xref: /netbsd-src/sys/kern/sys_generic.c (revision 7c7c171d130af9949261bc7dce2150a03c3d239c)
1 /*	$NetBSD: sys_generic.c,v 1.36 1998/03/01 02:22:32 fvdl 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/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 		auio.uio_resid += iov->iov_len;
181 		if (auio.uio_resid < 0) {
182 			error = EINVAL;
183 			goto done;
184 		}
185 		iov++;
186 	}
187 #ifdef KTRACE
188 	/*
189 	 * if tracing, save a copy of iovec
190 	 */
191 	if (KTRPOINT(p, KTR_GENIO))  {
192 		MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK);
193 		bcopy((caddr_t)auio.uio_iov, (caddr_t)ktriov, iovlen);
194 	}
195 #endif
196 	cnt = auio.uio_resid;
197 	error = (*fp->f_ops->fo_read)(fp, &auio, fp->f_cred);
198 	if (error)
199 		if (auio.uio_resid != cnt && (error == ERESTART ||
200 		    error == EINTR || error == EWOULDBLOCK))
201 			error = 0;
202 	cnt -= auio.uio_resid;
203 #ifdef KTRACE
204 	if (KTRPOINT(p, KTR_GENIO))
205 		if (error == 0) {
206 			ktrgenio(p->p_tracep, fd, UIO_READ, ktriov, cnt,
207 			    error);
208 		FREE(ktriov, M_TEMP);
209 	}
210 #endif
211 	*retval = cnt;
212 done:
213 	if (needfree)
214 		FREE(needfree, M_IOV);
215 	return (error);
216 }
217 
218 /*
219  * Write system call
220  */
221 int
222 sys_write(p, v, retval)
223 	struct proc *p;
224 	void *v;
225 	register_t *retval;
226 {
227 	register struct sys_write_args /* {
228 		syscallarg(int) fd;
229 		syscallarg(const void *) buf;
230 		syscallarg(size_t) nbyte;
231 	} */ *uap = v;
232 	int fd = SCARG(uap, fd);
233 	register struct file *fp;
234 	register struct filedesc *fdp = p->p_fd;
235 	struct uio auio;
236 	struct iovec aiov;
237 	long cnt, error = 0;
238 #ifdef KTRACE
239 	struct iovec ktriov;
240 #endif
241 
242 	if ((u_int)fd >= fdp->fd_nfiles ||
243 	    (fp = fdp->fd_ofiles[fd]) == NULL ||
244 	    (fp->f_flag & FWRITE) == 0)
245 		return (EBADF);
246 	aiov.iov_base = (char *)SCARG(uap, buf);	/* XXX kills const */
247 	aiov.iov_len = SCARG(uap, nbyte);
248 	auio.uio_iov = &aiov;
249 	auio.uio_iovcnt = 1;
250 	auio.uio_resid = SCARG(uap, nbyte);
251 	auio.uio_rw = UIO_WRITE;
252 	auio.uio_segflg = UIO_USERSPACE;
253 	auio.uio_procp = p;
254 	if (auio.uio_resid < 0)
255 		return EINVAL;
256 #ifdef KTRACE
257 	/*
258 	 * if tracing, save a copy of iovec
259 	 */
260 	if (KTRPOINT(p, KTR_GENIO))
261 		ktriov = aiov;
262 #endif
263 	cnt = SCARG(uap, nbyte);
264 	error = (*fp->f_ops->fo_write)(fp, &auio, fp->f_cred);
265 	if (error) {
266 		if (auio.uio_resid != cnt && (error == ERESTART ||
267 		    error == EINTR || error == EWOULDBLOCK))
268 			error = 0;
269 		if (error == EPIPE)
270 			psignal(p, SIGPIPE);
271 	}
272 	cnt -= auio.uio_resid;
273 #ifdef KTRACE
274 	if (KTRPOINT(p, KTR_GENIO) && error == 0)
275 		ktrgenio(p->p_tracep, fd, UIO_WRITE, &ktriov, cnt, error);
276 #endif
277 	*retval = cnt;
278 	return (error);
279 }
280 
