1 /* $NetBSD: sys_generic.c,v 1.82 2005/02/25 19:56:07 matt 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. Neither the name of the University nor the names of its contributors
21 * may be used to endorse or promote products derived from this software
22 * without specific prior written permission.
23 *
24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * SUCH DAMAGE.
35 *
36 * @(#)sys_generic.c 8.9 (Berkeley) 2/14/95
37 */
38
39 #include <sys/cdefs.h>
40 __KERNEL_RCSID(0, "$NetBSD: sys_generic.c,v 1.82 2005/02/25 19:56:07 matt Exp $");
41
42 #include "opt_ktrace.h"
43
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/filedesc.h>
47 #include <sys/ioctl.h>
48 #include <sys/file.h>
49 #include <sys/proc.h>
50 #include <sys/socketvar.h>
51 #include <sys/signalvar.h>
52 #include <sys/uio.h>
53 #include <sys/kernel.h>
54 #include <sys/stat.h>
55 #include <sys/malloc.h>
56 #include <sys/poll.h>
57 #ifdef KTRACE
58 #include <sys/ktrace.h>
59 #endif
60
61 #include <sys/mount.h>
62 #include <sys/sa.h>
63 #include <sys/syscallargs.h>
64
65 int selscan(struct proc *, fd_mask *, fd_mask *, int, register_t *);
66 int pollscan(struct proc *, struct pollfd *, int, register_t *);
67
68
69 /*
70 * Read system call.
71 */
72 /* ARGSUSED */
73 int
74 sys_read(struct lwp *l, void *v, register_t *retval)
75 {
76 struct sys_read_args /* {
77 syscallarg(int) fd;
78 syscallarg(void *) buf;
79 syscallarg(size_t) nbyte;
80 } */ *uap = v;
81 int fd;
82 struct file *fp;
83 struct proc *p;
84 struct filedesc *fdp;
85
86 fd = SCARG(uap, fd);
87 p = l->l_proc;
88 fdp = p->p_fd;
89
90 if ((fp = fd_getfile(fdp, fd)) == NULL)
91 return (EBADF);
92
93 if ((fp->f_flag & FREAD) == 0) {
94 simple_unlock(&fp->f_slock);
95 return (EBADF);
96 }
97
98 FILE_USE(fp);
99
100 /* dofileread() will unuse the descriptor for us */
101 return (dofileread(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
102 &fp->f_offset, FOF_UPDATE_OFFSET, retval));
103 }
104
105 int
106 dofileread(struct proc *p, int fd, struct file *fp, void *buf, size_t nbyte,
107 off_t *offset, int flags, register_t *retval)
108 {
109 struct uio auio;
110 struct iovec aiov;
111 size_t cnt;
112 int error;
113 #ifdef KTRACE
114 struct iovec ktriov = {0};
115 #endif
116 error = 0;
117
118 aiov.iov_base = (caddr_t)buf;
119 aiov.iov_len = nbyte;
120 auio.uio_iov = &aiov;
121 auio.uio_iovcnt = 1;
122 auio.uio_resid = nbyte;
123 auio.uio_rw = UIO_READ;
124 auio.uio_segflg = UIO_USERSPACE;
125 auio.uio_procp = p;
126
127 /*
128 * Reads return ssize_t because -1 is returned on error. Therefore
129 * we must restrict the length to SSIZE_MAX to avoid garbage return
130 * values.
131 */
132 if (auio.uio_resid > SSIZE_MAX) {
133 error = EINVAL;
134 goto out;
135 }
136
137 #ifdef KTRACE
138 /*
139 * if tracing, save a copy of iovec
140 */
141 if (KTRPOINT(p, KTR_GENIO))
142 ktriov = aiov;
143 #endif
144 cnt = auio.uio_resid;
145 error = (*fp->f_ops->fo_read)(fp, offset, &auio, fp->f_cred, flags);
146 if (error)
147 if (auio.uio_resid != cnt && (error == ERESTART ||
148 error == EINTR || error == EWOULDBLOCK))
149 error = 0;
150 cnt -= auio.uio_resid;
151 #ifdef KTRACE
152 if (KTRPOINT(p, KTR_GENIO) && error == 0)
153 ktrgenio(p, fd, UIO_READ, &ktriov, cnt, error);
154 #endif
155 *retval = cnt;
156 out:
157 FILE_UNUSE(fp, p);
158 return (error);
159 }
160
161 /*
162 * Scatter read system call.
