]> git.saurik.com Git - apple/xnu.git/blob - bsd/kern/kern_descrip.c
projects / apple / xnu.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
xnu-792.6.56.tar.gz
[apple/xnu.git] / bsd / kern / kern_descrip.c
1 /*
2 * Copyright (c) 2000-2004 Apple Computer, Inc. All rights reserved.
3 *
4 * @APPLE_LICENSE_HEADER_START@
5 *
6 * This file contains Original Code and/or Modifications of Original Code
7 * as defined in and that are subject to the Apple Public Source License
8 * Version 2.0 (the 'License'). You may not use this file except in
9 * compliance with the License. Please obtain a copy of the License at
10 * http://www.opensource.apple.com/apsl/ and read it before using this
11 * file.
12 *
13 * The Original Code and all software distributed under the License are
14 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
15 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
16 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
17 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
18 * Please see the License for the specific language governing rights and
19 * limitations under the License.
20 *
21 * @APPLE_LICENSE_HEADER_END@
22 */
23 /* Copyright (c) 1995, 1997 Apple Computer, Inc. All Rights Reserved */
24 /*
25 * Copyright (c) 1982, 1986, 1989, 1991, 1993
26 * The Regents of the University of California. All rights reserved.
27 * (c) UNIX System Laboratories, Inc.
28 * All or some portions of this file are derived from material licensed
29 * to the University of California by American Telephone and Telegraph
30 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
31 * the permission of UNIX System Laboratories, Inc.
32 *
33 * Redistribution and use in source and binary forms, with or without
34 * modification, are permitted provided that the following conditions
35 * are met:
36 * 1. Redistributions of source code must retain the above copyright
37 * notice, this list of conditions and the following disclaimer.
38 * 2. Redistributions in binary form must reproduce the above copyright
39 * notice, this list of conditions and the following disclaimer in the
40 * documentation and/or other materials provided with the distribution.
41 * 3. All advertising materials mentioning features or use of this software
42 * must display the following acknowledgement:
43 * This product includes software developed by the University of
44 * California, Berkeley and its contributors.
45 * 4. Neither the name of the University nor the names of its contributors
46 * may be used to endorse or promote products derived from this software
47 * without specific prior written permission.
48 *
49 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59 * SUCH DAMAGE.
60 *
61 * @(#)kern_descrip.c 8.8 (Berkeley) 2/14/95
62 */
63
64 #include <sys/param.h>
65 #include <sys/systm.h>
66 #include <sys/filedesc.h>
67 #include <sys/kernel.h>
68 #include <sys/vnode_internal.h>
69 #include <sys/proc_internal.h>
70 #include <sys/kauth.h>
71 #include <sys/file_internal.h>
72 #include <sys/socket.h>
73 #include <sys/socketvar.h>
74 #include <sys/stat.h>
75 #include <sys/ioctl.h>
76 #include <sys/fcntl.h>
77 #include <sys/malloc.h>
78 #include <sys/mman.h>
79 #include <sys/syslog.h>
80 #include <sys/unistd.h>
81 #include <sys/resourcevar.h>
82 #include <sys/aio_kern.h>
83 #include <sys/ev.h>
84 #include <kern/lock.h>
85
86 #include <bsm/audit_kernel.h>
87
88 #include <sys/mount_internal.h>
89 #include <sys/kdebug.h>
90 #include <sys/sysproto.h>
91 #include <sys/pipe.h>
92 #include <kern/kern_types.h>
93 #include <kern/kalloc.h>
94 #include <libkern/OSAtomic.h>
95
96 struct psemnode;
97 struct pshmnode;
98
99 int fdopen(dev_t dev, int mode, int type, struct proc *p);
100 int ogetdtablesize(struct proc *p, void *uap, register_t *retval);
101 int finishdup(struct proc * p, struct filedesc *fdp, int old, int new, register_t *retval);
102
103 int closef(struct fileglob *fg, struct proc *p);
104 int falloc_locked(struct proc *p, struct fileproc **resultfp, int *resultfd, int locked);
105 void fddrop(struct proc *p, int fd);
106 int fdgetf_noref(struct proc *p, int fd, struct fileproc **resultfp);
107 void fg_drop(struct fileproc * fp);
108 void fg_free(struct fileglob *fg);
109 void fg_ref(struct fileproc * fp);
110 int fp_getfpshm(struct proc *p, int fd, struct fileproc **resultfp, struct pshmnode **resultpshm);
111
112 static int closef_finish(struct fileproc *fp, struct fileglob *fg, struct proc *p);
113
114 extern void file_lock_init(void);
115 extern int is_suser(void);
116 extern int kqueue_stat(struct fileproc *fp, struct stat *st, struct proc *p);
117 extern int soo_stat(struct socket *so, struct stat *ub);
118 extern int vn_path_package_check(vnode_t vp, char *path, int pathlen, int *component);
119
120 extern kauth_scope_t kauth_scope_fileop;
121
122 #define f_flag f_fglob->fg_flag
123 #define f_type f_fglob->fg_type
124 #define f_msgcount f_fglob->fg_msgcount
125 #define f_cred f_fglob->fg_cred
126 #define f_ops f_fglob->fg_ops
127 #define f_offset f_fglob->fg_offset
128 #define f_data f_fglob->fg_data
129 /*
130 * Descriptor management.
131 */
132 struct filelist filehead; /* head of list of open files */
133 struct fmsglist fmsghead; /* head of list of open files */
134 struct fmsglist fmsg_ithead; /* head of list of open files */
135 int nfiles; /* actual number of open files */
136
137
138 lck_grp_attr_t * file_lck_grp_attr;
139 lck_grp_t * file_lck_grp;
140 lck_attr_t * file_lck_attr;
141
142 lck_mtx_t * uipc_lock;
143 lck_mtx_t * file_iterate_lcok;
144 lck_mtx_t * file_flist_lock;
145
146
147 void
148 file_lock_init(void)
149 {
150
151 /* allocate file lock group attribute and group */
152 file_lck_grp_attr= lck_grp_attr_alloc_init();
153 lck_grp_attr_setstat(file_lck_grp_attr);
154
155 file_lck_grp = lck_grp_alloc_init("file", file_lck_grp_attr);
156
157 /* Allocate file lock attribute */
158 file_lck_attr = lck_attr_alloc_init();
159 //lck_attr_setdebug(file_lck_attr);
160
161 uipc_lock = lck_mtx_alloc_init(file_lck_grp, file_lck_attr);
162 file_iterate_lcok = lck_mtx_alloc_init(file_lck_grp, file_lck_attr);
163 file_flist_lock = lck_mtx_alloc_init(file_lck_grp, file_lck_attr);
164
165
166
167 }
168
169
170 void
171 proc_fdlock(struct proc *p)
172 {
173 lck_mtx_lock(&p->p_fdmlock);
174 }
175
176 void
177 proc_fdunlock(struct proc *p)
178 {
179 lck_mtx_unlock(&p->p_fdmlock);
180 }
181
182 /*
183 * System calls on descriptors.