281 /*
282  * Gather write system call
283  */
284 int
285 sys_writev(p, v, retval)
286 	struct proc *p;
287 	void *v;
288 	register_t *retval;
289 {
290 	register struct sys_writev_args /* {
291 		syscallarg(int) fd;
292 		syscallarg(const struct iovec *) iovp;
293 		syscallarg(u_int) iovcnt;
294 	} */ *uap = v;
295 	int fd = SCARG(uap, fd);
296 	int iovcnt = SCARG(uap, iovcnt);
297 	register struct file *fp;
298 	register struct filedesc *fdp = p->p_fd;
299 	struct uio auio;
300 	register struct iovec *iov;
301 	struct iovec *needfree;
302 	struct iovec aiov[UIO_SMALLIOV];
303 	long i, cnt, error = 0;
304 	u_int iovlen;
305 #ifdef KTRACE
306 	struct iovec *ktriov = NULL;
307 #endif
308 
309 	if ((u_int)fd >= fdp->fd_nfiles ||
310 	    (fp = fdp->fd_ofiles[fd]) == NULL ||
311 	    (fp->f_flag & FWRITE) == 0)
312 		return (EBADF);
313 	/* note: can't use iovlen until iovcnt is validated */
314 	iovlen = iovcnt * sizeof (struct iovec);
315 	if ((u_int)iovcnt > UIO_SMALLIOV) {
316 		if ((u_int)iovcnt > UIO_MAXIOV)
317 			return (EINVAL);
318 		MALLOC(iov, struct iovec *, iovlen, M_IOV, M_WAITOK);
319 		needfree = iov;
320 	} else {
321 		iov = aiov;
322 		needfree = NULL;
323 	}
324 	auio.uio_iov = iov;
325 	auio.uio_iovcnt = iovcnt;
326 	auio.uio_rw = UIO_WRITE;
327 	auio.uio_segflg = UIO_USERSPACE;
328 	auio.uio_procp = p;
329 	error = copyin(SCARG(uap, iovp), iov, iovlen);
330 	if (error)
331 		goto done;
332 	auio.uio_resid = 0;
333 	for (i = 0; i < iovcnt; i++) {
334 		auio.uio_resid += iov->iov_len;
335 		if (auio.uio_resid < 0) {
336 			error = EINVAL;
337 			goto done;
338 		}
339 		iov++;
340 	}
341 #ifdef KTRACE
342 	/*
343 	 * if tracing, save a copy of iovec
344 	 */
345 	if (KTRPOINT(p, KTR_GENIO))  {
346 		MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK);
347 		bcopy((caddr_t)auio.uio_iov, (caddr_t)ktriov, iovlen);
348 	}
349 #endif
350 	cnt = auio.uio_resid;
351 	error = (*fp->f_ops->fo_write)(fp, &auio, fp->f_cred);
352 	if (error) {
353 		if (auio.uio_resid != cnt && (error == ERESTART ||
354 		    error == EINTR || error == EWOULDBLOCK))
355 			error = 0;
356 		if (error == EPIPE)
357 			psignal(p, SIGPIPE);
358 	}
359 	cnt -= auio.uio_resid;
360 #ifdef KTRACE
361 	if (KTRPOINT(p, KTR_GENIO))
362 		if (error == 0) {
363 			ktrgenio(p->p_tracep, fd, UIO_WRITE, ktriov, cnt,
364 			    error);
365 		FREE(ktriov, M_TEMP);
366 	}
367 #endif
368 	*retval = cnt;
369 done:
370 	if (needfree)
371 		FREE(needfree, M_IOV);
372 	return (error);
373 }
374 
375 /*
376  * Ioctl system call
377  */
378 /* ARGSUSED */
379 int
380 sys_ioctl(p, v, retval)
381 	struct proc *p;
382 	void *v;
383 	register_t *retval;
384 {
385 	register struct sys_ioctl_args /* {
386 		syscallarg(int) fd;
387 		syscallarg(u_long) com;
388 		syscallarg(caddr_t) data;
389 	} */ *uap = v;
390 	register struct file *fp;
391 	register struct filedesc *fdp;
