1 /* $OpenBSD: kern_descrip.c,v 1.210 2024/12/30 02:46:00 guenther Exp $ */
2 /* $NetBSD: kern_descrip.c,v 1.42 1996/03/30 22:24:38 christos Exp $ */
3
4 /*
5 * Copyright (c) 1982, 1986, 1989, 1991, 1993
6 * The Regents of the University of California. All rights reserved.
7 * (c) UNIX System Laboratories, Inc.
8 * All or some portions of this file are derived from material licensed
9 * to the University of California by American Telephone and Telegraph
10 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
11 * the permission of UNIX System Laboratories, Inc.
12 *
13 * Redistribution and use in source and binary forms, with or without
14 * modification, are permitted provided that the following conditions
15 * are met:
16 * 1. Redistributions of source code must retain the above copyright
17 * notice, this list of conditions and the following disclaimer.
18 * 2. Redistributions in binary form must reproduce the above copyright
19 * notice, this list of conditions and the following disclaimer in the
20 * documentation and/or other materials provided with the distribution.
21 * 3. Neither the name of the University nor the names of its contributors
22 * may be used to endorse or promote products derived from this software
23 * without specific prior written permission.
24 *
25 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
26 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
30 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
31 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35 * SUCH DAMAGE.
36 *
37 * @(#)kern_descrip.c 8.6 (Berkeley) 4/19/94
38 */
39
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/filedesc.h>
43 #include <sys/vnode.h>
44 #include <sys/proc.h>
45 #include <sys/file.h>
46 #include <sys/socket.h>
47 #include <sys/stat.h>
48 #include <sys/ioctl.h>
49 #include <sys/fcntl.h>
50 #include <sys/lock.h>
51 #include <sys/malloc.h>
52 #include <sys/ucred.h>
53 #include <sys/unistd.h>
54 #include <sys/resourcevar.h>
55 #include <sys/mount.h>
56 #include <sys/syscallargs.h>
57 #include <sys/event.h>
58 #include <sys/pool.h>
59 #include <sys/ktrace.h>
60 #include <sys/pledge.h>
61
62 /*
63 * Descriptor management.
64 *
65 * We need to block interrupts as long as `fhdlk' is being taken
66 * with and without the KERNEL_LOCK().
67 */
68 struct mutex fhdlk = MUTEX_INITIALIZER(IPL_MPFLOOR);
69 struct filelist filehead; /* head of list of open files */
70 int numfiles; /* actual number of open files */
71
72 static __inline void fd_used(struct filedesc *, int);
73 static __inline void fd_unused(struct filedesc *, int);
74 static __inline int find_next_zero(u_int *, int, u_int);
75 static __inline int fd_inuse(struct filedesc *, int);
76 int finishdup(struct proc *, struct file *, int, int, register_t *, int);
77 int find_last_set(struct filedesc *, int);
78 int dodup3(struct proc *, int, int, int, register_t *);
79
80 #define DUPF_CLOEXEC 0x01
81 #define DUPF_DUP2 0x02
82
83 struct pool file_pool;
84 struct pool fdesc_pool;
85
86 void
filedesc_init(void)87 filedesc_init(void)
88 {
89 pool_init(&file_pool, sizeof(struct file), 0, IPL_MPFLOOR,
90 PR_WAITOK, "filepl", NULL);
91 pool_init(&fdesc_pool, sizeof(struct filedesc0), 0, IPL_NONE,
92 PR_WAITOK, "fdescpl", NULL);
93 LIST_INIT(&filehead);
94 }
95
96 static __inline int
find_next_zero(u_int * bitmap,int want,u_int bits)97 find_next_zero(u_int *bitmap, int want, u_int bits)
98 {
99 int i, off, maxoff;
100 u_int sub;
101
102 if (want > bits)
103 return -1;
104
105 off = want >> NDENTRYSHIFT;
106 i = want & NDENTRYMASK;
107 if (i) {
108 sub = bitmap[off] | ((u_int)~0 >> (NDENTRIES - i));
109 if (sub != ~0)
110 goto found;
111 off++;
112 }
113
114 maxoff = NDLOSLOTS(bits);
115 while (off < maxoff) {
116 if ((sub = bitmap[off]) != ~0)
117 goto found;
118 off++;
119 }
120
121 return -1;
122
123 found:
124 return (off << NDENTRYSHIFT) + ffs(~sub) - 1;
125 }
126
127 int
find_last_set(struct filedesc * fd,int last)128 find_last_set(struct filedesc *fd, int last)
129 {
130 int off, i;
131 u_int *bitmap = fd->fd_lomap;
132
133 off = (last - 1) >> NDENTRYSHIFT;
134
135 while (off >= 0 && !bitmap[off])
136 off--;
137 if (off < 0)
138 return 0;
139
140 i = ((off + 1) << NDENTRYSHIFT) - 1;
141 if (i >= last)
142 i = last - 1;
143
144 while (i > 0 && !fd_inuse(fd, i))
145 i--;
146 return i;
147 }
148
149 static __inline int
fd_inuse(struct filedesc * fdp,int fd)150 fd_inuse(struct filedesc *fdp, int fd)
151 {
152 u_int off = fd >> NDENTRYSHIFT;
153
154 if (fdp->fd_lomap[off] & (1U << (fd & NDENTRYMASK)))
155 return 1;
156
157 return 0;
158 }
159
160 static __inline void
fd_used(struct filedesc * fdp,int fd)161 fd_used(struct filedesc *fdp, int fd)
162 {
163 u_int off = fd >> NDENTRYSHIFT;
164
165 fdp->fd_lomap[off] |= 1U << (fd & NDENTRYMASK);
166 if (fdp->fd_lomap[off] == ~0)
167 fdp->fd_himap[off >> NDENTRYSHIFT] |= 1U << (off & NDENTRYMASK);
168
169 if (fd > fdp->fd_lastfile)
170 fdp->fd_lastfile = fd;
171 fdp->fd_openfd++;
172 }
173
174 static __inline void
fd_unused(struct filedesc * fdp,int fd)175 fd_unused(struct filedesc *fdp, int fd)
176 {
177 u_int off = fd >> NDENTRYSHIFT;
178
179 if (fd < fdp->fd_freefile)
180 fdp->fd_freefile = fd;
181
