xref: /netbsd-src/sys/kern/sys_generic.c (revision f5d3fbbc6ff4a77159fb268d247bd94cb7d7e332)
1 /*	$NetBSD: sys_generic.c,v 1.34 1997/10/15 17:04:14 mycroft 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	8.5 (Berkeley) 1/21/94
41  */
42 
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/filedesc.h>
46 #include <sys/ioctl.h>
47 #include <sys/file.h>
48 #include <sys/proc.h>
49 #include <sys/socketvar.h>
50 #include <sys/signalvar.h>
51 #include <sys/uio.h>
52 #include <sys/kernel.h>
53 #include <sys/stat.h>
54 #include <sys/malloc.h>
55 #include <sys/poll.h>
56 #ifdef KTRACE
57 #include <sys/ktrace.h>
58 #endif
59 
60 #include <sys/mount.h>
61 #include <sys/syscallargs.h>
62 
63 int selscan __P((struct proc *, fd_mask *, fd_mask *, int, register_t *));
64 int pollscan __P((struct proc *, struct pollfd *, int, register_t *));
65 
66 /*
67  * Read system call.
68  */
69 /* ARGSUSED */
70 int
71 sys_read(p, v, retval)
72 	struct proc *p;
73 	void *v;
74 	register_t *retval;
75 {
76 	register struct sys_read_args /* {
77 		syscallarg(int) fd;
78 		syscallarg(void *) buf;
79 		syscallarg(size_t) nbyte;
80 	} */ *uap = v;
81 	int fd = SCARG(uap, fd);
82 	register struct file *fp;
83 	register struct filedesc *fdp = p->p_fd;
84 	struct uio auio;
85 	struct iovec aiov;
86 	long cnt, error = 0;
87 #ifdef KTRACE
88 	struct iovec ktriov;
89 #endif
90 
91 	if ((u_int)fd >= fdp->fd_nfiles ||
92 	    (fp = fdp->fd_ofiles[fd]) == NULL ||
93 	    (fp->f_flag & FREAD) == 0)
94 		return (EBADF);
95 	aiov.iov_base = (caddr_t)SCARG(uap, buf);
96 	aiov.iov_len = SCARG(uap, nbyte);
97 	auio.uio_iov = &aiov;
98 	auio.uio_iovcnt = 1;
99 	auio.uio_resid = SCARG(uap, nbyte);
100 	auio.uio_rw = UIO_READ;
101 	auio.uio_segflg = UIO_USERSPACE;
102 	auio.uio_procp = p;
103 	if (auio.uio_resid < 0)
104 		return EINVAL;
105 #ifdef KTRACE
106 	/*
107 	 * if tracing, save a copy of iovec
108 	 */
109 	if (KTRPOINT(p, KTR_GENIO))
110 		ktriov = aiov;
111 #endif
112 	cnt = SCARG(uap, nbyte);
113 	error = (*fp->f_ops->fo_read)(fp, &auio, fp->f_cred);
114 	if (error)
115 		if (auio.uio_resid != cnt && (error == ERESTART ||
116 		    error == EINTR || error == EWOULDBLOCK))
117 			error = 0;
118 	cnt -= auio.uio_resid;
119 #ifdef KTRACE
120 	if (KTRPOINT(p, KTR_GENIO) && error == 0)
121 		ktrgenio(p->p_tracep, fd, UIO_READ, &ktriov, cnt, error);
122 #endif
123 	*retval = cnt;
124 	return (error);
125 }
126 
127 /*
128  * Scatter read system call.