392 	register u_long com;
393 	register int error;
394 	register u_int size;
395 	caddr_t data, memp;
396 	int tmp;
397 #define STK_PARAMS	128
398 	char stkbuf[STK_PARAMS];
399 
400 	fdp = p->p_fd;
401 	if ((u_int)SCARG(uap, fd) >= fdp->fd_nfiles ||
402 	    (fp = fdp->fd_ofiles[SCARG(uap, fd)]) == NULL)
403 		return (EBADF);
404 
405 	if ((fp->f_flag & (FREAD | FWRITE)) == 0)
406 		return (EBADF);
407 
408 	switch (com = SCARG(uap, com)) {
409 	case FIONCLEX:
410 		fdp->fd_ofileflags[SCARG(uap, fd)] &= ~UF_EXCLOSE;
411 		return (0);
412 	case FIOCLEX:
413 		fdp->fd_ofileflags[SCARG(uap, fd)] |= UF_EXCLOSE;
414 		return (0);
415 	}
416 
417 	/*
418 	 * Interpret high order word to find amount of data to be
419 	 * copied to/from the user's address space.
420 	 */
421 	size = IOCPARM_LEN(com);
422 	if (size > IOCPARM_MAX)
423 		return (ENOTTY);
424 	memp = NULL;
425 	if (size > sizeof (stkbuf)) {
426 		memp = (caddr_t)malloc((u_long)size, M_IOCTLOPS, M_WAITOK);
427 		data = memp;
428 	} else
429 		data = stkbuf;
430 	if (com&IOC_IN) {
431 		if (size) {
432 			error = copyin(SCARG(uap, data), data, size);
433 			if (error) {
434 				if (memp)
435 					free(memp, M_IOCTLOPS);
436 				return (error);
437 			}
438 		} else
439 			*(caddr_t *)data = SCARG(uap, data);
440 	} else if ((com&IOC_OUT) && size)
441 		/*
442 		 * Zero the buffer so the user always
443 		 * gets back something deterministic.
444 		 */
445 		bzero(data, size);
446 	else if (com&IOC_VOID)
447 		*(caddr_t *)data = SCARG(uap, data);
448 
449 	switch (com) {
450 
451 	case FIONBIO:
452 		if ((tmp = *(int *)data) != 0)
453 			fp->f_flag |= FNONBLOCK;
454 		else
455 			fp->f_flag &= ~FNONBLOCK;
456 		error = (*fp->f_ops->fo_ioctl)(fp, FIONBIO, (caddr_t)&tmp, p);
457 		break;
458 
459 	case FIOASYNC:
460 		if ((tmp = *(int *)data) != 0)
461 			fp->f_flag |= FASYNC;
462 		else
463 			fp->f_flag &= ~FASYNC;
464 		error = (*fp->f_ops->fo_ioctl)(fp, FIOASYNC, (caddr_t)&tmp, p);
465 		break;
466 
467 	case FIOSETOWN:
468 		tmp = *(int *)data;
469 		if (fp->f_type == DTYPE_SOCKET) {
470 			((struct socket *)fp->f_data)->so_pgid = tmp;
471 			error = 0;
472 			break;
473 		}
474 		if (tmp <= 0) {
475 			tmp = -tmp;
476 		} else {
477 			struct proc *p1 = pfind(tmp);
478 			if (p1 == 0) {
479 				error = ESRCH;
480 				break;
481 			}
482 			tmp = p1->p_pgrp->pg_id;
483 		}
484 		error = (*fp->f_ops->fo_ioctl)
485 			(fp, TIOCSPGRP, (caddr_t)&tmp, p);
486 		break;
487 
488 	case FIOGETOWN:
489 		if (fp->f_type == DTYPE_SOCKET) {
490 			error = 0;
491 			*(int *)data = ((struct socket *)fp->f_data)->so_pgid;
492 			break;
493 		}
494 		error = (*fp->f_ops->fo_ioctl)(fp, TIOCGPGRP, data, p);
495 		*(int *)data = -*(int *)data;
496 		break;
497 
498 	default:
499 		error = (*fp->f_ops->fo_ioctl)(fp, com, data, p);
500 		/*
501 		 * Copy any data to user, size was
502 		 * already set and checked above.