182 if (fdp->fd_lomap[off] == ~0)
183 fdp->fd_himap[off >> NDENTRYSHIFT] &= ~(1U << (off & NDENTRYMASK));
184 fdp->fd_lomap[off] &= ~(1U << (fd & NDENTRYMASK));
185
186 #ifdef DIAGNOSTIC
187 if (fd > fdp->fd_lastfile)
188 panic("fd_unused: fd_lastfile inconsistent");
189 #endif
190 if (fd == fdp->fd_lastfile)
191 fdp->fd_lastfile = find_last_set(fdp, fd);
192 fdp->fd_openfd--;
193 }
194
195 struct file *
fd_iterfile(struct file * fp,struct proc * p)196 fd_iterfile(struct file *fp, struct proc *p)
197 {
198 struct file *nfp;
199 unsigned int count;
200
201 mtx_enter(&fhdlk);
202 if (fp == NULL)
203 nfp = LIST_FIRST(&filehead);
204 else
205 nfp = LIST_NEXT(fp, f_list);
206
207 /* don't refcount when f_count == 0 to avoid race in fdrop() */
208 while (nfp != NULL) {
209 count = nfp->f_count;
210 if (count == 0) {
211 nfp = LIST_NEXT(nfp, f_list);
212 continue;
213 }
214 if (atomic_cas_uint(&nfp->f_count, count, count + 1) == count)
215 break;
216 }
217 mtx_leave(&fhdlk);
218
219 if (fp != NULL)
220 FRELE(fp, p);
221
222 return nfp;
223 }
224
225 struct file *
fd_getfile(struct filedesc * fdp,int fd)226 fd_getfile(struct filedesc *fdp, int fd)
227 {
228 struct file *fp;
229
230 vfs_stall_barrier();
231
232 if ((u_int)fd >= fdp->fd_nfiles)
233 return (NULL);
234
235 mtx_enter(&fdp->fd_fplock);
236 fp = fdp->fd_ofiles[fd];
237 if (fp != NULL)
238 atomic_inc_int(&fp->f_count);
239 mtx_leave(&fdp->fd_fplock);
240
241 return (fp);
242 }
243
244 struct file *
fd_getfile_mode(struct filedesc * fdp,int fd,int mode)245 fd_getfile_mode(struct filedesc *fdp, int fd, int mode)
246 {
247 struct file *fp;
248
249 KASSERT(mode != 0);
250
251 fp = fd_getfile(fdp, fd);
252 if (fp == NULL)
253 return (NULL);
254
255 if ((fp->f_flag & mode) == 0) {
256 FRELE(fp, curproc);
257 return (NULL);
258 }
259
260 return (fp);
261 }
262
263 int
fd_checkclosed(struct filedesc * fdp,int fd,struct file * fp)264 fd_checkclosed(struct filedesc *fdp, int fd, struct file *fp)
265 {
266 int closed;
267
268 mtx_enter(&fdp->fd_fplock);
269 KASSERT(fd < fdp->fd_nfiles);
270 closed = (fdp->fd_ofiles[fd] != fp);
271 mtx_leave(&fdp->fd_fplock);
272 return (closed);
273 }
274
275 /*
276 * System calls on descriptors.
277 */
278
279 /*
280 * Duplicate a file descriptor.
281 */
282 int
sys_dup(struct proc * p,void * v,register_t * retval)283 sys_dup(struct proc *p, void *v, register_t *retval)
284 {
285 struct sys_dup_args /* {
286 syscallarg(int) fd;
287 } */ *uap = v;
288 struct filedesc *fdp = p->p_fd;
289 int old = SCARG(uap, fd);
290 struct file *fp;
291 int new;
292 int error;
293
294 restart:
295 if ((fp = fd_getfile(fdp, old)) == NULL)
296 return (EBADF);
297 fdplock(fdp);
298 if ((error = fdalloc(p, 0, &new)) != 0) {
299 if (error == ENOSPC) {
300 fdexpand(p);
301 fdpunlock(fdp);
302 FRELE(fp, p);
303 goto restart;
304 }
305 fdpunlock(fdp);
306 FRELE(fp, p);
307 return (error);
308 }
309 /* No need for FRELE(), finishdup() uses current ref. */
310 return (finishdup(p, fp, old, new, retval, 0));
311 }
312
313 /*
314 * Duplicate a file descriptor to a particular value.
315 */
316 int
sys_dup2(struct proc * p,void * v,register_t * retval)317 sys_dup2(struct proc *p, void *v, register_t *retval)
318 {
319 struct sys_dup2_args /* {
320 syscallarg(int) from;
321 syscallarg(int) to;
322 } */ *uap = v;
323
324 return (dodup3(p, SCARG(uap, from), SCARG(uap, to), 0, retval));
325 }
326
327 int
sys_dup3(struct proc * p,void * v,register_t * retval)328 sys_dup3(struct proc *p, void *v, register_t *retval)
329 {
330 struct sys_dup3_args /* {
331 syscallarg(int) from;
332 syscallarg(int) to;
333 syscallarg(int) flags;
334 } */ *uap = v;
335
336 if (SCARG(uap, from) == SCARG(uap, to))
337 return (EINVAL);
338 if (SCARG(uap, flags) & ~O_CLOEXEC)
339 return (EINVAL);
340 return (dodup3(p, SCARG(uap, from), SCARG(uap, to),
341 SCARG(uap, flags), retval));
342 }
343
344 int
dodup3(struct proc * p,int old,int new,int flags,register_t * retval)345 dodup3(struct proc *p, int old, int new, int flags, register_t *retval)
346 {
347 struct filedesc *fdp = p->p_fd;
348 struct file *fp;
349 int dupflags, error, i;
350
351 restart:
352 if ((fp = fd_getfile(fdp, old)) == NULL)
353 return (EBADF);
354 if (old == new) {
355 /*
356 * NOTE! This doesn't clear the close-on-exec flag. This might
357 * or might not be the intended behavior from the start, but
358 * this is what everyone else does.
359 */
360 *retval = new;
361 FRELE(fp, p);
362 return (0);
363 }
364 if ((u_int)new >= lim_cur(RLIMIT_NOFILE) ||
365 (u_int)new >= atomic_load_int(&maxfiles)) {
366 FRELE(fp, p);
367 return (EBADF);
368 }
369 fdplock(fdp);
370 if (new >= fdp->fd_nfiles) {
371 if ((error = fdalloc(p, new, &i)) != 0) {
372 if (error == ENOSPC) {
373 fdexpand(p);
374 fdpunlock(fdp);
375 FRELE(fp, p);
376 goto restart;
377 }
378 fdpunlock(fdp);
379 FRELE(fp, p);
380 return (error);
381 }
382 if (new != i)
383 panic("dup2: fdalloc");
384 fd_unused(fdp, new);
385 }
386
387 dupflags = DUPF_DUP2;
388 if (flags & O_CLOEXEC)
389 dupflags |= DUPF_CLOEXEC;
390
391 /* No need for FRELE(), finishdup() uses current ref. */
392 return (finishdup(p, fp, old, new, retval, dupflags));
393 }
394
395 /*
396 * The file control system call.