129  */
130 int
131 sys_readv(p, v, retval)
132 	struct proc *p;
133 	void *v;
134 	register_t *retval;
135 {
136 	register struct sys_readv_args /* {
137 		syscallarg(int) fd;
138 		syscallarg(const struct iovec *) iovp;
139 		syscallarg(int) iovcnt;
140 	} */ *uap = v;
141 	int fd = SCARG(uap, fd);
142 	int iovcnt = SCARG(uap, iovcnt);
143 	register struct file *fp;
144 	register struct filedesc *fdp = p->p_fd;
145 	struct uio auio;
146 	register struct iovec *iov;
147 	struct iovec *needfree;
148 	struct iovec aiov[UIO_SMALLIOV];
149 	long i, cnt, error = 0;
150 	u_int iovlen;
151 #ifdef KTRACE
152 	struct iovec *ktriov = NULL;
153 #endif
154 
155 	if ((u_int)fd >= fdp->fd_nfiles ||
156 	    (fp = fdp->fd_ofiles[fd]) == NULL ||
157 	    (fp->f_flag & FREAD) == 0)
158 		return (EBADF);
159 	/* note: can't use iovlen until iovcnt is validated */
160 	iovlen = iovcnt * sizeof (struct iovec);
161 	if ((u_int)iovcnt > UIO_SMALLIOV) {
162 		if ((u_int)iovcnt > UIO_MAXIOV)
163 			return (EINVAL);
164 		MALLOC(iov, struct iovec *, iovlen, M_IOV, M_WAITOK);
165 		needfree = iov;
166 	} else {
167 		iov = aiov;
168 		needfree = NULL;
169 	}
170 	auio.uio_iov = iov;
171 	auio.uio_iovcnt = iovcnt;
172 	auio.uio_rw = UIO_READ;
173 	auio.uio_segflg = UIO_USERSPACE;
174 	auio.uio_procp = p;
175 	error = copyin(SCARG(uap, iovp), iov, iovlen);
176 	if (error)
177 		goto done;
178 	auio.uio_resid = 0;
179 	for (i = 0; i < iovcnt; i++) {
180 #if 0
181 		/* Cannot happen iov_len is unsigned */
182 		if (iov->iov_len < 0) {
183 			error = EINVAL;
184 			goto done;
185 		}
186 #endif
187 		auio.uio_resid += iov->iov_len;
188 		if (auio.uio_resid < 0) {
189 			error = EINVAL;
190 			goto done;
191 		}
192 		iov++;
193 	}
194 #ifdef KTRACE
195 	/*
196 	 * if tracing, save a copy of iovec
197 	 */
198 	if (KTRPOINT(p, KTR_GENIO))  {
199 		MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK);
200 		bcopy((caddr_t)auio.uio_iov, (caddr_t)ktriov, iovlen);
201 	}
202 #endif
203 	cnt = auio.uio_resid;
204 	error = (*fp->f_ops->fo_read)(fp, &auio, fp->f_cred);
205 	if (error)
206 		if (auio.uio_resid != cnt && (error == ERESTART ||
207 		    error == EINTR || error == EWOULDBLOCK))
208 			error = 0;
209 	cnt -= auio.uio_resid;
210 #ifdef KTRACE
211 	if (KTRPOINT(p, KTR_GENIO))
212 		if (error == 0) {
213 			ktrgenio(p->p_tracep, fd, UIO_READ, ktriov, cnt,
214 			    error);
215 		FREE(ktriov, M_TEMP);
216 	}
217 #endif
218 	*retval = cnt;
219 done:
220 	if (needfree)
221 		FREE(needfree, M_IOV);
222 	return (error);
223 }
224 
225 /*
226  * Write system call
227  */
228 int
229 sys_write(p, v, retval)
230 	struct proc *p;
231 	void *v;
232 	register_t *retval;
233 {
234 	register struct sys_write_args /* {
235 		syscallarg(int) fd;
236 		syscallarg(const void *) buf;
237 		syscallarg(size_t) nbyte;
238 	} */ *uap = v;
239 	int fd = SCARG(uap, fd);
240 	register struct file *fp;