503 		 */
504 		if (error == 0 && (com&IOC_OUT) && size)
505 			error = copyout(data, SCARG(uap, data), size);
506 		break;
507 	}
508 	if (memp)
509 		free(memp, M_IOCTLOPS);
510 	return (error);
511 }
512 
513 int	selwait, nselcoll;
514 
515 /*
516  * Select system call.
517  */
518 int
519 sys_select(p, v, retval)
520 	register struct proc *p;
521 	void *v;
522 	register_t *retval;
523 {
524 	register struct sys_select_args /* {
525 		syscallarg(int) nd;
526 		syscallarg(fd_set *) in;
527 		syscallarg(fd_set *) ou;
528 		syscallarg(fd_set *) ex;
529 		syscallarg(struct timeval *) tv;
530 	} */ *uap = v;
531 	caddr_t bits;
532 	char smallbits[howmany(FD_SETSIZE, NFDBITS) * sizeof(fd_mask) * 6];
533 	struct timeval atv;
534 	int s, ncoll, error = 0, timo;
535 	size_t ni;
536 
537 	if (SCARG(uap, nd) < 0)
538 		return (EINVAL);
539 	if (SCARG(uap, nd) > p->p_fd->fd_nfiles) {
540 		/* forgiving; slightly wrong */
541 		SCARG(uap, nd) = p->p_fd->fd_nfiles;
542 	}
543 	ni = howmany(SCARG(uap, nd), NFDBITS) * sizeof(fd_mask);
544 	if (ni * 6 > sizeof(smallbits))
545 		bits = malloc(ni * 6, M_TEMP, M_WAITOK);
546 	else
547 		bits = smallbits;
548 
549 #define	getbits(name, x) \
550 	if (SCARG(uap, name)) { \
551 		error = copyin(SCARG(uap, name), bits + ni * x, ni); \
552 		if (error) \
553 			goto done; \
554 	} else \
555 		bzero(bits + ni * x, ni);
556 	getbits(in, 0);
557 	getbits(ou, 1);
558 	getbits(ex, 2);
559 #undef	getbits
560 
561 	if (SCARG(uap, tv)) {
562 		error = copyin(SCARG(uap, tv), (caddr_t)&atv,
563 			sizeof (atv));
564 		if (error)
565 			goto done;
566 		if (itimerfix(&atv)) {
567 			error = EINVAL;
568 			goto done;
569 		}
570 		s = splclock();
571 		timeradd(&atv, &time, &atv);
572 		timo = hzto(&atv);
573 		/*
574 		 * Avoid inadvertently sleeping forever.