163 */
164 int
165 sys_readv(struct lwp *l, void *v, register_t *retval)
166 {
167 struct sys_readv_args /* {
168 syscallarg(int) fd;
169 syscallarg(const struct iovec *) iovp;
170 syscallarg(int) iovcnt;
171 } */ *uap = v;
172 int fd;
173 struct file *fp;
174 struct proc *p;
175 struct filedesc *fdp;
176
177 fd = SCARG(uap, fd);
178 p = l->l_proc;
179 fdp = p->p_fd;
180
181 if ((fp = fd_getfile(fdp, fd)) == NULL)
182 return (EBADF);
183
184 if ((fp->f_flag & FREAD) == 0) {
185 simple_unlock(&fp->f_slock);
186 return (EBADF);
187 }
188
189 FILE_USE(fp);
190
191 /* dofilereadv() will unuse the descriptor for us */
192 return (dofilereadv(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt),
193 &fp->f_offset, FOF_UPDATE_OFFSET, retval));
194 }
195
196 int
197 dofilereadv(struct proc *p, int fd, struct file *fp, const struct iovec *iovp,
198 int iovcnt, off_t *offset, int flags, register_t *retval)
199 {
200 struct uio auio;
201 struct iovec *iov, *needfree, aiov[UIO_SMALLIOV];
202 int i, error;
203 size_t cnt;
204 u_int iovlen;
205 #ifdef KTRACE
206 struct iovec *ktriov;
207 #endif
208
209 error = 0;
210 #ifdef KTRACE
211 ktriov = NULL;
212 #endif
213 /* note: can't use iovlen until iovcnt is validated */
214 iovlen = iovcnt * sizeof(struct iovec);
215 if ((u_int)iovcnt > UIO_SMALLIOV) {
216 if ((u_int)iovcnt > IOV_MAX) {
217 error = EINVAL;
218 goto out;
219 }
220 iov = malloc(iovlen, M_IOV, M_WAITOK);
221 needfree = iov;
222 } else if ((u_int)iovcnt > 0) {
223 iov = aiov;
224 needfree = NULL;
225 } else {
226 error = EINVAL;
227 goto out;
228 }
229
230 auio.uio_iov = iov;
231 auio.uio_iovcnt = iovcnt;
232 auio.uio_rw = UIO_READ;
233 auio.uio_segflg = UIO_USERSPACE;
234 auio.uio_procp = p;
235 error = copyin(iovp, iov, iovlen);
236 if (error)
237 goto done;
238 auio.uio_resid = 0;
239 for (i = 0; i < iovcnt; i++) {
240 auio.uio_resid += iov->iov_len;
241 /*
242 * Reads return ssize_t because -1 is returned on error.
243 * Therefore we must restrict the length to SSIZE_MAX to
244 * avoid garbage return values.
245 */
246 if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
247 error = EINVAL;
248 goto done;
249 }
250 iov++;
251 }
252 #ifdef KTRACE
253 /*
254 * if tracing, save a copy of iovec
255 */
256 if (KTRPOINT(p, KTR_GENIO)) {
257 ktriov = malloc(iovlen, M_TEMP, M_WAITOK);
258 memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen);
259 }
260 #endif
261 cnt = auio.uio_resid;
262 error = (*fp->f_ops->fo_read)(fp, offset, &auio, fp->f_cred, flags);
263 if (error)
264 if (auio.uio_resid != cnt && (error == ERESTART ||
265 error == EINTR || error == EWOULDBLOCK))
266 error = 0;
267 cnt -= auio.uio_resid;
268 #ifdef KTRACE
269 if (ktriov != NULL) {
270 if (KTRPOINT(p, KTR_GENIO) && (error == 0))
271 ktrgenio(p, fd, UIO_READ, ktriov, cnt, error);
272 free(ktriov, M_TEMP);
273 }
274 #endif
275 *retval = cnt;
276 done:
277 if (needfree)
278 free(needfree, M_IOV);
279 out:
280 FILE_UNUSE(fp, p);
281 return (error);
282 }
283
284 /*
285 * Write system call
286 */
287 int
288 sys_write(struct lwp *l, void *v, register_t *retval)
289 {
290 struct sys_write_args /* {
291 syscallarg(int) fd;
292 syscallarg(const void *) buf;
293 syscallarg(size_t) nbyte;
294 } */ *uap = v;
295 int fd;
296 struct file *fp;
297 struct proc *p;
298 struct filedesc *fdp;
299
300 fd = SCARG(uap, fd);
301 p = l->l_proc;
302 fdp = p->p_fd;
303
304 if ((fp = fd_getfile(fdp, fd)) == NULL)
305 return (EBADF);
306
307 if ((fp->f_flag & FWRITE) == 0) {
308 simple_unlock(&fp->f_slock);
309 return (EBADF);
310 }
311
312 FILE_USE(fp);
313
314 /* dofilewrite() will unuse the descriptor for us */