241 	register struct filedesc *fdp = p->p_fd;
242 	struct uio auio;
243 	struct iovec aiov;
244 	long cnt, error = 0;
245 #ifdef KTRACE
246 	struct iovec ktriov;
247 #endif
248 
249 	if ((u_int)fd >= fdp->fd_nfiles ||
250 	    (fp = fdp->fd_ofiles[fd]) == NULL ||
251 	    (fp->f_flag & FWRITE) == 0)
252 		return (EBADF);
253 	aiov.iov_base = (char *)SCARG(uap, buf);	/* XXX kills const */
254 	aiov.iov_len = SCARG(uap, nbyte);
255 	auio.uio_iov = &aiov;
256 	auio.uio_iovcnt = 1;
257 	auio.uio_resid = SCARG(uap, nbyte);
258 	auio.uio_rw = UIO_WRITE;
259 	auio.uio_segflg = UIO_USERSPACE;
260 	auio.uio_procp = p;
261 	if (auio.uio_resid < 0)
262 		return EINVAL;
263 #ifdef KTRACE
264 	/*
265 	 * if tracing, save a copy of iovec
266 	 */
267 	if (KTRPOINT(p, KTR_GENIO))
268 		ktriov = aiov;
269 #endif
270 	cnt = SCARG(uap, nbyte);
271 	error = (*fp->f_ops->fo_write)(fp, &auio, fp->f_cred);
272 	if (error) {
273 		if (auio.uio_resid != cnt && (error == ERESTART ||
274 		    error == EINTR || error == EWOULDBLOCK))
275 			error = 0;
276 		if (error == EPIPE)
277 			psignal(p, SIGPIPE);
278 	}
279 	cnt -= auio.uio_resid;
280 #ifdef KTRACE
281 	if (KTRPOINT(p, KTR_GENIO) && error == 0)
282 		ktrgenio(p->p_tracep, fd, UIO_WRITE, &ktriov, cnt, error);
283 #endif
284 	*retval = cnt;
285 	return (error);
286 }
287 
288 /*
289  * Gather write system call
290  */
291 int
292 sys_writev(p, v, retval)
293 	struct proc *p;
294 	void *v;
295 	register_t *retval;
296 {
297 	register struct sys_writev_args /* {
298 		syscallarg(int) fd;
299 		syscallarg(const struct iovec *) iovp;
300 		syscallarg(u_int) iovcnt;
301 	} */ *uap = v;
302 	int fd = SCARG(uap, fd);
303 	int iovcnt = SCARG(uap, iovcnt);
304 	register struct file *fp;
305 	register struct filedesc *fdp = p->p_fd;
306 	struct uio auio;
307 	register struct iovec *iov;
308 	struct iovec *needfree;
309 	struct iovec aiov[UIO_SMALLIOV];
310 	long i, cnt, error = 0;
311 	u_int iovlen;
312 #ifdef KTRACE
313 	struct iovec *ktriov = NULL;
314 #endif
315 
316 	if ((u_int)fd >= fdp->fd_nfiles ||
317 	    (fp = fdp->fd_ofiles[fd]) == NULL ||
318 	    (fp->f_flag & FWRITE) == 0)
319 		return (EBADF);
320 	/* note: can't use iovlen until iovcnt is validated */
321 	iovlen = iovcnt * sizeof (struct iovec);
322 	if ((u_int)iovcnt > UIO_SMALLIOV) {
323 		if ((u_int)iovcnt > UIO_MAXIOV)
324 			return (EINVAL);
325 		MALLOC(iov, struct iovec *, iovlen, M_IOV, M_WAITOK);
326 		needfree = iov;
327 	} else {
328 		iov = aiov;
329 		needfree = NULL;
330 	}
331 	auio.uio_iov = iov;
332 	auio.uio_iovcnt = iovcnt;
333 	auio.uio_rw = UIO_WRITE;
334 	auio.uio_segflg = UIO_USERSPACE;
335 	auio.uio_procp = p;
336 	error = copyin(SCARG(uap, iovp), iov, iovlen);
337 	if (error)
338 		goto done;
339 	auio.uio_resid = 0;
340 	for (i = 0; i < iovcnt; i++) {
341 #if 0
342 		/* Cannot happen iov_len is unsigned */