575 		 */
576 		if (timo == 0)
577 			timo = 1;
578 		splx(s);
579 	} else
580 		timo = 0;
581 retry:
582 	ncoll = nselcoll;
583 	p->p_flag |= P_SELECT;
584 	error = selscan(p, (fd_mask *)(bits + ni * 0),
585 			   (fd_mask *)(bits + ni * 3), SCARG(uap, nd), retval);
586 	if (error || *retval)
587 		goto done;
588 	s = splhigh();
589 	if (timo && timercmp(&time, &atv, >=)) {
590 		splx(s);
591 		goto done;
592 	}
593 	if ((p->p_flag & P_SELECT) == 0 || nselcoll != ncoll) {
594 		splx(s);
595 		goto retry;
596 	}
597 	p->p_flag &= ~P_SELECT;
598 	error = tsleep((caddr_t)&selwait, PSOCK | PCATCH, "select", timo);
599 	splx(s);
600 	if (error == 0)
601 		goto retry;
602 done:
603 	p->p_flag &= ~P_SELECT;
604 	/* select is not restarted after signals... */
605 	if (error == ERESTART)
606 		error = EINTR;
607 	if (error == EWOULDBLOCK)
608 		error = 0;
609 	if (error == 0) {
610 #define	putbits(name, x) \
611 		if (SCARG(uap, name)) { \
612 			error = copyout(bits + ni * x, SCARG(uap, name), ni); \
613 			if (error) \
614 				goto out; \
615 		}
616 		putbits(in, 3);
617 		putbits(ou, 4);
618 		putbits(ex, 5);
619 #undef putbits
620 	}
621 out:
622 	if (ni * 6 > sizeof(smallbits))
623 		free(bits, M_TEMP);
624 	return (error);
625 }
626 
627 int
628 selscan(p, ibitp, obitp, nfd, retval)
629 	struct proc *p;
630 	fd_mask *ibitp, *obitp;
631 	int nfd;
632 	register_t *retval;
633 {
634 	register struct filedesc *fdp = p->p_fd;
635 	register int msk, i, j, fd;
636 	register fd_mask ibits, obits;
637 	struct file *fp;
638 	int n = 0;
639 	static int flag[3] = { POLLRDNORM | POLLHUP | POLLERR,
640 			       POLLWRNORM | POLLHUP | POLLERR,
641 			       POLLRDBAND };
642 
643 	for (msk = 0; msk < 3; msk++) {
644 		for (i = 0; i < nfd; i += NFDBITS) {
645 			ibits = *ibitp++;
646 			obits = 0;
647 			while ((j = ffs(ibits)) && (fd = i + --j) < nfd) {
648 				ibits &= ~(1 << j);
649 				fp = fdp->fd_ofiles[fd];
650 				if (fp == NULL)
651 					return (EBADF);
652 				if ((*fp->f_ops->fo_poll)(fp, flag[msk], p)) {
653 					obits |= (1 << j);
654 					n++;
655 				}
656 			}
657 			*obitp++ = obits;
658 		}
659 	}
660 	*retval = n;
661 	return (0);
662 }
663 
664 /*
665  * Poll system call.
666  */
667 int
668 sys_poll(p, v, retval)
669 	register struct proc *p;
670 	void *v;
671 	register_t *retval;
672 {
673 	register struct sys_poll_args /* {
674 		syscallarg(struct pollfd *) fds;
675 		syscallarg(u_int) nfds;
676 		syscallarg(int) timeout;
677 	} */ *uap = v;
678 	caddr_t bits;
679 	char smallbits[32 * sizeof(struct pollfd)];
680 	struct timeval atv;
681 	int s, ncoll, error = 0, timo;
682 	size_t ni;
683 
684 	if (SCARG(uap, nfds) > p->p_fd->fd_nfiles) {
685 		/* forgiving; slightly wrong */
686 		SCARG(uap, nfds) = p->p_fd->fd_nfiles;
687 	}
688 	ni = SCARG(uap, nfds) * sizeof(struct pollfd);
689 	if (ni > sizeof(smallbits))
690 		bits = malloc(ni, M_TEMP, M_WAITOK);
691 	else
692 		bits = smallbits;
693 
694 	error = copyin(SCARG(uap, fds), bits, ni);
695 	if (error)
696 		goto done;
697 
698 	if (SCARG(uap, timeout) != INFTIM) {
699 		atv.tv_sec = SCARG(uap, timeout) / 1000;
700 		atv.tv_usec = (SCARG(uap, timeout) % 1000) * 1000;
701 		if (itimerfix(&atv)) {
702 			error = EINVAL;
703 			goto done;
704 		}
705 		s = splclock();
706 		timeradd(&atv, &time, &atv);
707 		timo = hzto(&atv);
708 		/*
709 		 * Avoid inadvertently sleeping forever.