184 */
185
186 int
187 getdtablesize(struct proc *p, __unused struct getdtablesize_args *uap, register_t *retval)
188 {
189 proc_fdlock(p);
190 *retval = min((int)p->p_rlimit[RLIMIT_NOFILE].rlim_cur, maxfiles);
191 proc_fdunlock(p);
192
193 return (0);
194 }
195
196 int
197 ogetdtablesize(struct proc *p, __unused void *uap, register_t *retval)
198 {
199 proc_fdlock(p);
200 *retval = min((int)p->p_rlimit[RLIMIT_NOFILE].rlim_cur, NOFILE);
201 proc_fdunlock(p);
202
203 return (0);
204 }
205
206 static __inline__ void
207 _fdrelse(struct filedesc *fdp, int fd)
208 {
209 if (fd < fdp->fd_freefile)
210 fdp->fd_freefile = fd;
211 #if DIAGNOSTIC
212 if (fd > fdp->fd_lastfile)
213 panic("fdrelse: fd_lastfile inconsistent");
214 #endif
215 fdp->fd_ofiles[fd] = NULL;
216 fdp->fd_ofileflags[fd] = 0;
217
218 while ((fd = fdp->fd_lastfile) > 0 &&
219 fdp->fd_ofiles[fd] == NULL &&
220 !(fdp->fd_ofileflags[fd] & UF_RESERVED))
221 fdp->fd_lastfile--;
222 }
223
224 /*
225 * Duplicate a file descriptor.
226 */
227 /* ARGSUSED */
228 int
229 dup(p, uap, retval)
230 struct proc *p;
231 struct dup_args *uap;
232 register_t *retval;
233 {
234 register struct filedesc *fdp = p->p_fd;
235 register int old = uap->fd;
236 int new, error;
237 struct fileproc *fp;
238
239 proc_fdlock(p);
240 if ( (error = fp_lookup(p, old, &fp, 1)) ) {
241 proc_fdunlock(p);
242 return(error);
243 }
244 if ( (error = fdalloc(p, 0, &new)) ) {
245 fp_drop(p, old, fp, 1);
246 proc_fdunlock(p);
247 return (error);
248 }
249 error = finishdup(p, fdp, old, new, retval);
250 fp_drop(p, old, fp, 1);
251 proc_fdunlock(p);
252
253 return (error);
254 }
255
256 /*
257 * Duplicate a file descriptor to a particular value.
258 */
259 /* ARGSUSED */
260 int
261 dup2(p, uap, retval)
262 struct proc *p;
263 struct dup2_args *uap;
264 register_t *retval;
265 {
266 register struct filedesc *fdp = p->p_fd;
267 register int old = uap->from, new = uap->to;
268 int i, error;
269 struct fileproc *fp;
270
271 proc_fdlock(p);
272
273 if ( (error = fp_lookup(p, old, &fp, 1)) ) {
274 proc_fdunlock(p);
275 return(error);
276 }
277 if (new < 0 ||
278 new >= p->p_rlimit[RLIMIT_NOFILE].rlim_cur ||
279 new >= maxfiles) {
280 fp_drop(p, old, fp, 1);
281 proc_fdunlock(p);
282 return (EBADF);
283 }
284 if (old == new) {
285 fp_drop(p, old, fp, 1);
286 *retval = new;
287 proc_fdunlock(p);
288 return (0);
289 }
290 if (new < 0 || new >= fdp->fd_nfiles) {
291 if ( (error = fdalloc(p, new, &i)) ) {
292 fp_drop(p, old, fp, 1);
293 proc_fdunlock(p);
294 return (error);
295 }
296 if (new != i) {
297 _fdrelse(fdp, i);
298 goto closeit;
299 }
300 } else {
301 struct fileproc **fpp;
302 char flags;
303 closeit:
304 flags = fdp->fd_ofileflags[new];
305 if ((flags & (UF_RESERVED | UF_CLOSING)) == UF_RESERVED) {
306 fp_drop(p, old, fp, 1);
307 proc_fdunlock(p);
308 return (EBADF);
309 }
310
311 /*
312 * dup2() must succeed even if the close has an error.
313 */
314 if (*(fpp = &fdp->fd_ofiles[new])) {
315 struct fileproc *nfp = *fpp;
316
317 close_internal(p, new, nfp, (CLOSEINT_LOCKED | CLOSEINT_WAITONCLOSE | CLOSEINT_NOFDRELSE | CLOSEINT_NOFDNOREF));
318 *fpp = NULL;
319 }
320 }
321 error = finishdup(p, fdp, old, new, retval);
322 fp_drop(p, old, fp, 1);
323 proc_fdunlock(p);
324
325 return(error);
326 }
327
328 /*
329 * The file control system call.
330 */
331 int
332 fcntl(p, uap, retval)
333 struct proc *p;
334 struct fcntl_args *uap;
335 register_t *retval;
336 {
337 int fd = uap->fd;
338 struct filedesc *fdp = p->p_fd;
339 struct fileproc *fp;
340 char *pop;
341 struct vnode *vp;
342 int i, tmp, error, error2, flg = F_POSIX;
343 struct flock fl;
344 struct vfs_context context;
345 off_t offset;
346 int newmin;
347 daddr64_t lbn, bn;
348 int devBlockSize = 0;
349 unsigned int fflag;
350 user_addr_t argp;
351
352 AUDIT_ARG(fd, uap->fd);
353 AUDIT_ARG(cmd, uap->cmd);
354
355 proc_fdlock(p);
356 if ( (error = fp_lookup(p, fd, &fp, 1)) ) {
357 proc_fdunlock(p);
358 return(error);
359 }
360 context.vc_proc = p;
361 context.vc_ucred = fp->f_cred;
362 if (proc_is64bit(p)) {
363 argp = uap->arg;
364 }
365 else {
366 /* since the arg parameter is defined as a long but may be either
367 * a long or a pointer we must take care to handle sign extension
368 * issues. Our sys call munger will sign extend a long when we are
369 * called from a 32-bit process. Since we can never have an address
370 * greater than 32-bits from a 32-bit process we lop off the top
371 * 32-bits to avoid getting the wrong address
372 */
373 argp = CAST_USER_ADDR_T(uap->arg);
374 }
375
376 pop = &fdp->fd_ofileflags[fd];
377
378 switch (uap->cmd) {
379
380 case F_DUPFD:
381 newmin = CAST_DOWN(int, uap->arg);