343 		if (iov->iov_len < 0) {
344 			error = EINVAL;
345 			goto done;
346 		}
347 #endif
348 		auio.uio_resid += iov->iov_len;
349 		if (auio.uio_resid < 0) {
350 			error = EINVAL;
351 			goto done;
352 		}
353 		iov++;
354 	}
355 #ifdef KTRACE
356 	/*
357 	 * if tracing, save a copy of iovec
358 	 */
359 	if (KTRPOINT(p, KTR_GENIO))  {
360 		MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK);
361 		bcopy((caddr_t)auio.uio_iov, (caddr_t)ktriov, iovlen);
362 	}
363 #endif
364 	cnt = auio.uio_resid;
365 	error = (*fp->f_ops->fo_write)(fp, &auio, fp->f_cred);
366 	if (error) {
367 		if (auio.uio_resid != cnt && (error == ERESTART ||
368 		    error == EINTR || error == EWOULDBLOCK))
369 			error = 0;
370 		if (error == EPIPE)
371 			psignal(p, SIGPIPE);
372 	}
373 	cnt -= auio.uio_resid;
374 #ifdef KTRACE
375 	if (KTRPOINT(p, KTR_GENIO))
376 		if (error == 0) {
377 			ktrgenio(p->p_tracep, fd, UIO_WRITE, ktriov, cnt,
378 			    error);
379 		FREE(ktriov, M_TEMP);
380 	}
381 #endif
382 	*retval = cnt;
383 done:
384 	if (needfree)
385 		FREE(needfree, M_IOV);
386 	return (error);
387 }
388 
389 /*
390  * Ioctl system call
391  */
392 /* ARGSUSED */
393 int
394 sys_ioctl(p, v, retval)
395 	struct proc *p;
396 	void *v;
397 	register_t *retval;
398 {
399 	register struct sys_ioctl_args /* {
400 		syscallarg(int) fd;
401 		syscallarg(u_long) com;
402 		syscallarg(caddr_t) data;
403 	} */ *uap = v;
404 	register struct file *fp;
405 	register struct filedesc *fdp;
406 	register u_long com;
407 	register int error;
408 	register u_int size;
409 	caddr_t data, memp;
410 	int tmp;
411 #define STK_PARAMS	128
412 	char stkbuf[STK_PARAMS];
413 
414 	fdp = p->p_fd;
415 	if ((u_int)SCARG(uap, fd) >= fdp->fd_nfiles ||
416 	    (fp = fdp->fd_ofiles[SCARG(uap, fd)]) == NULL)
417 		return (EBADF);
418 
419 	if ((fp->f_flag & (FREAD | FWRITE)) == 0)
420 		return (EBADF);
421 
422 	switch (com = SCARG(uap, com)) {
423 	case FIONCLEX:
424 		fdp->fd_ofileflags[SCARG(uap, fd)] &= ~UF_EXCLOSE;
425 		return (0);
426 	case FIOCLEX:
427 		fdp->fd_ofileflags[SCARG(uap, fd)] |= UF_EXCLOSE;
428 		return (0);
429 	}
430 
431 	/*
432 	 * Interpret high order word to find amount of data to be
433 	 * copied to/from the user's address space.
434 	 */
435 	size = IOCPARM_LEN(com);
436 	if (size > IOCPARM_MAX)
437 		return (ENOTTY);
438 	memp = NULL;
439 	if (size > sizeof (stkbuf)) {
440 		memp = (caddr_t)malloc((u_long)size, M_IOCTLOPS, M_WAITOK);
441 		data = memp;
442 	} else
443 		data = stkbuf;
444 	if (com&IOC_IN) {
445 		if (size) {
446 			error = copyin(SCARG(uap, data), data, size);
447 			if (error) {
448 				if (memp)
449 					free(memp, M_IOCTLOPS);
450 				return (error);
451 			}
452 		} else
453 			*(caddr_t *)data = SCARG(uap, data);
454 	} else if ((com&IOC_OUT) && size)
455 		/*
456 		 * Zero the buffer so the user always
457 		 * gets back something deterministic.