315 return (dofilewrite(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte),
316 &fp->f_offset, FOF_UPDATE_OFFSET, retval));
317 }
318
319 int
320 dofilewrite(struct proc *p, int fd, struct file *fp, const void *buf,
321 size_t nbyte, off_t *offset, int flags, register_t *retval)
322 {
323 struct uio auio;
324 struct iovec aiov;
325 size_t cnt;
326 int error;
327 #ifdef KTRACE
328 struct iovec ktriov = {0};
329 #endif
330
331 error = 0;
332 aiov.iov_base = (caddr_t)buf; /* XXX kills const */
333 aiov.iov_len = nbyte;
334 auio.uio_iov = &aiov;
335 auio.uio_iovcnt = 1;
336 auio.uio_resid = nbyte;
337 auio.uio_rw = UIO_WRITE;
338 auio.uio_segflg = UIO_USERSPACE;
339 auio.uio_procp = p;
340
341 /*
342 * Writes return ssize_t because -1 is returned on error. Therefore
343 * we must restrict the length to SSIZE_MAX to avoid garbage return
344 * values.
345 */
346 if (auio.uio_resid > SSIZE_MAX) {
347 error = EINVAL;
348 goto out;
349 }
350
351 #ifdef KTRACE
352 /*
353 * if tracing, save a copy of iovec
354 */
355 if (KTRPOINT(p, KTR_GENIO))
356 ktriov = aiov;
357 #endif
358 cnt = auio.uio_resid;
359 error = (*fp->f_ops->fo_write)(fp, offset, &auio, fp->f_cred, flags);
360 if (error) {
361 if (auio.uio_resid != cnt && (error == ERESTART ||
362 error == EINTR || error == EWOULDBLOCK))
363 error = 0;
364 if (error == EPIPE)
365 psignal(p, SIGPIPE);
366 }
367 cnt -= auio.uio_resid;
368 #ifdef KTRACE
369 if (KTRPOINT(p, KTR_GENIO) && error == 0)
370 ktrgenio(p, fd, UIO_WRITE, &ktriov, cnt, error);
371 #endif
372 *retval = cnt;
373 out:
374 FILE_UNUSE(fp, p);
375 return (error);
376 }
377
378 /*
379 * Gather write system call
380 */
381 int
382 sys_writev(struct lwp *l, void *v, register_t *retval)
383 {
384 struct sys_writev_args /* {
385 syscallarg(int) fd;
386 syscallarg(const struct iovec *) iovp;
387 syscallarg(int) iovcnt;
388 } */ *uap = v;
389 int fd;
390 struct file *fp;
391 struct proc *p;
392 struct filedesc *fdp;
393
394 fd = SCARG(uap, fd);
395 p = l->l_proc;
396 fdp = p->p_fd;
397
398 if ((fp = fd_getfile(fdp, fd)) == NULL)
399 return (EBADF);
400
401 if ((fp->f_flag & FWRITE) == 0) {
402 simple_unlock(&fp->f_slock);
403 return (EBADF);
404 }
405
406 FILE_USE(fp);
407
408 /* dofilewritev() will unuse the descriptor for us */
409 return (dofilewritev(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt),
410 &fp->f_offset, FOF_UPDATE_OFFSET, retval));
411 }
412
413 int
414 dofilewritev(struct proc *p, int fd, struct file *fp, const struct iovec *iovp,
415 int iovcnt, off_t *offset, int flags, register_t *retval)
416 {
417 struct uio auio;
418 struct iovec *iov, *needfree, aiov[UIO_SMALLIOV];
419 int i, error;
420 size_t cnt;
421 u_int iovlen;
422 #ifdef KTRACE
423 struct iovec *ktriov;
424 #endif
425
426 error = 0;
427 #ifdef KTRACE
428 ktriov = NULL;
429 #endif
430 /* note: can't use iovlen until iovcnt is validated */
431 iovlen = iovcnt * sizeof(struct iovec);
432 if ((u_int)iovcnt > UIO_SMALLIOV) {
433 if ((u_int)iovcnt > IOV_MAX) {
434 error = EINVAL;
435 goto out;
436 }
437 iov = malloc(iovlen, M_IOV, M_WAITOK);
438 needfree = iov;
439 } else if ((u_int)iovcnt > 0) {
440 iov = aiov;
441 needfree = NULL;
442 } else {
443 error = EINVAL;
444 goto out;
445 }
446
447 auio.uio_iov = iov;
448 auio.uio_iovcnt = iovcnt;
449 auio.uio_rw = UIO_WRITE;
450 auio.uio_segflg = UIO_USERSPACE;
451 auio.uio_procp = p;
452 error = copyin(iovp, iov, iovlen);
453 if (error)
454 goto done;
455 auio.uio_resid = 0;
456 for (i = 0; i < iovcnt; i++) {
457 auio.uio_resid += iov->iov_len;
458 /*
459 * Writes return ssize_t because -1 is returned on error.