382 if ((u_int)newmin >= p->p_rlimit[RLIMIT_NOFILE].rlim_cur ||
383 newmin >= maxfiles) {
384 error = EINVAL;
385 goto out;
386 }
387 if ( (error = fdalloc(p, newmin, &i)) )
388 goto out;
389 error = finishdup(p, fdp, fd, i, retval);
390 goto out;
391
392 case F_GETFD:
393 *retval = (*pop & UF_EXCLOSE)? 1 : 0;
394 error = 0;
395 goto out;
396
397 case F_SETFD:
398 *pop = (*pop &~ UF_EXCLOSE) |
399 (uap->arg & 1)? UF_EXCLOSE : 0;
400 error = 0;
401 goto out;
402
403 case F_GETFL:
404 *retval = OFLAGS(fp->f_flag);
405 error = 0;
406 goto out;
407
408 case F_SETFL:
409 fp->f_flag &= ~FCNTLFLAGS;
410 tmp = CAST_DOWN(int, uap->arg);
411 fp->f_flag |= FFLAGS(tmp) & FCNTLFLAGS;
412 tmp = fp->f_flag & FNONBLOCK;
413 error = fo_ioctl(fp, FIONBIO, (caddr_t)&tmp, p);
414 if (error)
415 goto out;
416 tmp = fp->f_flag & FASYNC;
417 error = fo_ioctl(fp, FIOASYNC, (caddr_t)&tmp, p);
418 if (!error)
419 goto out;
420 fp->f_flag &= ~FNONBLOCK;
421 tmp = 0;
422 (void)fo_ioctl(fp, FIONBIO, (caddr_t)&tmp, p);
423 goto out;
424
425 case F_GETOWN:
426 if (fp->f_type == DTYPE_SOCKET) {
427 *retval = ((struct socket *)fp->f_data)->so_pgid;
428 error = 0;
429 goto out;
430 }
431 error = fo_ioctl(fp, (int)TIOCGPGRP, (caddr_t)retval, p);
432 *retval = -*retval;
433 goto out;
434
435 case F_SETOWN:
436 tmp = CAST_DOWN(pid_t, uap->arg);
437 if (fp->f_type == DTYPE_SOCKET) {
438 ((struct socket *)fp->f_data)->so_pgid = tmp;
439 error =0;
440 goto out;
441 }
442 if (fp->f_type == DTYPE_PIPE) {
443 error = fo_ioctl(fp, (int)TIOCSPGRP, (caddr_t)&tmp, p);
444 goto out;
445 }
446
447 if (tmp <= 0) {
448 tmp = -tmp;
449 } else {
450 struct proc *p1 = pfind(tmp);
451 if (p1 == 0) {
452 error = ESRCH;
453 goto out;
454 }
455 tmp = (int)p1->p_pgrp->pg_id;
456 }
457 error = fo_ioctl(fp, (int)TIOCSPGRP, (caddr_t)&tmp, p);
458 goto out;
459
460 case F_SETLKW:
461 flg |= F_WAIT;
462 /* Fall into F_SETLK */
463
464 case F_SETLK:
465 if (fp->f_type != DTYPE_VNODE) {
466 error = EBADF;
467 goto out;
468 }
469 vp = (struct vnode *)fp->f_data;
470
471 fflag = fp->f_flag;
472 offset = fp->f_offset;
473 proc_fdunlock(p);
474
475 /* Copy in the lock structure */
476 error = copyin(argp, (caddr_t)&fl, sizeof (fl));
477 if (error) {
478 goto outdrop;
479 }
480 if ( (error = vnode_getwithref(vp)) ) {
481 goto outdrop;
482 }
483 if (fl.l_whence == SEEK_CUR)
484 fl.l_start += offset;
485
486 switch (fl.l_type) {
487
488 case F_RDLCK:
489 if ((fflag & FREAD) == 0) {
490 (void)vnode_put(vp);
491 error = EBADF;
492 goto outdrop;
493 }
494 OSBitOrAtomic(P_LADVLOCK, &p->p_ladvflag);
495 error = VNOP_ADVLOCK(vp, (caddr_t)p, F_SETLK, &fl, flg, &context);
496 (void)vnode_put(vp);
497 goto outdrop;
498
499 case F_WRLCK:
500 if ((fflag & FWRITE) == 0) {
501 (void)vnode_put(vp);
502 error = EBADF;
503 goto outdrop;
504 }
505 OSBitOrAtomic(P_LADVLOCK, &p->p_ladvflag);
506 error = VNOP_ADVLOCK(vp, (caddr_t)p, F_SETLK, &fl, flg, &context);
507 (void)vnode_put(vp);
508 goto outdrop;
509
510 case F_UNLCK:
511 error = VNOP_ADVLOCK(vp, (caddr_t)p, F_UNLCK, &fl,
512 F_POSIX, &context);
513 (void)vnode_put(vp);
514 goto outdrop;
515
516 default:
517 (void)vnode_put(vp);
518 error = EINVAL;
519 goto outdrop;
520 }
521
522 case F_GETLK:
523 if (fp->f_type != DTYPE_VNODE) {
524 error = EBADF;
525 goto out;
526 }
527 vp = (struct vnode *)fp->f_data;
528
529 offset = fp->f_offset;
530 proc_fdunlock(p);
531
532 /* Copy in the lock structure */
533 error = copyin(argp, (caddr_t)&fl, sizeof (fl));
534 if (error)
535 goto outdrop;
536
537 if ( (error = vnode_getwithref(vp)) == 0 ) {
538 if (fl.l_whence == SEEK_CUR)
539 fl.l_start += offset;
540
541 error = VNOP_ADVLOCK(vp, (caddr_t)p, F_GETLK, &fl, F_POSIX, &context);
542
543 (void)vnode_put(vp);
544
545 if (error == 0)
546 error = copyout((caddr_t)&fl, argp, sizeof (fl));
547 }
548 goto outdrop;
549
550 case F_PREALLOCATE: {
551 fstore_t alloc_struct; /* structure for allocate command */
552 u_int32_t alloc_flags = 0;
553
554 if (fp->f_type != DTYPE_VNODE) {
555 error = EBADF;
556 goto out;
557 }
558
559 vp = (struct vnode *)fp->f_data;
560 proc_fdunlock(p);
561
562 /* make sure that we have write permission */
563 if ((fp->f_flag & FWRITE) == 0) {
564 error = EBADF;
565 goto outdrop;
566 }
567
568 error = copyin(argp, (caddr_t)&alloc_struct, sizeof (alloc_struct));
569 if (error)
570 goto outdrop;
571
572 /* now set the space allocated to 0 */
573 alloc_struct.fst_bytesalloc = 0;
574
575 /*
576 * Do some simple parameter checking
577 */
578
579 /* set up the flags */
580
581 alloc_flags |= PREALLOCATE;
582
583 if (alloc_struct.fst_flags & F_ALLOCATECONTIG)
584 alloc_flags |= ALLOCATECONTIG;
585
586 if (alloc_struct.fst_flags & F_ALLOCATEALL)
587 alloc_flags |= ALLOCATEALL;
588