458 		 */
459 		bzero(data, size);
460 	else if (com&IOC_VOID)
461 		*(caddr_t *)data = SCARG(uap, data);
462 
463 	switch (com) {
464 
465 	case FIONBIO:
466 		if ((tmp = *(int *)data) != 0)
467 			fp->f_flag |= FNONBLOCK;
468 		else
469 			fp->f_flag &= ~FNONBLOCK;
470 		error = (*fp->f_ops->fo_ioctl)(fp, FIONBIO, (caddr_t)&tmp, p);
471 		break;
472 
473 	case FIOASYNC:
474 		if ((tmp = *(int *)data) != 0)
475 			fp->f_flag |= FASYNC;
476 		else
477 			fp->f_flag &= ~FASYNC;
478 		error = (*fp->f_ops->fo_ioctl)(fp, FIOASYNC, (caddr_t)&tmp, p);
479 		break;
480 
481 	case FIOSETOWN:
482 		tmp = *(int *)data;
483 		if (fp->f_type == DTYPE_SOCKET) {
484 			((struct socket *)fp->f_data)->so_pgid = tmp;
485 			error = 0;
486 			break;
487 		}
488 		if (tmp <= 0) {
489 			tmp = -tmp;
490 		} else {
491 			struct proc *p1 = pfind(tmp);
492 			if (p1 == 0) {
493 				error = ESRCH;
494 				break;
495 			}
496 			tmp = p1->p_pgrp->pg_id;
497 		}
498 		error = (*fp->f_ops->fo_ioctl)
499 			(fp, TIOCSPGRP, (caddr_t)&tmp, p);
500 		break;
501 
502 	case FIOGETOWN:
503 		if (fp->f_type == DTYPE_SOCKET) {
504 			error = 0;
505 			*(int *)data = ((struct socket *)fp->f_data)->so_pgid;
506 			break;
507 		}
508 		error = (*fp->f_ops->fo_ioctl)(fp, TIOCGPGRP, data, p);
509 		*(int *)data = -*(int *)data;
510 		break;
511 
512 	default:
513 		error = (*fp->f_ops->fo_ioctl)(fp, com, data, p);
514 		/*
515 		 * Copy any data to user, size was
516 		 * already set and checked above.
517 		 */
518 		if (error == 0 && (com&IOC_OUT) && size)
519 			error = copyout(data, SCARG(uap, data), size);
520 		break;
521 	}
522 	if (memp)
523 		free(memp, M_IOCTLOPS);
524 	return (error);
525 }
526 
527 int	selwait, nselcoll;
528 
529 /*
530  * Select system call.
531  */
532 int
533 sys_select(p, v, retval)
534 	register struct proc *p;
535 	void *v;
536 	register_t *retval;
537 {
538 	register struct sys_select_args /* {
539 		syscallarg(u_int) nd;
540 		syscallarg(fd_set *) in;
541 		syscallarg(fd_set *) ou;
542 		syscallarg(fd_set *) ex;
543 		syscallarg(struct timeval *) tv;
544 	} */ *uap = v;
545 	caddr_t bits;
546 	char smallbits[howmany(FD_SETSIZE, NFDBITS) * sizeof(fd_mask) * 6];
547 	struct timeval atv;
548 	int s, ncoll, error = 0, timo;
549 	size_t ni;
550 
551 	if (SCARG(uap, nd) > p->p_fd->fd_nfiles) {
552 		/* forgiving; slightly wrong */
553 		SCARG(uap, nd) = p->p_fd->fd_nfiles;
554 	}
555 	ni = howmany(SCARG(uap, nd), NFDBITS) * sizeof(fd_mask);
556 	if (ni * 6 > sizeof(smallbits))
557 		bits = malloc(ni * 6, M_TEMP, M_WAITOK);
558 	else
559 		bits = smallbits;
560 
561 #define	getbits(name, x) \
562 	if (SCARG(uap, name)) { \
563 		error = copyin(SCARG(uap, name), bits + ni * x, ni); \
564 		if (error) \
565 			goto done; \
566 	} else \
567 		bzero(bits + ni * x, ni);
568 	getbits(in, 0);
569 	getbits(ou, 1);
570 	getbits(ex, 2);
571 #undef	getbits
572 
573 	if (SCARG(uap, tv)) {
574 		error = copyin(SCARG(uap, tv), (caddr_t)&atv,
575 			sizeof (atv));
576 		if (error)
577 			goto done;
578 		if (itimerfix(&atv)) {
579 			error = EINVAL;
580 			goto done;
581 		}
582 		s = splclock();
583 		timeradd(&atv, &time, &atv);
584 		timo = hzto(&atv);
585 		/*
586 		 * Avoid inadvertently sleeping forever.