460 * Therefore we must restrict the length to SSIZE_MAX to
461 * avoid garbage return values.
462 */
463 if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) {
464 error = EINVAL;
465 goto done;
466 }
467 iov++;
468 }
469 #ifdef KTRACE
470 /*
471 * if tracing, save a copy of iovec
472 */
473 if (KTRPOINT(p, KTR_GENIO)) {
474 ktriov = malloc(iovlen, M_TEMP, M_WAITOK);
475 memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen);
476 }
477 #endif
478 cnt = auio.uio_resid;
479 error = (*fp->f_ops->fo_write)(fp, offset, &auio, fp->f_cred, flags);
480 if (error) {
481 if (auio.uio_resid != cnt && (error == ERESTART ||
482 error == EINTR || error == EWOULDBLOCK))
483 error = 0;
484 if (error == EPIPE)
485 psignal(p, SIGPIPE);
486 }
487 cnt -= auio.uio_resid;
488 #ifdef KTRACE
489 if (ktriov != NULL) {
490 if (KTRPOINT(p, KTR_GENIO) && (error == 0))
491 ktrgenio(p, fd, UIO_WRITE, ktriov, cnt, error);
492 free(ktriov, M_TEMP);
493 }
494 #endif
495 *retval = cnt;
496 done:
497 if (needfree)
498 free(needfree, M_IOV);
499 out:
500 FILE_UNUSE(fp, p);
501 return (error);
502 }
503
504 /*
505 * Ioctl system call
506 */
507 /* ARGSUSED */
508 int
509 sys_ioctl(struct lwp *l, void *v, register_t *retval)
510 {
511 struct sys_ioctl_args /* {
512 syscallarg(int) fd;
513 syscallarg(u_long) com;
514 syscallarg(caddr_t) data;
515 } */ *uap = v;
516 struct file *fp;
517 struct proc *p;
518 struct filedesc *fdp;
519 u_long com;
520 int error;
521 u_int size;
522 caddr_t data, memp;
523 #define STK_PARAMS 128
524 u_long stkbuf[STK_PARAMS/sizeof(u_long)];
525
526 error = 0;
527 p = l->l_proc;
528 fdp = p->p_fd;
529
530 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL)
531 return (EBADF);
532
533 FILE_USE(fp);
534
535 if ((fp->f_flag & (FREAD | FWRITE)) == 0) {
536 error = EBADF;
537 com = 0;
538 goto out;
539 }
540
541 switch (com = SCARG(uap, com)) {
542 case FIONCLEX:
543 fdp->fd_ofileflags[SCARG(uap, fd)] &= ~UF_EXCLOSE;
544 goto out;
545
546 case FIOCLEX:
547 fdp->fd_ofileflags[SCARG(uap, fd)] |= UF_EXCLOSE;
548 goto out;
549 }
550
551 /*
552 * Interpret high order word to find amount of data to be
553 * copied to/from the user's address space.