589 /*
590 * Do any position mode specific stuff. The only
591 * position mode supported now is PEOFPOSMODE
592 */
593
594 switch (alloc_struct.fst_posmode) {
595
596 case F_PEOFPOSMODE:
597 if (alloc_struct.fst_offset != 0) {
598 error = EINVAL;
599 goto outdrop;
600 }
601
602 alloc_flags |= ALLOCATEFROMPEOF;
603 break;
604
605 case F_VOLPOSMODE:
606 if (alloc_struct.fst_offset <= 0) {
607 error = EINVAL;
608 goto outdrop;
609 }
610
611 alloc_flags |= ALLOCATEFROMVOL;
612 break;
613
614 default: {
615 error = EINVAL;
616 goto outdrop;
617 }
618 }
619 if ( (error = vnode_getwithref(vp)) == 0 ) {
620 /*
621 * call allocate to get the space
622 */
623 error = VNOP_ALLOCATE(vp,alloc_struct.fst_length,alloc_flags,
624 &alloc_struct.fst_bytesalloc, alloc_struct.fst_offset,
625 &context);
626 (void)vnode_put(vp);
627
628 error2 = copyout((caddr_t)&alloc_struct, argp, sizeof (alloc_struct));
629
630 if (error == 0)
631 error = error2;
632 }
633 goto outdrop;
634
635 }
636 case F_SETSIZE:
637 if (fp->f_type != DTYPE_VNODE) {
638 error = EBADF;
639 goto out;
640 }
641 proc_fdunlock(p);
642
643 error = copyin(argp, (caddr_t)&offset, sizeof (off_t));
644 if (error)
645 goto outdrop;
646
647 /*
648 * Make sure that we are root. Growing a file
649 * without zero filling the data is a security hole
650 * root would have access anyway so we'll allow it
651 */
652
653 if (!is_suser()) {
654 error = EACCES;
655 goto outdrop;
656 }
657 vp = (struct vnode *)fp->f_data;
658
659 if ( (error = vnode_getwithref(vp)) == 0 ) {
660 /*
661 * set the file size
662 */
663 error = vnode_setsize(vp, offset, IO_NOZEROFILL, &context);
664
665 (void)vnode_put(vp);
666 }
667 goto outdrop;
668
669 case F_RDAHEAD:
670 if (fp->f_type != DTYPE_VNODE) {
671 error = EBADF;
672 goto out;
673 }
674 vp = (struct vnode *)fp->f_data;
675 proc_fdunlock(p);
676
677 if ( (error = vnode_getwithref(vp)) == 0) {
678 if (uap->arg)
679 vnode_clearnoreadahead(vp);
680 else
681 vnode_setnoreadahead(vp);
682
683 (void)vnode_put(vp);
684 }
685 goto outdrop;
686
687 case F_NOCACHE:
688 if (fp->f_type != DTYPE_VNODE) {
689 error = EBADF;
690 goto out;
691 }
692 vp = (struct vnode *)fp->f_data;
693 proc_fdunlock(p);
694
695 if ( (error = vnode_getwithref(vp)) == 0 ) {
696 if (uap->arg)
697 vnode_setnocache(vp);
698 else
699 vnode_clearnocache(vp);
700
701 (void)vnode_put(vp);
702 }
703 goto outdrop;
704
705 case F_GLOBAL_NOCACHE:
706 if (fp->f_type != DTYPE_VNODE) {
707 error = EBADF;
708 goto out;
709 }
710 vp = (struct vnode *)fp->f_data;
711 proc_fdunlock(p);
712
713 if ( (error = vnode_getwithref(vp)) == 0 ) {
714
715 *retval = vnode_isnocache(vp);
716
717 if (uap->arg)
718 vnode_setnocache(vp);
719 else
720 vnode_clearnocache(vp);
721
722 (void)vnode_put(vp);
723 }
724 goto outdrop;
725
726 case F_RDADVISE: {
727 struct radvisory ra_struct;
728
729 if (fp->f_type != DTYPE_VNODE) {
730 error = EBADF;
731 goto out;
732 }
733 vp = (struct vnode *)fp->f_data;
734 proc_fdunlock(p);
735
736 if ( (error = copyin(argp, (caddr_t)&ra_struct, sizeof (ra_struct))) )
737 goto outdrop;
738 if ( (error = vnode_getwithref(vp)) == 0 ) {
739 error = VNOP_IOCTL(vp, F_RDADVISE, (caddr_t)&ra_struct, 0, &context);
740
741 (void)vnode_put(vp);
742 }
743 goto outdrop;
744 }
745
746 case F_READBOOTSTRAP:
747 case F_WRITEBOOTSTRAP: {
748 fbootstraptransfer_t fbt_struct;
749 user_fbootstraptransfer_t user_fbt_struct;
750 int sizeof_struct;
751 caddr_t boot_structp;
752
753 if (fp->f_type != DTYPE_VNODE) {
754 error = EBADF;
755 goto out;
756 }
757 vp = (struct vnode *)fp->f_data;
758 proc_fdunlock(p);
759
760 if (IS_64BIT_PROCESS(p)) {
761 sizeof_struct = sizeof(user_fbt_struct);
762 boot_structp = (caddr_t) &user_fbt_struct;
763 }
764 else {
765 sizeof_struct = sizeof(fbt_struct);
766 boot_structp = (caddr_t) &fbt_struct;
767 }
768 error = copyin(argp, boot_structp, sizeof_struct);
769 if (error)
770 goto outdrop;
771 if ( (error = vnode_getwithref(vp)) ) {
772 goto outdrop;
773 }
774 if (uap->cmd == F_WRITEBOOTSTRAP) {
775 /*
776 * Make sure that we are root. Updating the
777 * bootstrap on a disk could be a security hole
778 */
779 if (!is_suser()) {
780 (void)vnode_put(vp);
781 error = EACCES;
782 goto outdrop;
783 }
784 }
785 if (strcmp(vnode_mount(vp)->mnt_vfsstat.f_fstypename, "hfs") != 0) {
786 error = EINVAL;
787 } else {
788 /*
789 * call vnop_ioctl to handle the I/O
790 */
791 error = VNOP_IOCTL(vp, uap->cmd, boot_structp, 0, &context);
792 }
793 (void)vnode_put(vp);
794 goto outdrop;
795 }
796 case F_LOG2PHYS: {
797 struct log2phys l2p_struct; /* structure for allocate command */
798
799 if (fp->f_type != DTYPE_VNODE) {
800 error = EBADF;
801 goto out;
802 }
803 vp = (struct vnode *)fp->f_data;