587 		 */
588 		if (timo == 0)
589 			timo = 1;
590 		splx(s);
591 	} else
592 		timo = 0;
593 retry:
594 	ncoll = nselcoll;
595 	p->p_flag |= P_SELECT;
596 	error = selscan(p, (fd_mask *)(bits + ni * 0),
597 			   (fd_mask *)(bits + ni * 3), SCARG(uap, nd), retval);
598 	if (error || *retval)
599 		goto done;
600 	s = splhigh();
601 	if (timo && timercmp(&time, &atv, >=)) {
602 		splx(s);
603 		goto done;
604 	}
605 	if ((p->p_flag & P_SELECT) == 0 || nselcoll != ncoll) {
606 		splx(s);
607 		goto retry;
608 	}
609 	p->p_flag &= ~P_SELECT;
610 	error = tsleep((caddr_t)&selwait, PSOCK | PCATCH, "select", timo);
611 	splx(s);
612 	if (error == 0)
613 		goto retry;
614 done:
615 	p->p_flag &= ~P_SELECT;
616 	/* select is not restarted after signals... */
617 	if (error == ERESTART)
618 		error = EINTR;
619 	if (error == EWOULDBLOCK)
620 		error = 0;
621 	if (error == 0) {
622 #define	putbits(name, x) \
623 		if (SCARG(uap, name)) { \
624 			error = copyout(bits + ni * x, SCARG(uap, name), ni); \
625 			if (error) \
626 				goto out; \
627 		}
628 		putbits(in, 3);
629 		putbits(ou, 4);
630 		putbits(ex, 5);
631 #undef putbits
632 	}
633 out:
634 	if (ni * 6 > sizeof(smallbits))
635 		free(bits, M_TEMP);
636 	return (error);
637 }
638 
639 int
640 selscan(p, ibitp, obitp, nfd, retval)
641 	struct proc *p;
642 	fd_mask *ibitp, *obitp;
643 	int nfd;
644 	register_t *retval;
645 {
646 	register struct filedesc *fdp = p->p_fd;
647 	register int msk, i, j, fd;
648 	register fd_mask ibits, obits;
649 	struct file *fp;
650 	int n = 0;
651 	static int flag[3] = { POLLRDNORM | POLLHUP | POLLERR,
652 			       POLLWRNORM | POLLHUP | POLLERR,
653 			       POLLRDBAND };
654 
655 	for (msk = 0; msk < 3; msk++) {
656 		for (i = 0; i < nfd; i += NFDBITS) {
657 			ibits = *ibitp++;
658 			obits = 0;
659 			while ((j = ffs(ibits)) && (fd = i + --j) < nfd) {
660 				ibits &= ~(1 << j);
661 				fp = fdp->fd_ofiles[fd];
662 				if (fp == NULL)
663 					return (EBADF);
664 				if ((*fp->f_ops->fo_poll)(fp, flag[msk], p)) {
665 					obits |= (1 << j);
666 					n++;
667 				}
668 			}
669 			*obitp++ = obits;
670 		}
671 	}
672 	*retval = n;
673 	return (0);
674 }
675 
676 /*
677  * Poll system call.
678  */
679 int
680 sys_poll(p, v, retval)
681 	register struct proc *p;
682 	void *v;
683 	register_t *retval;
684 {
685 	register struct sys_poll_args /* {
686 		syscallarg(struct pollfd *) fds;
687 		syscallarg(u_int) nfds;
688 		syscallarg(int) timeout;
689 	} */ *uap = v;
690 	caddr_t bits;
691 	char smallbits[32 * sizeof(struct pollfd)];
692 	struct timeval atv;
693 	int s, ncoll, error = 0, timo;
694 	size_t ni;
695 
696 	if (SCARG(uap, nfds) > p->p_fd->fd_nfiles) {
697 		/* forgiving; slightly wrong */
698 		SCARG(uap, nfds) = p->p_fd->fd_nfiles;
699 	}
700 	ni = SCARG(uap, nfds) * sizeof(struct pollfd);
701 	if (ni > sizeof(smallbits))
702 		bits = malloc(ni, M_TEMP, M_WAITOK);
703 	else
704 		bits = smallbits;
705 
706 	error = copyin(SCARG(uap, fds), bits, ni);
707 	if (error)
708 		goto done;
709 
710 	if (SCARG(uap, timeout) != INFTIM) {
711 		atv.tv_sec = SCARG(uap, timeout) / 1000;
712 		atv.tv_usec = (SCARG(uap, timeout) % 1000) * 1000;
713 		if (itimerfix(&atv)) {
714 			error = EINVAL;
715 			goto done;
716 		}
717 		s = splclock();
718 		timeradd(&atv, &time, &atv);
719 		timo = hzto(&atv);
720 		/*
721 		 * Avoid inadvertently sleeping forever.
