (sip->sel_pid && (p = pfind(sip->sel_pid))) {
966 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
967 			if (l->l_wchan == (caddr_t)&selwait) {
968 				collision = 1;
969 				sip->sel_flags |= SI_COLL;
970 			}
971 		}
972 	}
973 
974 	if (collision == 0)
975 		sip->sel_pid = mypid;
976 }
977 
978 /*
979  * Do a wakeup when a selectable event occurs.
980  */
981 void
982 selwakeup(sip)
983 	struct selinfo *sip;
984 {
985 	struct lwp *l;
986 	struct proc *p;
987 	int s;
988 
989 	if (sip->sel_pid == 0)
990 		return;
991 	if (sip->sel_flags & SI_COLL) {
992 		sip->sel_pid = 0;
993 		nselcoll++;
994 		sip->sel_flags &= ~SI_COLL;
995 		wakeup((caddr_t)&selwait);
996 		return;
997 	}
998 	p = pfind(sip->sel_pid);
999 	sip->sel_pid = 0;
1000 	if (p != NULL) {
1001 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
1002 			SCHED_LOCK(s);
1003 			if (l->l_wchan == (caddr_t)&selwait) {
1004 				if (l->l_stat == LSSLEEP)
1005 					setrunnable(l);
1006 				else
1007 					unsleep(l);
1008 			} else if (l->l_flag & L_SELECT)
1009 				l->l_flag &= ~L_SELECT;
1010 			SCHED_UNLOCK(s);
1011 		}
1012 	}
1013 }
1014