397 */
398 int
sys_fcntl(struct proc * p,void * v,register_t * retval)399 sys_fcntl(struct proc *p, void *v, register_t *retval)
400 {
401 struct sys_fcntl_args /* {
402 syscallarg(int) fd;
403 syscallarg(int) cmd;
404 syscallarg(void *) arg;
405 } */ *uap = v;
406 int fd = SCARG(uap, fd);
407 struct filedesc *fdp = p->p_fd;
408 struct file *fp;
409 struct vnode *vp;
410 int i, prev, tmp, newmin, flg = F_POSIX;
411 struct flock fl;
412 int error = 0;
413
414 error = pledge_fcntl(p, SCARG(uap, cmd));
415 if (error)
416 return (error);
417
418 restart:
419 if ((fp = fd_getfile(fdp, fd)) == NULL)
420 return (EBADF);
421 switch (SCARG(uap, cmd)) {
422
423 case F_DUPFD:
424 case F_DUPFD_CLOEXEC:
425 newmin = (long)SCARG(uap, arg);
426 if ((u_int)newmin >= lim_cur(RLIMIT_NOFILE) ||
427 (u_int)newmin >= atomic_load_int(&maxfiles)) {
428 error = EINVAL;
429 break;
430 }
431 fdplock(fdp);
432 if ((error = fdalloc(p, newmin, &i)) != 0) {
433 if (error == ENOSPC) {
434 fdexpand(p);
435 fdpunlock(fdp);
436 FRELE(fp, p);
437 goto restart;
438 }
439 fdpunlock(fdp);
440 FRELE(fp, p);
441 } else {
442 int dupflags = 0;
443
444 if (SCARG(uap, cmd) == F_DUPFD_CLOEXEC)
445 dupflags |= DUPF_CLOEXEC;
446
447 /* No need for FRELE(), finishdup() uses current ref. */
448 error = finishdup(p, fp, fd, i, retval, dupflags);
449 }
450 return (error);
451
452 case F_GETFD:
453 fdplock(fdp);
454 *retval = fdp->fd_ofileflags[fd] & UF_EXCLOSE ? 1 : 0;
455 fdpunlock(fdp);
456 break;
457
458 case F_SETFD:
459 fdplock(fdp);
460 if ((long)SCARG(uap, arg) & 1)
461 fdp->fd_ofileflags[fd] |= UF_EXCLOSE;
462 else
463 fdp->fd_ofileflags[fd] &= ~UF_EXCLOSE;
464 fdpunlock(fdp);
465 break;
466
467 case F_GETFL:
468 *retval = OFLAGS(fp->f_flag);
469 break;
470
471 case F_ISATTY:
472 vp = fp->f_data;
473 if (fp->f_type == DTYPE_VNODE && (vp->v_flag & VISTTY))
474 *retval = 1;
475 else {
476 *retval = 0;
477 error = ENOTTY;
478 }
479 break;
480
481 case F_SETFL:
482 do {
483 tmp = prev = fp->f_flag;
484 tmp &= ~FCNTLFLAGS;
485 tmp |= FFLAGS((long)SCARG(uap, arg)) & FCNTLFLAGS;
486 } while (atomic_cas_uint(&fp->f_flag, prev, tmp) != prev);
487 tmp = fp->f_flag & FASYNC;
488 error = (*fp->f_ops->fo_ioctl)(fp, FIOASYNC, (caddr_t)&tmp, p);
489 break;
490
491 case F_GETOWN:
492 tmp = 0;
493 error = (*fp->f_ops->fo_ioctl)
494 (fp, FIOGETOWN, (caddr_t)&tmp, p);
495 *retval = tmp;
496 break;
497
498 case F_SETOWN:
499 tmp = (long)SCARG(uap, arg);
500 error = ((*fp->f_ops->fo_ioctl)
501 (fp, FIOSETOWN, (caddr_t)&tmp, p));
502 break;
503
504 case F_SETLKW:
505 flg |= F_WAIT;
506 /* FALLTHROUGH */
507
508 case F_SETLK:
509 error = pledge_flock(p);
510 if (error != 0)
511 break;
512
513 if (fp->f_type != DTYPE_VNODE) {
514 error = EINVAL;
515 break;
516 }
517 vp = fp->f_data;
518 /* Copy in the lock structure */
519 error = copyin((caddr_t)SCARG(uap, arg), (caddr_t)&fl,
520 sizeof (fl));
521 if (error)
522 break;
523 #ifdef KTRACE
524 if (KTRPOINT(p, KTR_STRUCT))
525 ktrflock(p, &fl);
526 #endif
527 if (fl.l_whence == SEEK_CUR) {
528 off_t offset = foffset(fp);
529
530 if (fl.l_start == 0 && fl.l_len < 0) {
531 /* lockf(3) compliance hack */
532 fl.l_len = -fl.l_len;
533 fl.l_start = offset - fl.l_len;
534 } else
535 fl.l_start += offset;
536 }
537 switch (fl.l_type) {
538
539 case F_RDLCK:
540 if ((fp->f_flag & FREAD) == 0) {
541 error = EBADF;
542 goto out;
543 }
544 atomic_setbits_int(&fdp->fd_flags, FD_ADVLOCK);
545 error = VOP_ADVLOCK(vp, fdp, F_SETLK, &fl, flg);
546 break;
547
548 case F_WRLCK:
549 if ((fp->f_flag & FWRITE) == 0) {
550 error = EBADF;
551 goto out;
552 }
553 atomic_setbits_int(&fdp->fd_flags, FD_ADVLOCK);
554 error = VOP_ADVLOCK(vp, fdp, F_SETLK, &fl, flg);
555 break;
556
557 case F_UNLCK:
558 error = VOP_ADVLOCK(vp, fdp, F_UNLCK, &fl, F_POSIX);
559 goto out;
560
561 default:
562 error = EINVAL;
563 goto out;
564 }
565
566 if (fd_checkclosed(fdp, fd, fp)) {
567 /*
568 * We have lost the race with close() or dup2();
569 * unlock, pretend that we've won the race and that
570 * lock had been removed by close()
571 */
572 fl.l_whence = SEEK_SET;
573 fl.l_start = 0;
574 fl.l_len = 0;
575 VOP_ADVLOCK(vp, fdp, F_UNLCK, &fl, F_POSIX);
576 fl.l_type = F_UNLCK;
577 }
578 goto out;
579
580
581 case F_GETLK:
582 error = pledge_flock(p);
583 if (error != 0)
584 break;
585
586 if (fp->f_type != DTYPE_VNODE) {
587 error = EINVAL;
588 break;
589 }
590 vp = fp->f_data;
591 /* Copy in the lock structure */
592 error = copyin((caddr_t)SCARG(uap, arg), (caddr_t)&fl,
593 sizeof (fl));
594 if (error)
595 break;
596 if (fl.l_whence == SEEK_CUR) {
597 off_t offset = foffset(fp);
598
599 if (fl.l_start == 0 && fl.l_len < 0) {
600 /* lockf(3) compliance hack */
601 fl.l_len = -fl.l_len;
602 fl.l_start = offset - fl.l_len;
603 } else
604 fl.l_start += offset;
605 }
606 if (fl.l_type != F_RDLCK &&
607 fl.l_type != F_WRLCK &&
608 fl.l_type != F_UNLCK &&
609 fl.l_type != 0) {
610 error = EINVAL;
611 break;
612 }
613 error = VOP_ADVLOCK(vp, fdp, F_GETLK, &fl, F_POSIX);
614 if (error)
615 break;
616 #ifdef KTRACE
617 if (KTRPOINT(p, KTR_STRUCT))
618 ktrflock(p, &fl);
619 #endif
620 error = (copyout((caddr_t)&fl, (caddr_t)SCARG(uap, arg),
621 sizeof (fl)));
622 break;
623
624 default:
625 error = EINVAL;
626 break;
627 }
628 out:
629 FRELE(fp, p);
630 return (error);
631 }
632
633 /*
634 * Common code for dup, dup2, and fcntl(F_DUPFD).