804 proc_fdunlock(p);
805 if ( (error = vnode_getwithref(vp)) ) {
806 goto outdrop;
807 }
808 error = VNOP_OFFTOBLK(vp, fp->f_offset, &lbn);
809 if (error) {
810 (void)vnode_put(vp);
811 goto outdrop;
812 }
813 error = VNOP_BLKTOOFF(vp, lbn, &offset);
814 if (error) {
815 (void)vnode_put(vp);
816 goto outdrop;
817 }
818 devBlockSize = vfs_devblocksize(vnode_mount(vp));
819
820 error = VNOP_BLOCKMAP(vp, offset, devBlockSize, &bn, NULL, NULL, 0, &context);
821
822 (void)vnode_put(vp);
823
824 if (!error) {
825 l2p_struct.l2p_flags = 0; /* for now */
826 l2p_struct.l2p_contigbytes = 0; /* for now */
827 l2p_struct.l2p_devoffset = bn * devBlockSize;
828 l2p_struct.l2p_devoffset += fp->f_offset - offset;
829 error = copyout((caddr_t)&l2p_struct, argp, sizeof (l2p_struct));
830 }
831 goto outdrop;
832 }
833 case F_GETPATH: {
834 char *pathbufp;
835 int pathlen;
836
837 if (fp->f_type != DTYPE_VNODE) {
838 error = EBADF;
839 goto out;
840 }
841 vp = (struct vnode *)fp->f_data;
842 proc_fdunlock(p);
843
844 pathlen = MAXPATHLEN;
845 MALLOC(pathbufp, char *, pathlen, M_TEMP, M_WAITOK);
846 if (pathbufp == NULL) {
847 error = ENOMEM;
848 goto outdrop;
849 }
850 if ( (error = vnode_getwithref(vp)) == 0 ) {
851 error = vn_getpath(vp, pathbufp, &pathlen);
852 (void)vnode_put(vp);
853
854 if (error == 0)
855 error = copyout((caddr_t)pathbufp, argp, pathlen);
856 }
857 FREE(pathbufp, M_TEMP);
858 goto outdrop;
859 }
860
861 case F_PATHPKG_CHECK: {
862 char *pathbufp;
863 size_t pathlen;
864
865 if (fp->f_type != DTYPE_VNODE) {
866 error = EBADF;
867 goto out;
868 }
869 vp = (struct vnode *)fp->f_data;
870 proc_fdunlock(p);
871
872 pathlen = MAXPATHLEN;
873 pathbufp = kalloc(MAXPATHLEN);
874
875 if ( (error = copyinstr(argp, pathbufp, MAXPATHLEN, &pathlen)) == 0 ) {
876 if ( (error = vnode_getwithref(vp)) == 0 ) {
877 error = vn_path_package_check(vp, pathbufp, pathlen, retval);
878
879 (void)vnode_put(vp);
880 }
881 }
882 kfree(pathbufp, MAXPATHLEN);
883 goto outdrop;
884 }
885
886 case F_CHKCLEAN: // used by regression tests to see if all dirty pages got cleaned by fsync()
887 case F_FULLFSYNC: // fsync + flush the journal + DKIOCSYNCHRONIZECACHE
888 case F_FREEZE_FS: // freeze all other fs operations for the fs of this fd
889 case F_THAW_FS: { // thaw all frozen fs operations for the fs of this fd
890 if (fp->f_type != DTYPE_VNODE) {
891 error = EBADF;
892 goto out;
893 }
894 vp = (struct vnode *)fp->f_data;
895 proc_fdunlock(p);
896
897 if ( (error = vnode_getwithref(vp)) == 0 ) {
898 error = VNOP_IOCTL(vp, uap->cmd, (caddr_t)NULL, 0, &context);
899
900 (void)vnode_put(vp);
901 }
902 break;
903 }
904
905 default:
906 if (uap->cmd < FCNTL_FS_SPECIFIC_BASE) {
907 error = EINVAL;
908 goto out;
909 }
910
911 // if it's a fs-specific fcntl() then just pass it through
912
913 if (fp->f_type != DTYPE_VNODE) {
914 error = EBADF;
915 goto out;
916 }
917 vp = (struct vnode *)fp->f_data;
918 proc_fdunlock(p);
919
920 if ( (error = vnode_getwithref(vp)) == 0 ) {
921 error = VNOP_IOCTL(vp, uap->cmd, CAST_DOWN(caddr_t, argp), 0, &context);
922
923 (void)vnode_put(vp);
924 }
925 break;
926
927 }
928
929 outdrop:
930 AUDIT_ARG(vnpath_withref, vp, ARG_VNODE1);
931 fp_drop(p, fd, fp, 0);
932 return(error);
933 out:
934 fp_drop(p, fd, fp, 1);
935 proc_fdunlock(p);
936 return(error);
937 }
938
939 /*
940 * Common code for dup, dup2, and fcntl(F_DUPFD).
941 */
942 int
943 finishdup(struct proc * p, struct filedesc *fdp, int old, int new, register_t *retval)
944 {
945 struct fileproc *nfp;
946 struct fileproc *ofp;
947
948 if ((ofp = fdp->fd_ofiles[old]) == NULL ||
949 (fdp->fd_ofileflags[old] & UF_RESERVED)) {
950 _fdrelse(fdp, new);
951 return (EBADF);
952 }
953 fg_ref(ofp);
954 proc_fdunlock(p);
955
956 MALLOC_ZONE(nfp, struct fileproc *, sizeof(struct fileproc), M_FILEPROC, M_WAITOK);
957 bzero(nfp, sizeof(struct fileproc));
958
959 proc_fdlock(p);
960 nfp->f_flags = ofp->f_flags;
961 nfp->f_fglob = ofp->f_fglob;
962 nfp->f_iocount = 0;
963
964 fdp->fd_ofiles[new] = nfp;
965 fdp->fd_ofileflags[new] = fdp->fd_ofileflags[old] &~ UF_EXCLOSE;
966 if (new > fdp->fd_lastfile)
967 fdp->fd_lastfile = new;
968 *retval = new;
969 return (0);
970 }
971
972
973 int
974 close(struct proc *p, struct close_args *uap, __unused register_t *retval)
975 {
976 struct fileproc *fp;
977 int fd = uap->fd;
978 int error =0;
979
980 AUDIT_SYSCLOSE(p, fd);
981
982 proc_fdlock(p);
983
984 if ( (error = fp_lookup(p,fd,&fp, 1)) ) {
985 proc_fdunlock(p);
986 return(error);
987 }
988
989 error = close_internal(p, fd, fp, CLOSEINT_LOCKED | CLOSEINT_WAITONCLOSE);
990
991 proc_fdunlock(p);
992
993 return(error);
994 }
995
996
997 /*
998 * Close a file descriptor.