722 		 */
723 		if (timo == 0)
724 			timo = 1;
725 		splx(s);
726 	} else
727 		timo = 0;
728 retry:
729 	ncoll = nselcoll;
730 	p->p_flag |= P_SELECT;
731 	error = pollscan(p, (struct pollfd *)bits, SCARG(uap, nfds), retval);
732 	if (error || *retval)
733 		goto done;
734 	s = splhigh();
735 	if (timo && timercmp(&time, &atv, >=)) {
736 		splx(s);
737 		goto done;
738 	}
739 	if ((p->p_flag & P_SELECT) == 0 || nselcoll != ncoll) {
740 		splx(s);
741 		goto retry;
742 	}
743 	p->p_flag &= ~P_SELECT;
744 	error = tsleep((caddr_t)&selwait, PSOCK | PCATCH, "select", timo);
745 	splx(s);
746 	if (error == 0)
747 		goto retry;
748 done:
749 	p->p_flag &= ~P_SELECT;
750 	/* poll is not restarted after signals... */
751 	if (error == ERESTART)
752 		error = EINTR;
753 	if (error == EWOULDBLOCK)
754 		error = 0;
755 	if (error == 0) {
756 		error = copyout(bits, SCARG(uap, fds), ni);
757 		if (error)
758 			goto out;
759 	}
760 out:
761 	if (ni > sizeof(smallbits))
762 		free(bits, M_TEMP);
763 	return (error);
764 }
765 
766 int
767 pollscan(p, fds, nfd, retval)
768 	struct proc *p;
769 	struct pollfd *fds;
770 	int nfd;
771 	register_t *retval;
772 {
773 	register struct filedesc *fdp = p->p_fd;
774 	int i;
775 	struct file *fp;
776 	int n = 0;
777 
778 	for (i = 0; i < nfd; i++, fds++) {
779 		if ((u_int)fds->fd >= fdp->fd_nfiles) {
780 			fds->revents = POLLNVAL;
781 			n++;
782 		} else {
783 			fp = fdp->fd_ofiles[fds->fd];
784 			if (fp == 0) {
785 				fds->revents = POLLNVAL;
786 				n++;
787 			} else {
788 				fds->revents = (*fp->f_ops->fo_poll)(fp,
789 				    fds->events | POLLERR | POLLHUP, p);
790 				if (fds->revents != 0)
791 					n++;
792 			}
793 		}
794 	}
795 	*retval = n;
796 	return (0);
797 }
798 
799 /*ARGSUSED*/
800 int
801 seltrue(dev, events, p)
802 	dev_t dev;
803 	int events;
804 	struct proc *p;
805 {
806 
807 	return (events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
808 }
809 
810 /*
811  * Record a select request.
812  */
813 void
814 selrecord(selector, sip)
815 	struct proc *selector;
816 	struct selinfo *sip;
817 {
818 	struct proc *p;
819 	pid_t mypid;
820 
821 	mypid = selector->p_pid;
822 	if (sip->si_pid == mypid)
823 		return;
824 	if (sip->si_pid && (p = pfind(sip->si_pid)) &&
825 	    p->p_wchan == (caddr_t)&selwait)
826 		sip->si_flags |= SI_COLL;
827 	else
828 		sip->si_pid = mypid;
829 }
830 
831 /*
832  * Do a wakeup when a selectable event occurs.
833  */
834 void
835 selwakeup(sip)
836 	register struct selinfo *sip;
837 {
838 	register struct proc *p;
839 	int s;
840 
841 	if (sip->si_pid == 0)
842 		return;
843 	if (sip->si_flags & SI_COLL) {
844 		nselcoll++;
845 		sip->si_flags &= ~SI_COLL;
846 		wakeup((caddr_t)&selwait);
847 	}
848 	p = pfind(sip->si_pid);
849 	sip->si_pid = 0;
850 	if (p != NULL) {
851 		s = splhigh();
852 		if (p->p_wchan == (caddr_t)&selwait) {
853 			if (p->p_stat == SSLEEP)
854 				setrunnable(p);
855 			else
856 				unsleep(p);
857 		} else if (p->p_flag & P_SELECT)
858 			p->p_flag &= ~P_SELECT;
859 		splx(s);
860 	}
861 }
862