635 */
636 int
finishdup(struct proc * p,struct file * fp,int old,int new,register_t * retval,int dupflags)637 finishdup(struct proc *p, struct file *fp, int old, int new,
638 register_t *retval, int dupflags)
639 {
640 struct file *oldfp;
641 struct filedesc *fdp = p->p_fd;
642 int error;
643
644 fdpassertlocked(fdp);
645 KASSERT(fp->f_iflags & FIF_INSERTED);
646
647 if (fp->f_count >= FDUP_MAX_COUNT) {
648 error = EDEADLK;
649 goto fail;
650 }
651
652 oldfp = fd_getfile(fdp, new);
653 if ((dupflags & DUPF_DUP2) && oldfp == NULL) {
654 if (fd_inuse(fdp, new)) {
655 error = EBUSY;
656 goto fail;
657 }
658 fd_used(fdp, new);
659 }
660
661 /*
662 * Use `fd_fplock' to synchronize with fd_getfile() so that
663 * the function no longer creates a new reference to the old file.
664 */
665 mtx_enter(&fdp->fd_fplock);
666 fdp->fd_ofiles[new] = fp;
667 mtx_leave(&fdp->fd_fplock);
668
669 fdp->fd_ofileflags[new] = fdp->fd_ofileflags[old] & ~UF_EXCLOSE;
670 if (dupflags & DUPF_CLOEXEC)
671 fdp->fd_ofileflags[new] |= UF_EXCLOSE;
672 *retval = new;
673
674 if (oldfp != NULL) {
675 knote_fdclose(p, new);
676 fdpunlock(fdp);
677 closef(oldfp, p);
678 } else {
679 fdpunlock(fdp);
680 }
681
682 return (0);
683
684 fail:
685 fdpunlock(fdp);
686 FRELE(fp, p);
687 return (error);
688 }
689
690 void
fdinsert(struct filedesc * fdp,int fd,int flags,struct file * fp)691 fdinsert(struct filedesc *fdp, int fd, int flags, struct file *fp)
692 {
693 struct file *fq;
694
695 fdpassertlocked(fdp);
696
697 mtx_enter(&fhdlk);
698 if ((fp->f_iflags & FIF_INSERTED) == 0) {
699 atomic_setbits_int(&fp->f_iflags, FIF_INSERTED);
700 if ((fq = fdp->fd_ofiles[0]) != NULL) {
701 LIST_INSERT_AFTER(fq, fp, f_list);
702 } else {
703 LIST_INSERT_HEAD(&filehead, fp, f_list);
704 }
705 }
706 mtx_leave(&fhdlk);
707
708 mtx_enter(&fdp->fd_fplock);
709 KASSERT(fdp->fd_ofiles[fd] == NULL);
710 fdp->fd_ofiles[fd] = fp;
711 mtx_leave(&fdp->fd_fplock);
712
713 fdp->fd_ofileflags[fd] |= (flags & UF_EXCLOSE);
714 }
715
716 void
fdremove(struct filedesc * fdp,int fd)717 fdremove(struct filedesc *fdp, int fd)
718 {
719 fdpassertlocked(fdp);
720
721 /*
722 * Use `fd_fplock' to synchronize with fd_getfile() so that
723 * the function no longer creates a new reference to the file.
724 */
725 mtx_enter(&fdp->fd_fplock);
726 fdp->fd_ofiles[fd] = NULL;
727 mtx_leave(&fdp->fd_fplock);
728
729 fdp->fd_ofileflags[fd] = 0;
730
731 fd_unused(fdp, fd);
732 }
733
734 int
fdrelease(struct proc * p,int fd)735 fdrelease(struct proc *p, int fd)
736 {
737 struct filedesc *fdp = p->p_fd;
738 struct file *fp;
739
740 fdpassertlocked(fdp);
741
742 fp = fd_getfile(fdp, fd);
743 if (fp == NULL) {
744 fdpunlock(fdp);
745 return (EBADF);
746 }
747 fdremove(fdp, fd);
748 knote_fdclose(p, fd);
749 fdpunlock(fdp);
750 return (closef(fp, p));
751 }
752
753 /*
754 * Close a file descriptor.
755 */
756 int
sys_close(struct proc * p,void * v,register_t * retval)757 sys_close(struct proc *p, void *v, register_t *retval)
758 {
759 struct sys_close_args /* {
760 syscallarg(int) fd;
761 } */ *uap = v;
762 int fd = SCARG(uap, fd), error;
763 struct filedesc *fdp = p->p_fd;
764
765 fdplock(fdp);
766 /* fdrelease unlocks fdp. */
767 error = fdrelease(p, fd);
768
769 return (error);
770 }
771
772 /*
773 * Return status information about a file descriptor.