999 */
1000 int
1001 close_internal(struct proc *p, int fd, struct fileproc *fp, int flags)
1002 {
1003 struct filedesc *fdp = p->p_fd;
1004 int error =0;
1005 int locked = flags & CLOSEINT_LOCKED;
1006 int waitonclose = flags & CLOSEINT_WAITONCLOSE;
1007 int norelse = flags & CLOSEINT_NOFDRELSE;
1008 int nofdref = flags & CLOSEINT_NOFDNOREF;
1009 int slpstate = PRIBIO;
1010
1011 if (!locked)
1012 proc_fdlock(p);
1013
1014 /* Keep people from using the filedesc while we are closing it */
1015 fdp->fd_ofileflags[fd] |= UF_RESERVED;
1016
1017 fdp->fd_ofileflags[fd] |= UF_CLOSING;
1018
1019
1020 if ((waitonclose && ((fp->f_flags & FP_CLOSING) == FP_CLOSING))) {
1021 if (nofdref == 0)
1022 fp_drop(p, fd, fp, 1);
1023 fp->f_flags |= FP_WAITCLOSE;
1024 if (!locked)
1025 slpstate |= PDROP;
1026 msleep(&fp->f_flags, &p->p_fdmlock, slpstate, "close wait",0) ;
1027 return(EBADF);
1028 }
1029
1030 fp->f_flags |= FP_CLOSING;
1031 if (nofdref)
1032 fp->f_iocount++;
1033
1034 if ( (fp->f_flags & FP_AIOISSUED) || kauth_authorize_fileop_has_listeners() ) {
1035
1036 proc_fdunlock(p);
1037
1038 if ( (fp->f_type == DTYPE_VNODE) && kauth_authorize_fileop_has_listeners() ) {
1039 /*
1040 * call out to allow 3rd party notification of close.
1041 * Ignore result of kauth_authorize_fileop call.
1042 */
1043 if (vnode_getwithref((vnode_t)fp->f_data) == 0) {
1044 u_int fileop_flags = 0;
1045 if ((fp->f_flags & FP_WRITTEN) != 0)
1046 fileop_flags |= KAUTH_FILEOP_CLOSE_MODIFIED;
1047 kauth_authorize_fileop(fp->f_fglob->fg_cred, KAUTH_FILEOP_CLOSE,
1048 (uintptr_t)fp->f_data, (uintptr_t)fileop_flags);
1049 vnode_put((vnode_t)fp->f_data);
1050 }
1051 }
1052 if (fp->f_flags & FP_AIOISSUED)
1053 /*
1054 * cancel all async IO requests that can be cancelled.
1055 */
1056 _aio_close( p, fd );
1057
1058 proc_fdlock(p);
1059 }
1060
1061 if (fd < fdp->fd_knlistsize)
1062 knote_fdclose(p, fd);
1063
1064 if (fp->f_flags & FP_WAITEVENT)
1065 (void)waitevent_close(p, fp);
1066
1067 if ((fp->f_flags & FP_INCHRREAD) == 0)
1068 fileproc_drain(p, fp);
1069 if (norelse == 0)
1070 _fdrelse(fdp, fd);
1071 error = closef_locked(fp, fp->f_fglob, p);
1072 if ((fp->f_flags & FP_WAITCLOSE) == FP_WAITCLOSE)
1073 wakeup(&fp->f_flags);
1074 fp->f_flags &= ~(FP_WAITCLOSE | FP_CLOSING);
1075
1076 if (!locked)
1077 proc_fdunlock(p);
1078
1079 FREE_ZONE(fp, sizeof *fp, M_FILEPROC);
1080 return(error);
1081 }
1082
1083 /*
1084 * Return status information about a file descriptor.
1085 *
1086 * XXX switch on node type is bogus; need a stat in struct fileops instead.
1087 */
1088 static int
1089 fstat1(struct proc *p, int fd, user_addr_t ub, user_addr_t xsecurity, user_addr_t xsecurity_size)
1090 {
1091 struct fileproc *fp;
1092 struct stat sb;
1093 struct user_stat user_sb;
1094 int error, my_size;
1095 int funnel_state;
1096 short type;
1097 caddr_t data;
1098 kauth_filesec_t fsec;
1099 ssize_t xsecurity_bufsize;
1100 int entrycount;
1101 struct vfs_context context;
1102
1103
1104 AUDIT_ARG(fd, fd);
1105
1106 if ((error = fp_lookup(p, fd, &fp, 0)) != 0)
1107 return(error);
1108 type = fp->f_type;
1109 data = fp->f_data;
1110 fsec = KAUTH_FILESEC_NONE;
1111
1112 switch (type) {
1113
1114 case DTYPE_VNODE:
1115 context.vc_proc = current_proc();
1116 context.vc_ucred = kauth_cred_get();
1117 if ((error = vnode_getwithref((vnode_t)data)) == 0) {
1118 /*
1119 * If the caller has the file open, and is not requesting extended security,
1120 * we are going to let them get the basic stat information.
1121 */
1122 if (xsecurity == USER_ADDR_NULL) {
1123 error = vn_stat_noauth((vnode_t)data, &sb, NULL, &context);
1124 } else {
1125 error = vn_stat((vnode_t)data, &sb, &fsec, &context);
1126 }
1127
1128 AUDIT_ARG(vnpath, (struct vnode *)data, ARG_VNODE1);
1129 (void)vnode_put((vnode_t)data);
1130 }
1131 break;
1132
1133 case DTYPE_SOCKET:
1134 error = soo_stat((struct socket *)data, &sb);
1135 break;
1136
1137 case DTYPE_PIPE:
1138 error = pipe_stat((void *)data, &sb);
1139 break;
1140
1141 case DTYPE_PSXSHM:
1142 error = pshm_stat((void *)data, &sb);
1143 break;
1144
1145 case DTYPE_KQUEUE:
1146 funnel_state = thread_funnel_set(kernel_flock, TRUE);
1147 error = kqueue_stat(fp, &sb, p);
1148 thread_funnel_set(kernel_flock, funnel_state);
1149 break;
1150
1151 default:
1152 error = EBADF;
1153 goto out;
1154 }
1155 /* Zap spare fields */
1156 sb.st_lspare = 0;
1157 sb.st_qspare[0] = 0LL;
1158 sb.st_qspare[1] = 0LL;
1159 if (error == 0) {
1160 caddr_t sbp;
1161 if (IS_64BIT_PROCESS(current_proc())) {
1162 munge_stat(&sb, &user_sb);
1163 my_size = sizeof(user_sb);
1164 sbp = (caddr_t)&user_sb;
1165 }
1166 else {
1167 my_size = sizeof(sb);
1168 sbp = (caddr_t)&sb;
1169 }
1170 error = copyout(sbp, ub, my_size);
1171 }
1172
1173 /* caller wants extended security information? */
1174 if (xsecurity != USER_ADDR_NULL) {
1175
1176 /* did we get any? */
1177 if (fsec == KAUTH_FILESEC_NONE) {
1178 if (susize(xsecurity_size, 0) != 0) {
1179 error = EFAULT;
1180 goto out;
1181 }
1182 } else {
1183 /* find the user buffer size */
1184 xsecurity_bufsize = fusize(xsecurity_size);
1185
1186 /* copy out the actual data size */
1187 if (susize(xsecurity_size, KAUTH_FILESEC_COPYSIZE(fsec)) != 0) {
1188 error = EFAULT;
1189 goto out;
1190 }
1191
1192 /* if the caller supplied enough room, copy out to it */
1193 if (xsecurity_bufsize >= KAUTH_FILESEC_COPYSIZE(fsec))
1194 error = copyout(fsec, xsecurity, KAUTH_FILESEC_COPYSIZE(fsec));
1195 }
1196 }
1197 out:
1198 fp_drop(p, fd, fp, 0);
1199 if (fsec != NULL)
1200 kauth_filesec_free(fsec);
1201 return (error);
1202 }
1203
1204 int
1205 fstat_extended(struct proc *p, struct fstat_extended_args *uap, __unused register_t *retval)
1206 {
1207 return(fstat1(p, uap->fd, uap->ub, uap->xsecurity, uap->xsecurity_size));
1208 }
1209
1210 int
1211 fstat(struct proc *p, register struct fstat_args *uap, __unused register_t *retval)
1212 {
1213 return(fstat1(p, uap->fd, uap->ub, 0, 0));
1214 }
1215
1216 /*
1217 * Return pathconf information about a file descriptor.