554 */
555 size = IOCPARM_LEN(com);
556 if (size > IOCPARM_MAX) {
557 error = ENOTTY;
558 goto out;
559 }
560 memp = NULL;
561 if (size > sizeof(stkbuf)) {
562 memp = (caddr_t)malloc((u_long)size, M_IOCTLOPS, M_WAITOK);
563 data = memp;
564 } else
565 data = (caddr_t)stkbuf;
566 if (com&IOC_IN) {
567 if (size) {
568 error = copyin(SCARG(uap, data), data, size);
569 if (error) {
570 if (memp)
571 free(memp, M_IOCTLOPS);
572 goto out;
573 }
574 #ifdef KTRACE
575 if (KTRPOINT(p, KTR_GENIO)) {
576 struct iovec iov;
577 iov.iov_base = SCARG(uap, data);
578 iov.iov_len = size;
579 ktrgenio(p, SCARG(uap, fd), UIO_WRITE, &iov,
580 size, 0);
581 }
582 #endif
583 } else
584 *(caddr_t *)data = SCARG(uap, data);
585 } else if ((com&IOC_OUT) && size)
586 /*
587 * Zero the buffer so the user always
588 * gets back something deterministic.
589 */
590 memset(data, 0, size);
591 else if (com&IOC_VOID)
592 *(caddr_t *)data = SCARG(uap, data);
593
594 switch (com) {
595
596 case FIONBIO:
597 if (*(int *)data != 0)
598 fp->f_flag |= FNONBLOCK;
599 else
600 fp->f_flag &= ~FNONBLOCK;
601 error = (*fp->f_ops->fo_ioctl)(fp, FIONBIO, data, p);
602 break;
603
604 case FIOASYNC:
605 if (*(int *)data != 0)
606 fp->f_flag |= FASYNC;
607 else
608 fp->f_flag &= ~FASYNC;
609 error = (*fp->f_ops->fo_ioctl)(fp, FIOASYNC, data, p);
610 break;
611
612 default:
613 error = (*fp->f_ops->fo_ioctl)(fp, com, data, p);
614 /*
615 * Copy any data to user, size was
616 * already set and checked above.
617 */
618 if (error == 0 && (com&IOC_OUT) && size) {
619 error = copyout(data, SCARG(uap, data), size);
620 #ifdef KTRACE
621 if (KTRPOINT(p, KTR_GENIO)) {
622 struct iovec iov;
623 iov.iov_base = SCARG(uap, data);
624 iov.iov_len = size;
625 ktrgenio(p, SCARG(uap, fd), UIO_READ, &iov,
626 size, error);
627 }
628 #endif
629 }
630 break;
631 }
632 if (memp)
633 free(memp, M_IOCTLOPS);
634 out:
635 FILE_UNUSE(fp, p);
636 switch (error) {
637 case -1:
638 printf("sys_ioctl: _IO%s%s('%c', %lu, %lu) returned -1: "
639 "pid=%d comm=%s\n",
640 (com & IOC_IN) ? "W" : "", (com & IOC_OUT) ? "R" : "",
641 (char)IOCGROUP(com), (com & 0xff), IOCPARM_LEN(com),
642 p->p_pid, p->p_comm);
643 /* FALLTHROUGH */
644 case EPASSTHROUGH:
645 error = ENOTTY;
646 /* FALLTHROUGH */
647 default:
648 return (error);
649 }
650 }
651
652 int selwait, nselcoll;
653
654 /*
655 * Select system call.
656 */
657 int
658 sys_pselect(struct lwp *l, void *v, register_t *retval)
659 {
660 struct sys_pselect_args /* {
661 syscallarg(int) nd;
662 syscallarg(fd_set *) in;
663 syscallarg(fd_set *) ou;
664 syscallarg(fd_set *) ex;
665 syscallarg(const struct timespec *) ts;
666 syscallarg(sigset_t *) mask;
667 } */ * const uap = v;
668 struct timespec ats;
669 struct timeval atv, *tv = NULL;
670 sigset_t amask, *mask = NULL;
671 int error;
672
673 if (SCARG(uap, ts)) {
674 error = copyin(SCARG(uap, ts), &ats, sizeof(ats));
675 if (error)
676 return error;
677 atv.tv_sec = ats.tv_sec;
678 atv.tv_usec = ats.tv_nsec / 1000;
679 tv = &atv;
680 }
681 if (SCARG(uap, mask) != NULL) {
682 error = copyin(SCARG(uap, mask), &amask, sizeof(amask));
683 if (error)
684 return error;
685 mask = &amask;
686 }
687
688 return selcommon(l, retval, SCARG(uap, nd), SCARG(uap, in),
689 SCARG(uap, ou), SCARG(uap, ex), tv, mask);
690 }
691
692 int
693 sys_select(struct lwp *l, void *v, register_t *retval)
694 {
695 struct sys_select_args /* {
696 syscallarg(int) nd;
697 syscallarg(fd_set *) in;
698 syscallarg(fd_set *) ou;
699 syscallarg(fd_set *) ex;
700 syscallarg(struct timeval *) tv;
701 } */ * const uap = v;