774 */
775 int
sys_fstat(struct proc * p,void * v,register_t * retval)776 sys_fstat(struct proc *p, void *v, register_t *retval)
777 {
778 struct sys_fstat_args /* {
779 syscallarg(int) fd;
780 syscallarg(struct stat *) sb;
781 } */ *uap = v;
782 int fd = SCARG(uap, fd);
783 struct filedesc *fdp = p->p_fd;
784 struct file *fp;
785 struct stat ub;
786 int error;
787
788 if ((fp = fd_getfile(fdp, fd)) == NULL)
789 return (EBADF);
790 error = (*fp->f_ops->fo_stat)(fp, &ub, p);
791 FRELE(fp, p);
792 if (error == 0) {
793 /*
794 * Don't let non-root see generation numbers
795 * (for NFS security)
796 */
797 if (suser(p))
798 ub.st_gen = 0;
799 error = copyout((caddr_t)&ub, (caddr_t)SCARG(uap, sb),
800 sizeof (ub));
801 }
802 #ifdef KTRACE
803 if (error == 0 && KTRPOINT(p, KTR_STRUCT))
804 ktrstat(p, &ub);
805 #endif
806 return (error);
807 }
808
809 /*
810 * Return pathconf information about a file descriptor.
811 */
812 int
sys_fpathconf(struct proc * p,void * v,register_t * retval)813 sys_fpathconf(struct proc *p, void *v, register_t *retval)
814 {
815 struct sys_fpathconf_args /* {
816 syscallarg(int) fd;
817 syscallarg(int) name;
818 } */ *uap = v;
819 int fd = SCARG(uap, fd);
820 struct filedesc *fdp = p->p_fd;
821 struct file *fp;
822 struct vnode *vp;
823 int error;
824
825 if ((fp = fd_getfile(fdp, fd)) == NULL)
826 return (EBADF);
827 switch (fp->f_type) {
828 case DTYPE_PIPE:
829 case DTYPE_SOCKET:
830 if (SCARG(uap, name) != _PC_PIPE_BUF) {
831 error = EINVAL;
832 break;
833 }
834 *retval = PIPE_BUF;
835 error = 0;
836 break;
837
838 case DTYPE_VNODE:
839 vp = fp->f_data;
840 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
841 error = VOP_PATHCONF(vp, SCARG(uap, name), retval);
842 VOP_UNLOCK(vp);
843 break;
844
845 default:
846 error = EOPNOTSUPP;
847 break;
848 }
849 FRELE(fp, p);
850 return (error);
851 }
852
853 /*
854 * Allocate a file descriptor for the process.
855 */
856 int
fdalloc(struct proc * p,int want,int * result)857 fdalloc(struct proc *p, int want, int *result)
858 {
859 struct filedesc *fdp = p->p_fd;
860 int lim, last, i;
861 u_int new, off;
862
863 fdpassertlocked(fdp);
864
865 /*
866 * Search for a free descriptor starting at the higher
867 * of want or fd_freefile. If that fails, consider
868 * expanding the ofile array.
869 */
870 restart:
871 lim = min((int)lim_cur(RLIMIT_NOFILE), atomic_load_int(&maxfiles));
872 last = min(fdp->fd_nfiles, lim);
873 if ((i = want) < fdp->fd_freefile)
874 i = fdp->fd_freefile;
875 off = i >> NDENTRYSHIFT;
876 new = find_next_zero(fdp->fd_himap, off,
877 (last + NDENTRIES - 1) >> NDENTRYSHIFT);
878 if (new != -1) {
879 i = find_next_zero(&fdp->fd_lomap[new],
880 new > off ? 0 : i & NDENTRYMASK,
881 NDENTRIES);
882 if (i == -1) {
883 /*
884 * Free file descriptor in this block was
885 * below want, try again with higher want.
886 */
887 want = (new + 1) << NDENTRYSHIFT;
888 goto restart;
889 }
890 i += (new << NDENTRYSHIFT);
891 if (i < last) {
892 fd_used(fdp, i);
893 if (want <= fdp->fd_freefile)
894 fdp->fd_freefile = i;
895 *result = i;
896 fdp->fd_ofileflags[i] = 0;
897 if (ISSET(p->p_p->ps_flags, PS_PLEDGE))
898 fdp->fd_ofileflags[i] |= UF_PLEDGED;
899 return (0);
900 }
901 }
902 if (fdp->fd_nfiles >= lim)
903 return (EMFILE);
904
905 return (ENOSPC);
906 }
907
908 void
fdexpand(struct proc * p)909 fdexpand(struct proc *p)
910 {
911 struct filedesc *fdp = p->p_fd;
912 int nfiles, oldnfiles;
913 size_t copylen;
914 struct file **newofile, **oldofile;
915 char *newofileflags;
916 u_int *newhimap, *newlomap;
917
918 fdpassertlocked(fdp);
919
920 oldnfiles = fdp->fd_nfiles;
921 oldofile = fdp->fd_ofiles;
922
923 /*
924 * No space in current array.
925 */
926 if (fdp->fd_nfiles < NDEXTENT)
927 nfiles = NDEXTENT;
928 else
929 nfiles = 2 * fdp->fd_nfiles;
930
931 newofile = mallocarray(nfiles, OFILESIZE, M_FILEDESC, M_WAITOK);
932 /*
933 * Allocate all required chunks before calling free(9) to make
934 * sure that ``fd_ofiles'' stays valid if we go to sleep.
935 */
936 if (NDHISLOTS(nfiles) > NDHISLOTS(fdp->fd_nfiles)) {
937 newhimap = mallocarray(NDHISLOTS(nfiles), sizeof(u_int),
938 M_FILEDESC, M_WAITOK);
939 newlomap = mallocarray(NDLOSLOTS(nfiles), sizeof(u_int),
940 M_FILEDESC, M_WAITOK);
941 }
942 newofileflags = (char *) &newofile[nfiles];
943
944 /*
945 * Copy the existing ofile and ofileflags arrays
946 * and zero the new portion of each array.