1218 */
1219 int
1220 fpathconf(p, uap, retval)
1221 struct proc *p;
1222 register struct fpathconf_args *uap;
1223 register_t *retval;
1224 {
1225 int fd = uap->fd;
1226 struct fileproc *fp;
1227 struct vnode *vp;
1228 struct vfs_context context;
1229 int error = 0;
1230 short type;
1231 caddr_t data;
1232
1233
1234 AUDIT_ARG(fd, uap->fd);
1235 if ( (error = fp_lookup(p, fd, &fp, 0)) )
1236 return(error);
1237 type = fp->f_type;
1238 data = fp->f_data;
1239
1240 switch (type) {
1241
1242 case DTYPE_SOCKET:
1243 if (uap->name != _PC_PIPE_BUF) {
1244 error = EINVAL;
1245 goto out;
1246 }
1247 *retval = PIPE_BUF;
1248 error = 0;
1249 goto out;
1250
1251 case DTYPE_PIPE:
1252 *retval = PIPE_BUF;
1253 error = 0;
1254 goto out;
1255
1256 case DTYPE_VNODE:
1257 vp = (struct vnode *)data;
1258
1259 if ( (error = vnode_getwithref(vp)) == 0) {
1260 AUDIT_ARG(vnpath, vp, ARG_VNODE1);
1261
1262 context.vc_proc = p;
1263 context.vc_ucred = kauth_cred_get();
1264
1265 error = vn_pathconf(vp, uap->name, retval, &context);
1266
1267 (void)vnode_put(vp);
1268 }
1269 goto out;
1270
1271 case DTYPE_PSXSHM:
1272 case DTYPE_KQUEUE:
1273 error = EINVAL;
1274 goto out;
1275
1276 default:
1277 panic("fpathconf (unrecognized - %d)", type);
1278 }
1279 /*NOTREACHED*/
1280 out:
1281 fp_drop(p, fd, fp, 0);
1282 return(error);
1283 }
1284
1285 /*
1286 * Allocate a file descriptor for the process.
1287 */
1288 int fdexpand;
1289
1290 int
1291 fdalloc(p, want, result)
1292 struct proc *p;
1293 int want;
1294 int *result;
1295 {
1296 register struct filedesc *fdp = p->p_fd;
1297 register int i;
1298 int lim, last, numfiles, oldnfiles;
1299 struct fileproc **newofiles, **ofiles;
1300 char *newofileflags, *ofileflags;
1301
1302 /*
1303 * Search for a free descriptor starting at the higher
1304 * of want or fd_freefile. If that fails, consider
1305 * expanding the ofile array.
1306 */
1307 lim = min((int)p->p_rlimit[RLIMIT_NOFILE].rlim_cur, maxfiles);
1308 for (;;) {
1309 last = min(fdp->fd_nfiles, lim);
1310 if ((i = want) < fdp->fd_freefile)
1311 i = fdp->fd_freefile;
1312 ofiles = &fdp->fd_ofiles[i];
1313 ofileflags = &fdp->fd_ofileflags[i];
1314 for (; i < last; i++) {
1315 if (*ofiles == NULL && !(*ofileflags & UF_RESERVED)) {
1316 *ofileflags = UF_RESERVED;
1317 if (i > fdp->fd_lastfile)
1318 fdp->fd_lastfile = i;
1319 if (want <= fdp->fd_freefile)
1320 fdp->fd_freefile = i;
1321 *result = i;
1322 return (0);
1323 }
1324 ofiles++; ofileflags++;
1325 }
1326
1327 /*
1328 * No space in current array. Expand?
1329 */
1330 if (fdp->fd_nfiles >= lim)
1331 return (EMFILE);
1332 if (fdp->fd_nfiles < NDEXTENT)
1333 numfiles = NDEXTENT;
1334 else
1335 numfiles = 2 * fdp->fd_nfiles;
1336 /* Enforce lim */
1337 if (numfiles > lim)
1338 numfiles = lim;
1339 proc_fdunlock(p);
1340 MALLOC_ZONE(newofiles, struct fileproc **,
1341 numfiles * OFILESIZE, M_OFILETABL, M_WAITOK);
1342 proc_fdlock(p);
1343 if (newofiles == NULL) {
1344 return (ENOMEM);
1345 }
1346 if (fdp->fd_nfiles >= numfiles) {
1347 FREE_ZONE(newofiles, numfiles * OFILESIZE, M_OFILETABL);
1348 continue;
1349 }
1350 newofileflags = (char *) &newofiles[numfiles];
1351 /*
1352 * Copy the existing ofile and ofileflags arrays
1353 * and zero the new portion of each array.
1354 */
1355 oldnfiles = fdp->fd_nfiles;
1356 (void) memcpy(newofiles, fdp->fd_ofiles,
1357 oldnfiles * sizeof *fdp->fd_ofiles);
1358 (void) memset(&newofiles[oldnfiles], 0,
1359 (numfiles - oldnfiles) * sizeof *fdp->fd_ofiles);
1360
1361 (void) memcpy(newofileflags, fdp->fd_ofileflags,
1362 oldnfiles * sizeof *fdp->fd_ofileflags);
1363 (void) memset(&newofileflags[oldnfiles], 0,
1364 (numfiles - oldnfiles) *
1365 sizeof *fdp->fd_ofileflags);
1366 ofiles = fdp->fd_ofiles;
1367 fdp->fd_ofiles = newofiles;
1368 fdp->fd_ofileflags = newofileflags;
1369 fdp->fd_nfiles = numfiles;
1370 FREE_ZONE(ofiles, oldnfiles * OFILESIZE, M_OFILETABL);
1371 fdexpand++;
1372 }
1373 }
1374
1375 /*
1376 * Check to see whether n user file descriptors
1377 * are available to the process p.