702 struct timeval atv, *tv = NULL;
703 int error;
704
705 if (SCARG(uap, tv)) {
706 error = copyin(SCARG(uap, tv), (caddr_t)&atv,
707 sizeof(atv));
708 if (error)
709 return error;
710 tv = &atv;
711 }
712
713 return selcommon(l, retval, SCARG(uap, nd), SCARG(uap, in),
714 SCARG(uap, ou), SCARG(uap, ex), tv, NULL);
715 }
716
717 int
718 selcommon(struct lwp *l, register_t *retval, int nd, fd_set *u_in,
719 fd_set *u_ou, fd_set *u_ex, struct timeval *tv, sigset_t *mask)
720 {
721 struct proc * const p = l->l_proc;
722 caddr_t bits;
723 char smallbits[howmany(FD_SETSIZE, NFDBITS) *
724 sizeof(fd_mask) * 6];
725 int s, ncoll, error, timo;
726 size_t ni;
727 sigset_t oldmask;
728
729 error = 0;
730 if (nd < 0)
731 return (EINVAL);
732 if (nd > p->p_fd->fd_nfiles) {
733 /* forgiving; slightly wrong */
734 nd = p->p_fd->fd_nfiles;
735 }
736 ni = howmany(nd, NFDBITS) * sizeof(fd_mask);
737 if (ni * 6 > sizeof(smallbits))
738 bits = malloc(ni * 6, M_TEMP, M_WAITOK);
739 else
740 bits = smallbits;
741
742 #define getbits(name, x) \
743 if (u_ ## name) { \
744 error = copyin(u_ ## name, bits + ni * x, ni); \
745 if (error) \
746 goto done; \
747 } else \
748 memset(bits + ni * x, 0, ni);
749 getbits(in, 0);
750 getbits(ou, 1);
751 getbits(ex, 2);
752 #undef getbits
753
754 timo = 0;
755 if (tv) {
756 if (itimerfix(tv)) {
757 error = EINVAL;
758 goto done;
759 }
760 s = splclock();
761 timeradd(tv, &time, tv);
762 splx(s);
763 }
764 if (mask)
765 (void)sigprocmask1(p, SIG_SETMASK, mask, &oldmask);
766
767 retry:
768 ncoll = nselcoll;
769 l->l_flag |= L_SELECT;
770 error = selscan(p, (fd_mask *)(bits + ni * 0),
771 (fd_mask *)(bits + ni * 3), nd, retval);
772 if (error || *retval)
773 goto done;
774 if (tv) {
775 /*
776 * We have to recalculate the timeout on every retry.
777 */
778 timo = hzto(tv);
779 if (timo <= 0)
780 goto done;
781 }
782 s = splsched();
783 if ((l->l_flag & L_SELECT) == 0 || nselcoll != ncoll) {
784 splx(s);
785 goto retry;
786 }
787 l->l_flag &= ~L_SELECT;
788 error = tsleep((caddr_t)&selwait, PSOCK | PCATCH, "select", timo);
789 splx(s);
790 if (error == 0)
791 goto retry;
792 done:
793 if (mask)
794 (void)sigprocmask1(p, SIG_SETMASK, &oldmask, NULL);
795 l->l_flag &= ~L_SELECT;
796 /* select is not restarted after signals... */
797 if (error == ERESTART)
798 error = EINTR;
799 if (error == EWOULDBLOCK)
800 error = 0;
801 if (error == 0) {
802
803 #define putbits(name, x) \
804 if (u_ ## name) { \
805 error = copyout(bits + ni * x, u_ ## name, ni); \
806 if (error) \
807 goto out; \
808 }
809 putbits(in, 3);
810 putbits(ou, 4);
811 putbits(ex, 5);
812 #undef putbits
813 }
814 out:
815 if (ni * 6 > sizeof(smallbits))
816 free(bits, M_TEMP);
817 return (error);
818 }
819
820 int
821 selscan(struct proc *p, fd_mask *ibitp, fd_mask *obitp, int nfd,
822 register_t *retval)
823 {
824 struct filedesc *fdp;
825 int msk, i, j, fd, n;
826 fd_mask ibits, obits;
827 struct file *fp;
828 static const int flag[3] = { POLLRDNORM | POLLHUP | POLLERR,
829 POLLWRNORM | POLLHUP | POLLERR,
830 POLLRDBAND };
831
832 fdp = p->p_fd;
833 n = 0;
834 for (msk = 0; msk < 3; msk++) {
835 for (i = 0; i < nfd; i += NFDBITS) {
836 ibits = *ibitp++;
837 obits = 0;
838 while ((j = ffs(ibits)) && (fd = i + --j) < nfd) {
839 ibits &= ~(1 << j);
840 if ((fp = fd_getfile(fdp, fd)) == NULL)
841 return (EBADF);
842 FILE_USE(fp);
843 if ((*fp->f_ops->fo_poll)(fp, flag[msk], p)) {
844 obits |= (1 << j);
845 n++;
846 }
847 FILE_UNUSE(fp, p);
848 }
849 *obitp++ = obits;
850 }
851 }
852 *retval = n;
853 return (0);
854 }
855
856 /*
857 * Poll system call.