947 */
948 copylen = sizeof(struct file *) * fdp->fd_nfiles;
949 memcpy(newofile, fdp->fd_ofiles, copylen);
950 memset((char *)newofile + copylen, 0,
951 nfiles * sizeof(struct file *) - copylen);
952 copylen = sizeof(char) * fdp->fd_nfiles;
953 memcpy(newofileflags, fdp->fd_ofileflags, copylen);
954 memset(newofileflags + copylen, 0, nfiles * sizeof(char) - copylen);
955
956 if (NDHISLOTS(nfiles) > NDHISLOTS(fdp->fd_nfiles)) {
957 copylen = NDHISLOTS(fdp->fd_nfiles) * sizeof(u_int);
958 memcpy(newhimap, fdp->fd_himap, copylen);
959 memset((char *)newhimap + copylen, 0,
960 NDHISLOTS(nfiles) * sizeof(u_int) - copylen);
961
962 copylen = NDLOSLOTS(fdp->fd_nfiles) * sizeof(u_int);
963 memcpy(newlomap, fdp->fd_lomap, copylen);
964 memset((char *)newlomap + copylen, 0,
965 NDLOSLOTS(nfiles) * sizeof(u_int) - copylen);
966
967 if (NDHISLOTS(fdp->fd_nfiles) > NDHISLOTS(NDFILE)) {
968 free(fdp->fd_himap, M_FILEDESC,
969 NDHISLOTS(fdp->fd_nfiles) * sizeof(u_int));
970 free(fdp->fd_lomap, M_FILEDESC,
971 NDLOSLOTS(fdp->fd_nfiles) * sizeof(u_int));
972 }
973 fdp->fd_himap = newhimap;
974 fdp->fd_lomap = newlomap;
975 }
976
977 mtx_enter(&fdp->fd_fplock);
978 fdp->fd_ofiles = newofile;
979 mtx_leave(&fdp->fd_fplock);
980
981 fdp->fd_ofileflags = newofileflags;
982 fdp->fd_nfiles = nfiles;
983
984 if (oldnfiles > NDFILE)
985 free(oldofile, M_FILEDESC, oldnfiles * OFILESIZE);
986 }
987
988 /*
989 * Create a new open file structure and allocate
990 * a file descriptor for the process that refers to it.
991 */
992 int
falloc(struct proc * p,struct file ** resultfp,int * resultfd)993 falloc(struct proc *p, struct file **resultfp, int *resultfd)
994 {
995 struct file *fp;
996 int error, i;
997
998 KASSERT(resultfp != NULL);
999 KASSERT(resultfd != NULL);
1000
1001 fdpassertlocked(p->p_fd);
1002 restart:
1003 if ((error = fdalloc(p, 0, &i)) != 0) {
1004 if (error == ENOSPC) {
1005 fdexpand(p);
1006 goto restart;
1007 }
1008 return (error);
1009 }
1010
1011 fp = fnew(p);
1012 if (fp == NULL) {
1013 fd_unused(p->p_fd, i);
1014 return (ENFILE);
1015 }
1016
1017 FREF(fp);
1018 *resultfp = fp;
1019 *resultfd = i;
1020
1021 return (0);
1022 }
1023
1024 struct file *
fnew(struct proc * p)1025 fnew(struct proc *p)
1026 {
1027 struct file *fp;
1028 int nfiles;
1029
1030 nfiles = atomic_inc_int_nv(&numfiles);
1031 if (nfiles > atomic_load_int(&maxfiles)) {
1032 atomic_dec_int(&numfiles);
1033 tablefull("file");
1034 return (NULL);
1035 }
1036
1037 fp = pool_get(&file_pool, PR_WAITOK|PR_ZERO);
1038 /*
1039 * We need to block interrupts as long as `f_mtx' is being taken
1040 * with and without the KERNEL_LOCK().
1041 */
1042 mtx_init(&fp->f_mtx, IPL_MPFLOOR);
1043 fp->f_count = 1;
1044 fp->f_cred = p->p_ucred;
1045 crhold(fp->f_cred);
1046
1047 return (fp);
1048 }
1049
1050 /*
1051 * Build a new filedesc structure.
1052 */
1053 struct filedesc *
fdinit(void)1054 fdinit(void)
1055 {
1056 struct filedesc0 *newfdp;
1057
1058 newfdp = pool_get(&fdesc_pool, PR_WAITOK|PR_ZERO);
1059 rw_init(&newfdp->fd_fd.fd_lock, "fdlock");
1060 mtx_init(&newfdp->fd_fd.fd_fplock, IPL_MPFLOOR);
1061 LIST_INIT(&newfdp->fd_fd.fd_kqlist);
1062
1063 /* Create the file descriptor table. */
1064 newfdp->fd_fd.fd_refcnt = 1;
1065 newfdp->fd_fd.fd_cmask = S_IWGRP|S_IWOTH;
1066 newfdp->fd_fd.fd_ofiles = newfdp->fd_dfiles;
1067 newfdp->fd_fd.fd_ofileflags = newfdp->fd_dfileflags;
1068 newfdp->fd_fd.fd_nfiles = NDFILE;
1069 newfdp->fd_fd.fd_himap = newfdp->fd_dhimap;
1070 newfdp->fd_fd.fd_lomap = newfdp->fd_dlomap;
1071
1072 newfdp->fd_fd.fd_freefile = 0;
1073 newfdp->fd_fd.fd_lastfile = 0;
1074
1075 return (&newfdp->fd_fd);
1076 }
1077
1078 /*
1079 * Share a filedesc structure.
1080 */
1081 struct filedesc *
fdshare(struct process * pr)1082 fdshare(struct process *pr)
1083 {
1084 pr->ps_fd->fd_refcnt++;
1085 return (pr->ps_fd);
1086 }
1087
1088 /*
1089 * Copy a filedesc structure.
1090 */
1091 struct filedesc *
fdcopy(struct process * pr)1092 fdcopy(struct process *pr)
1093 {
1094 struct filedesc *newfdp, *fdp = pr->ps_fd;
1095 int i;
1096
1097 newfdp = fdinit();
1098
1099 fdplock(fdp);
1100 if (fdp->fd_cdir) {
1101 vref(fdp->fd_cdir);
1102 newfdp->fd_cdir = fdp->fd_cdir;
1103 }
1104 if (fdp->fd_rdir) {
1105 vref(fdp->fd_rdir);
1106 newfdp->fd_rdir = fdp->fd_rdir;
1107 }
1108
1109 /*
1110 * If the number of open files fits in the internal arrays
1111 * of the open file structure, use them, otherwise allocate
1112 * additional memory for the number of descriptors currently
1113 * in use.
1114 */
1115 if (fdp->fd_lastfile >= NDFILE) {
1116 /*
1117 * Compute the smallest multiple of NDEXTENT needed
1118 * for the file descriptors currently in use,
1119 * allowing the table to shrink.