710 		 */
711 		if (timo == 0)
712 			timo = 1;
713 		splx(s);
714 	} else
715 		timo = 0;
716 retry:
717 	ncoll = nselcoll;
718 	p->p_flag |= P_SELECT;
719 	error = pollscan(p, (struct pollfd *)bits, SCARG(uap, nfds), retval);
720 	if (error || *retval)
721 		goto done;
722 	s = splhigh();
723 	if (timo && timercmp(&time, &atv, >=)) {
724 		splx(s);
725 		goto done;
726 	}
727 	if ((p->p_flag & P_SELECT) == 0 || nselcoll != ncoll) {
728 		splx(s);
729 		goto retry;
730 	}
731 	p->p_flag &= ~P_SELECT;
732 	error = tsleep((caddr_t)&selwait, PSOCK | PCATCH, "select", timo);
733 	splx(s);
734 	if (error == 0)
735 		goto retry;
736 done:
737 	p->p_flag &= ~P_SELECT;
738 	/* poll is not restarted after signals... */
739 	if (error == ERESTART)
740 		error = EINTR;
741 	if (error == EWOULDBLOCK)
742 		error = 0;
743 	if (error == 0) {
744 		error = copyout(bits, SCARG(uap, fds), ni);
745 		if (error)
746 			goto out;
747 	}
748 out:
749 	if (ni > sizeof(smallbits))
750 		free(bits, M_TEMP);
751 	return (error);
752 }
753 
754 int
755 pollscan(p, fds, nfd, retval)
756 	struct proc *p;
757 	struct pollfd *fds;
758 	int nfd;
759 	register_t *retval;
760 {
761 	register struct filedesc *fdp = p->p_fd;
762 	int i;
763 	struct file *fp;
764 	int n = 0;
765 
766 	for (i = 0; i < nfd; i++, fds++) {
767 		if ((u_int)fds->fd >= fdp->fd_nfiles) {
768 			fds->revents = POLLNVAL;
769 			n++;
770 		} else {
771 			fp = fdp->fd_ofiles[fds->fd];
772 			if (fp == 0) {
773 				fds->revents = POLLNVAL;
774 				n++;
775 			} else {
776 				fds->revents = (*fp->f_ops->fo_poll)(fp,
777 				    fds->events | POLLERR | POLLHUP, p);
778 				if (fds->revents != 0)
779 					n++;
780 			}
781 		}
782 	}
783 	*retval = n;
784 	return (0);
785 }
786 
787 /*ARGSUSED*/
788 int
789 seltrue(dev, events, p)
790 	dev_t dev;
791 	int events;
792 	struct proc *p;
793 {
794 
795 	return (events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
796 }
797 
798 /*
799  * Record a select request.
800  */
801 void
802 selrecord(selector, sip)
803 	struct proc *selector;
804 	struct selinfo *sip;
805 {
806 	struct proc *p;
807 	pid_t mypid;
808 
809 	mypid = selector->p_pid;
810 	if (sip->si_pid == mypid)
811 		return;
812 	if (sip->si_pid && (p = pfind(sip->si_pid)) &&
813 	    p->p_wchan == (caddr_t)&selwait)
814 		sip->si_flags |= SI_COLL;
815 	else
816 		sip->si_pid = mypid;
817 }
818 
819 /*
820  * Do a wakeup when a selectable event occurs.
821  */
822 void
823 selwakeup(sip)
824 	register struct selinfo *sip;
825 {
826 	register struct proc *p;
827 	int s;
828 
829 	if (sip->si_pid == 0)
830 		return;
831 	if (sip->si_flags & SI_COLL) {
832 		nselcoll++;
833 		sip->si_flags &= ~SI_COLL;
834 		wakeup((caddr_t)&selwait);
835 	}
836 	p = pfind(sip->si_pid);
837 	sip->si_pid = 0;
838 	if (p != NULL) {
839 		s = splhigh();
840 		if (p->p_wchan == (caddr_t)&selwait) {
841 			if (p->p_stat == SSLEEP)
842 				setrunnable(p);
843 			else
844 				unsleep(p);
845 		} else if (p->p_flag & P_SELECT)
846 			p->p_flag &= ~P_SELECT;
847 		splx(s);
848 	}
849 }
850