858 */
859 int
860 sys_poll(struct lwp *l, void *v, register_t *retval)
861 {
862 struct sys_poll_args /* {
863 syscallarg(struct pollfd *) fds;
864 syscallarg(u_int) nfds;
865 syscallarg(int) timeout;
866 } */ * const uap = v;
867 struct timeval atv, *tv = NULL;
868
869 if (SCARG(uap, timeout) != INFTIM) {
870 atv.tv_sec = SCARG(uap, timeout) / 1000;
871 atv.tv_usec = (SCARG(uap, timeout) % 1000) * 1000;
872 tv = &atv;
873 }
874
875 return pollcommon(l, retval, SCARG(uap, fds), SCARG(uap, nfds),
876 tv, NULL);
877 }
878
879 /*
880 * Poll system call.
881 */
882 int
883 sys_pollts(struct lwp *l, void *v, register_t *retval)
884 {
885 struct sys_pollts_args /* {
886 syscallarg(struct pollfd *) fds;
887 syscallarg(u_int) nfds;
888 syscallarg(const struct timespec *) ts;
889 syscallarg(const sigset_t *) mask;
890 } */ * const uap = v;
891 struct timespec ats;
892 struct timeval atv, *tv = NULL;
893 sigset_t amask, *mask = NULL;
894 int error;
895
896 if (SCARG(uap, ts)) {
897 error = copyin(SCARG(uap, ts), &ats, sizeof(ats));
898 if (error)
899 return error;
900 atv.tv_sec = ats.tv_sec;
901 atv.tv_usec = ats.tv_nsec / 1000;
902 tv = &atv;
903 }
904 if (SCARG(uap, mask)) {
905 error = copyin(SCARG(uap, mask), &amask, sizeof(amask));
906 if (error)
907 return error;
908 mask = &amask;
909 }
910
911 return pollcommon(l, retval, SCARG(uap, fds), SCARG(uap, nfds),
912 tv, mask);
913 }
914
915 int
916 pollcommon(struct lwp *l, register_t *retval,
917 struct pollfd *u_fds, u_int nfds,
918 struct timeval *tv, sigset_t *mask)
919 {
920 struct proc * const p = l->l_proc;
921 caddr_t bits;
922 char smallbits[32 * sizeof(struct pollfd)];
923 sigset_t oldmask;
924 int s, ncoll, error, timo;
925 size_t ni;
926
927 if (nfds > p->p_fd->fd_nfiles) {
928 /* forgiving; slightly wrong */
929 nfds = p->p_fd->fd_nfiles;
930 }
931 ni = nfds * sizeof(struct pollfd);
932 if (ni > sizeof(smallbits))
933 bits = malloc(ni, M_TEMP, M_WAITOK);
934 else
935 bits = smallbits;
936
937 error = copyin(u_fds, bits, ni);
938 if (error)
939 goto done;
940
941 timo = 0;
942 if (tv) {
943 if (itimerfix(tv)) {
944 error = EINVAL;
945 goto done;
946 }
947 s = splclock();
948 timeradd(tv, &time, tv);
949 splx(s);
950 }
951 if (mask != NULL)
952 (void)sigprocmask1(p, SIG_SETMASK, mask, &oldmask);
953
954 retry:
955 ncoll = nselcoll;
956 l->l_flag |= L_SELECT;
957 error = pollscan(p, (struct pollfd *)bits, nfds, retval);
958 if (error || *retval)
959 goto done;
960 if (tv) {
961 /*
962 * We have to recalculate the timeout on every retry.