1120 */
1121 i = fdp->fd_nfiles;
1122 while (i >= 2 * NDEXTENT && i > fdp->fd_lastfile * 2)
1123 i /= 2;
1124 newfdp->fd_ofiles = mallocarray(i, OFILESIZE, M_FILEDESC,
1125 M_WAITOK | M_ZERO);
1126 newfdp->fd_ofileflags = (char *) &newfdp->fd_ofiles[i];
1127 newfdp->fd_nfiles = i;
1128 }
1129 if (NDHISLOTS(newfdp->fd_nfiles) > NDHISLOTS(NDFILE)) {
1130 newfdp->fd_himap = mallocarray(NDHISLOTS(newfdp->fd_nfiles),
1131 sizeof(u_int), M_FILEDESC, M_WAITOK | M_ZERO);
1132 newfdp->fd_lomap = mallocarray(NDLOSLOTS(newfdp->fd_nfiles),
1133 sizeof(u_int), M_FILEDESC, M_WAITOK | M_ZERO);
1134 }
1135 newfdp->fd_freefile = fdp->fd_freefile;
1136 newfdp->fd_flags = fdp->fd_flags;
1137 newfdp->fd_cmask = fdp->fd_cmask;
1138
1139 for (i = 0; i <= fdp->fd_lastfile; i++) {
1140 struct file *fp = fdp->fd_ofiles[i];
1141
1142 if (fp != NULL) {
1143 /*
1144 * XXX Gruesome hack. If count gets too high, fail
1145 * to copy an fd, since fdcopy()'s callers do not
1146 * permit it to indicate failure yet.
1147 * Meanwhile, kqueue files have to be
1148 * tied to the process that opened them to enforce
1149 * their internal consistency, so close them here.
1150 */
1151 if (fp->f_count >= FDUP_MAX_COUNT ||
1152 fp->f_type == DTYPE_KQUEUE) {
1153 if (i < newfdp->fd_freefile)
1154 newfdp->fd_freefile = i;
1155 continue;
1156 }
1157
1158 FREF(fp);
1159 newfdp->fd_ofiles[i] = fp;
1160 newfdp->fd_ofileflags[i] = fdp->fd_ofileflags[i];
1161 fd_used(newfdp, i);
1162 }
1163 }
1164 fdpunlock(fdp);
1165
1166 return (newfdp);
1167 }
1168
1169 /*
1170 * Release a filedesc structure.
1171 */
1172 void
fdfree(struct proc * p)1173 fdfree(struct proc *p)
1174 {
1175 struct filedesc *fdp = p->p_fd;
1176 struct file *fp;
1177 int fd;
1178
1179 if (--fdp->fd_refcnt > 0)
1180 return;
1181 for (fd = 0; fd <= fdp->fd_lastfile; fd++) {
1182 fp = fdp->fd_ofiles[fd];
1183 if (fp != NULL) {
1184 fdp->fd_ofiles[fd] = NULL;
1185 knote_fdclose(p, fd);
1186 /* closef() expects a refcount of 2 */
1187 FREF(fp);
1188 (void) closef(fp, p);
1189 }
1190 }
1191 p->p_fd = NULL;
1192 if (fdp->fd_nfiles > NDFILE)
1193 free(fdp->fd_ofiles, M_FILEDESC, fdp->fd_nfiles * OFILESIZE);
1194 if (NDHISLOTS(fdp->fd_nfiles) > NDHISLOTS(NDFILE)) {
1195 free(fdp->fd_himap, M_FILEDESC,
1196 NDHISLOTS(fdp->fd_nfiles) * sizeof(u_int));
1197 free(fdp->fd_lomap, M_FILEDESC,
1198 NDLOSLOTS(fdp->fd_nfiles) * sizeof(u_int));
1199 }
1200 if (fdp->fd_cdir)
1201 vrele(fdp->fd_cdir);
1202 if (fdp->fd_rdir)
1203 vrele(fdp->fd_rdir);
1204 pool_put(&fdesc_pool, fdp);
1205 }
1206
1207 /*
1208 * Internal form of close.
1209 * Decrement reference count on file structure.
1210 * Note: p may be NULL when closing a file
1211 * that was being passed in a message.
1212 *
1213 * The fp must have its usecount bumped and will be FRELEd here.
1214 */
1215 int
closef(struct file * fp,struct proc * p)1216 closef(struct file *fp, struct proc *p)
1217 {
1218 struct filedesc *fdp;
1219
1220 if (fp == NULL)
1221 return (0);
1222
1223 KASSERTMSG(fp->f_count >= 2, "count (%u) < 2", fp->f_count);
1224
1225 atomic_dec_int(&fp->f_count);
1226
1227 /*
1228 * POSIX record locking dictates that any close releases ALL
1229 * locks owned by this process. This is handled by setting
1230 * a flag in the unlock to free ONLY locks obeying POSIX
1231 * semantics, and not to free BSD-style file locks.
1232 * If the descriptor was in a message, POSIX-style locks
1233 * aren't passed with the descriptor.
1234 */
1235
1236 if (p && ((fdp = p->p_fd) != NULL) &&
1237 (fdp->fd_flags & FD_ADVLOCK) &&
1238 fp->f_type == DTYPE_VNODE) {
1239 struct vnode *vp = fp->f_data;
1240 struct flock lf;
1241
1242 lf.l_whence = SEEK_SET;
1243 lf.l_start = 0;
1244 lf.l_len = 0;
1245 lf.l_type = F_UNLCK;
1246 (void) VOP_ADVLOCK(vp, fdp, F_UNLCK, &lf, F_POSIX);
1247 }
1248
1249 return (FRELE(fp, p));
1250 }
1251
1252 int
fdrop(struct file * fp,struct proc * p)1253 fdrop(struct file *fp, struct proc *p)
1254 {
1255 int error;
1256
1257 KASSERTMSG(fp->f_count == 0, "count (%u) != 0", fp->f_count);
1258
1259 mtx_enter(&fhdlk);
1260 if (fp->f_iflags & FIF_INSERTED)
1261 LIST_REMOVE(fp, f_list);
1262 mtx_leave(&fhdlk);
1263
1264 if (fp->f_ops)
1265 error = (*fp->f_ops->fo_close)(fp, p);
1266 else
1267 error = 0;
1268
1269 crfree(fp->f_cred);
1270 atomic_dec_int(&numfiles);
1271 pool_put(&file_pool, fp);
1272
1273 return (error);
1274 }
1275
1276 /*
1277 * Apply an advisory lock on a file descriptor.
1278 *
1279 * Just attempt to get a record lock of the requested type on
1280 * the entire file (l_whence = SEEK_SET, l_start = 0, l_len = 0).