963 */
964 timo = hzto(tv);
965 if (timo <= 0)
966 goto done;
967 }
968 s = splsched();
969 if ((l->l_flag & L_SELECT) == 0 || nselcoll != ncoll) {
970 splx(s);
971 goto retry;
972 }
973 l->l_flag &= ~L_SELECT;
974 error = tsleep((caddr_t)&selwait, PSOCK | PCATCH, "poll", timo);
975 splx(s);
976 if (error == 0)
977 goto retry;
978 done:
979 if (mask != NULL)
980 (void)sigprocmask1(p, SIG_SETMASK, &oldmask, NULL);
981 l->l_flag &= ~L_SELECT;
982 /* poll is not restarted after signals... */
983 if (error == ERESTART)
984 error = EINTR;
985 if (error == EWOULDBLOCK)
986 error = 0;
987 if (error == 0) {
988 error = copyout(bits, u_fds, ni);
989 if (error)
990 goto out;
991 }
992 out:
993 if (ni > sizeof(smallbits))
994 free(bits, M_TEMP);
995 return (error);
996 }
997
998 int
999 pollscan(struct proc *p, struct pollfd *fds, int nfd, register_t *retval)
1000 {
1001 struct filedesc *fdp;
1002 int i, n;
1003 struct file *fp;
1004
1005 fdp = p->p_fd;
1006 n = 0;
1007 for (i = 0; i < nfd; i++, fds++) {
1008 if (fds->fd >= fdp->fd_nfiles) {
1009 fds->revents = POLLNVAL;
1010 n++;
1011 } else if (fds->fd < 0) {
1012 fds->revents = 0;
1013 } else {
1014 if ((fp = fd_getfile(fdp, fds->fd)) == NULL) {
1015 fds->revents = POLLNVAL;
1016 n++;
1017 } else {
1018 FILE_USE(fp);
1019 fds->revents = (*fp->f_ops->fo_poll)(fp,
1020 fds->events | POLLERR | POLLHUP, p);
1021 if (fds->revents != 0)
1022 n++;
1023 FILE_UNUSE(fp, p);
1024 }
1025 }
1026 }
1027 *retval = n;
1028 return (0);
1029 }
1030
1031 /*ARGSUSED*/
1032 int
1033 seltrue(dev_t dev, int events, struct proc *p)
1034 {
1035
1036 return (events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
1037 }
1038
1039 /*
1040 * Record a select request.
1041 */
1042 void
1043 selrecord(struct proc *selector, struct selinfo *sip)
1044 {
1045 struct lwp *l;
1046 struct proc *p;
1047 pid_t mypid;
1048
1049 mypid = selector->p_pid;
1050 if (sip->sel_pid == mypid)
1051 return;
1052 if (sip->sel_pid && (p = pfind(sip->sel_pid))) {
1053 LIST_FOREACH(l, &p->p_lwps, l_sibling) {
1054 if (l->l_wchan == (caddr_t)&selwait) {
1055 sip->sel_collision = 1;
1056 return;
1057 }
1058 }
1059 }
1060
1061 sip->sel_pid = mypid;
1062 }
1063
1064 /*
1065 * Do a wakeup when a selectable event occurs.
1066 */
1067 void
1068 selwakeup(sip)
1069 struct selinfo *sip;
1070 {
1071 struct lwp *l;
1072 struct proc *p;
1073 int s;
1074
1075 if (sip->sel_pid == 0)
1076 return;
1077 if (sip->sel_collision) {
1078 sip->sel_pid = 0;
1079 nselcoll++;
1080 sip->sel_collision = 0;
1081 wakeup((caddr_t)&selwait);
1082 return;
1083 }
1084 p = pfind(sip->sel_pid);
1085 sip->sel_pid = 0;
1086 if (p != NULL) {
1087 LIST_FOREACH(l, &p->p_lwps, l_sibling) {
1088 SCHED_LOCK(s);
1089 if (l->l_wchan == (caddr_t)&selwait) {
1090 if (l->l_stat == LSSLEEP)
1091 setrunnable(l);
1092 else
1093 unsleep(l);
1094 } else if (l->l_flag & L_SELECT)
1095 l->l_flag &= ~L_SELECT;
1096 SCHED_UNLOCK(s);
1097 }
1098 }
1099 }
Cache object: 68c5cd8106e8c9b63cbc114b136d6b41
|