1281 */
1282 int
sys_flock(struct proc * p,void * v,register_t * retval)1283 sys_flock(struct proc *p, void *v, register_t *retval)
1284 {
1285 struct sys_flock_args /* {
1286 syscallarg(int) fd;
1287 syscallarg(int) how;
1288 } */ *uap = v;
1289 int fd = SCARG(uap, fd);
1290 int how = SCARG(uap, how);
1291 struct filedesc *fdp = p->p_fd;
1292 struct file *fp;
1293 struct vnode *vp;
1294 struct flock lf;
1295 int error;
1296
1297 if ((fp = fd_getfile(fdp, fd)) == NULL)
1298 return (EBADF);
1299 if (fp->f_type != DTYPE_VNODE) {
1300 error = EOPNOTSUPP;
1301 goto out;
1302 }
1303 vp = fp->f_data;
1304 lf.l_whence = SEEK_SET;
1305 lf.l_start = 0;
1306 lf.l_len = 0;
1307 if (how & LOCK_UN) {
1308 lf.l_type = F_UNLCK;
1309 atomic_clearbits_int(&fp->f_iflags, FIF_HASLOCK);
1310 error = VOP_ADVLOCK(vp, (caddr_t)fp, F_UNLCK, &lf, F_FLOCK);
1311 goto out;
1312 }
1313 if (how & LOCK_EX)
1314 lf.l_type = F_WRLCK;
1315 else if (how & LOCK_SH)
1316 lf.l_type = F_RDLCK;
1317 else {
1318 error = EINVAL;
1319 goto out;
1320 }
1321 atomic_setbits_int(&fp->f_iflags, FIF_HASLOCK);
1322 if (how & LOCK_NB)
1323 error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, F_FLOCK);
1324 else
1325 error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf, F_FLOCK|F_WAIT);
1326 out:
1327 FRELE(fp, p);
1328 return (error);
1329 }
1330
1331 /*
1332 * File Descriptor pseudo-device driver (/dev/fd/).
1333 *
1334 * Opening minor device N dup()s the file (if any) connected to file
1335 * descriptor N belonging to the calling process. Note that this driver
1336 * consists of only the ``open()'' routine, because all subsequent
1337 * references to this file will be direct to the other driver.
1338 */
1339 int
filedescopen(dev_t dev,int mode,int type,struct proc * p)1340 filedescopen(dev_t dev, int mode, int type, struct proc *p)
1341 {
1342
1343 /*
1344 * XXX Kludge: set curproc->p_dupfd to contain the value of the
1345 * the file descriptor being sought for duplication. The error
1346 * return ensures that the vnode for this device will be released
1347 * by vn_open. Open will detect this special error and take the
1348 * actions in dupfdopen below. Other callers of vn_open or VOP_OPEN
1349 * will simply report the error.
1350 */
1351 p->p_dupfd = minor(dev);
1352 return (ENODEV);
1353 }
1354
1355 /*
1356 * Duplicate the specified descriptor to a free descriptor.
1357 */
1358 int
dupfdopen(struct proc * p,int indx,int mode)1359 dupfdopen(struct proc *p, int indx, int mode)
1360 {
1361 struct filedesc *fdp = p->p_fd;
1362 int dupfd = p->p_dupfd;
1363 struct file *wfp;
1364
1365 fdpassertlocked(fdp);
1366
1367 /*
1368 * Assume that the filename was user-specified; applications do
1369 * not tend to open /dev/fd/# when they can just call dup()
1370 */
1371 if ((p->p_p->ps_flags & (PS_SUGIDEXEC | PS_SUGID))) {
1372 if (p->p_descfd == 255)
1373 return (EPERM);
1374 if (p->p_descfd != dupfd)
1375 return (EPERM);
1376 }
1377
1378 /*
1379 * If the to-be-dup'd fd number is greater than the allowed number
1380 * of file descriptors, or the fd to be dup'd has already been
1381 * closed, reject. Note, there is no need to check for new == old
1382 * because fd_getfile will return NULL if the file at indx is
1383 * newly created by falloc.
1384 */
1385 if ((wfp = fd_getfile(fdp, dupfd)) == NULL)
1386 return (EBADF);
1387
1388 /*
1389 * Check that the mode the file is being opened for is a
1390 * subset of the mode of the existing descriptor.
1391 */
1392 if (((mode & (FREAD|FWRITE)) | wfp->f_flag) != wfp->f_flag) {
1393 FRELE(wfp, p);
1394 return (EACCES);
1395 }
1396 if (wfp->f_count >= FDUP_MAX_COUNT) {
1397 FRELE(wfp, p);
1398 return (EDEADLK);
1399 }
1400
1401 KASSERT(wfp->f_iflags & FIF_INSERTED);
1402
1403 mtx_enter(&fdp->fd_fplock);
1404 KASSERT(fdp->fd_ofiles[indx] == NULL);
1405 fdp->fd_ofiles[indx] = wfp;
1406 mtx_leave(&fdp->fd_fplock);
1407
1408 fdp->fd_ofileflags[indx] = (fdp->fd_ofileflags[indx] & UF_EXCLOSE) |
1409 (fdp->fd_ofileflags[dupfd] & ~UF_EXCLOSE);
1410
1411 return (0);
1412 }
1413
1414 /*
1415 * Close any files on exec?
1416 */
1417 void
fdcloseexec(struct proc * p)1418 fdcloseexec(struct proc *p)
1419 {
1420 struct filedesc *fdp = p->p_fd;
1421 int fd;
1422
1423 fdplock(fdp);
1424 for (fd = 0; fd <= fdp->fd_lastfile; fd++) {
1425 fdp->fd_ofileflags[fd] &= ~UF_PLEDGED;
1426 if (fdp->fd_ofileflags[fd] & UF_EXCLOSE) {
1427 /* fdrelease() unlocks fdp. */
1428 (void) fdrelease(p, fd);
1429 fdplock(fdp);
1430 }
1431 }
1432 fdpunlock(fdp);
1433 }
1434
1435 int
sys_closefrom(struct proc * p,void * v,register_t * retval)1436 sys_closefrom(struct proc *p, void *v, register_t *retval)
1437 {
1438 struct sys_closefrom_args *uap = v;
1439 struct filedesc *fdp = p->p_fd;
1440 u_int startfd, i;
1441
1442 startfd = SCARG(uap, fd);
1443 fdplock(fdp);
1444
1445 if (startfd > fdp->fd_lastfile) {
1446 fdpunlock(fdp);
1447 return (EBADF);
1448 }
1449
1450 for (i = startfd; i <= fdp->fd_lastfile; i++) {
1451 /* fdrelease() unlocks fdp. */
1452 fdrelease(p, i);
1453 fdplock(fdp);
1454 }
1455
1456 fdpunlock(fdp);
1457 return (0);
1458 }
1459
1460 int
sys_getdtablecount(struct proc * p,void * v,register_t * retval)1461 sys_getdtablecount(struct proc *p, void *v, register_t *retval)
1462 {
1463 *retval = p->p_fd->fd_openfd;
1464 return (0);
1465 }
1466