1378 */
1379 int
1380 fdavail(p, n)
1381 struct proc *p;
1382 int n;
1383 {
1384 struct filedesc *fdp = p->p_fd;
1385 struct fileproc **fpp;
1386 char *flags;
1387 int i, lim;
1388
1389 lim = min((int)p->p_rlimit[RLIMIT_NOFILE].rlim_cur, maxfiles);
1390 if ((i = lim - fdp->fd_nfiles) > 0 && (n -= i) <= 0)
1391 return (1);
1392 fpp = &fdp->fd_ofiles[fdp->fd_freefile];
1393 flags = &fdp->fd_ofileflags[fdp->fd_freefile];
1394 for (i = fdp->fd_nfiles - fdp->fd_freefile; --i >= 0; fpp++, flags++)
1395 if (*fpp == NULL && !(*flags & UF_RESERVED) && --n <= 0)
1396 return (1);
1397 return (0);
1398 }
1399
1400 void
1401 fdrelse(p, fd)
1402 struct proc *p;
1403 int fd;
1404 {
1405 _fdrelse(p->p_fd, fd);
1406 }
1407
1408 void
1409 fddrop(p, fd)
1410 struct proc *p;
1411 int fd;
1412 {
1413 struct filedesc *fdp = p->p_fd;
1414 struct fileproc *fp;
1415
1416 if (fd < fdp->fd_freefile)
1417 fdp->fd_freefile = fd;
1418 #if DIAGNOSTIC
1419 if (fd > fdp->fd_lastfile)
1420 panic("fdrelse: fd_lastfile inconsistent");
1421 #endif
1422 fp = fdp->fd_ofiles[fd];
1423 fdp->fd_ofiles[fd] = NULL;
1424 fdp->fd_ofileflags[fd] = 0;
1425
1426 while ((fd = fdp->fd_lastfile) > 0 &&
1427 fdp->fd_ofiles[fd] == NULL &&
1428 !(fdp->fd_ofileflags[fd] & UF_RESERVED))
1429 fdp->fd_lastfile--;
1430 FREE_ZONE(fp, sizeof *fp, M_FILEPROC);
1431 }
1432
1433
1434 int
1435 fdgetf_noref(p, fd, resultfp)
1436 struct proc *p;
1437 int fd;
1438 struct fileproc **resultfp;
1439 {
1440 struct filedesc *fdp = p->p_fd;
1441 struct fileproc *fp;
1442
1443 if (fd < 0 || fd >= fdp->fd_nfiles ||
1444 (fp = fdp->fd_ofiles[fd]) == NULL ||
1445 (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
1446 return (EBADF);
1447 }
1448 if (resultfp)
1449 *resultfp = fp;
1450 return (0);
1451 }
1452
1453
1454 /* should be called only when proc_fdlock is held */
1455 void
1456 fp_setflags(proc_t p, struct fileproc * fp, int flags)
1457 {
1458 proc_fdlock(p);
1459 fp->f_flags |= flags;
1460 proc_fdunlock(p);
1461 }
1462
1463 void
1464 fp_clearflags(proc_t p, struct fileproc * fp, int flags)
1465 {
1466
1467 proc_fdlock(p);
1468 if (fp)
1469 fp->f_flags &= ~flags;
1470 proc_fdunlock(p);
1471 }
1472
1473 int
1474 fp_getfvp(p, fd, resultfp, resultvp)
1475 struct proc *p;
1476 int fd;
1477 struct fileproc **resultfp;
1478 struct vnode **resultvp;
1479 {
1480 struct filedesc *fdp = p->p_fd;
1481 struct fileproc *fp;
1482
1483 proc_fdlock(p);
1484 if (fd < 0 || fd >= fdp->fd_nfiles ||
1485 (fp = fdp->fd_ofiles[fd]) == NULL ||
1486 (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
1487 proc_fdunlock(p);
1488 return (EBADF);
1489 }
1490 if (fp->f_type != DTYPE_VNODE) {
1491 proc_fdunlock(p);
1492 return(ENOTSUP);
1493 }
1494 fp->f_iocount++;
1495
1496 if (resultfp)
1497 *resultfp = fp;
1498 if (resultvp)
1499 *resultvp = (struct vnode *)fp->f_data;
1500 proc_fdunlock(p);
1501
1502 return (0);
1503 }
1504
1505
1506 /*
1507 * Returns: EBADF The file descriptor is invalid
1508 * EOPNOTSUPP The file descriptor is not a socket
1509 * 0 Success
1510 *
1511 * Notes: EOPNOTSUPP should probably be ENOTSOCK; this function is only
1512 * ever called from accept1().
1513 */
1514 int
1515 fp_getfsock(p, fd, resultfp, results)
1516 struct proc *p;
1517 int fd;
1518 struct fileproc **resultfp;
1519 struct socket **results;
1520 {
1521 struct filedesc *fdp = p->p_fd;
1522 struct fileproc *fp;
1523
1524 proc_fdlock(p);
1525 if (fd < 0 || fd >= fdp->fd_nfiles ||
1526 (fp = fdp->fd_ofiles[fd]) == NULL ||
1527 (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
1528 proc_fdunlock(p);
1529 return (EBADF);
1530 }
1531 if (fp->f_type != DTYPE_SOCKET) {
1532 proc_fdunlock(p);
1533 return(EOPNOTSUPP);
1534 }
1535 fp->f_iocount++;
1536
1537 if (resultfp)
1538 *resultfp = fp;
1539 if (results)
1540 *results = (struct socket *)fp->f_data;
1541 proc_fdunlock(p);
1542
1543 return (0);
1544 }
1545
1546
1547 int
1548 fp_getfkq(p, fd, resultfp, resultkq)
1549 struct proc *p;
1550 int fd;
1551 struct fileproc **resultfp;
1552 struct kqueue **resultkq;
1553 {
1554 struct filedesc *fdp = p->p_fd;
1555 struct fileproc *fp;
1556
1557 proc_fdlock(p);
1558 if ( fd < 0 || fd >= fdp->fd_nfiles ||
1559 (fp = fdp->fd_ofiles[fd]) == NULL ||
1560 (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
1561 proc_fdunlock(p);
1562 return (EBADF);
1563 }
1564 if (fp->f_type != DTYPE_KQUEUE) {
1565 proc_fdunlock(p);
1566 return(EBADF);
1567 }
1568 fp->f_iocount++;
1569
1570 if (resultfp)
1571 *resultfp = fp;
1572 if (resultkq)
1573 *resultkq = (struct kqueue *)fp->f_data;
1574 proc_fdunlock(p);
1575
1576 return (0);
1577 }
1578
1579 int
1580 fp_getfpshm(p, fd, resultfp, resultpshm)
1581 struct proc *p;
1582 int fd;
1583 struct fileproc **resultfp;
1584 struct pshmnode **resultpshm;
1585 {
1586 struct filedesc *fdp = p->p_fd;
1587 struct fileproc *fp;
1588
1589 proc_fdlock(p);
1590 if (fd < 0 || fd >= fdp->fd_nfiles ||
1591 (fp = fdp->fd_ofiles[fd]) == NULL ||
1592 (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
1593 proc_fdunlock(p);
1594 return (EBADF);
1595 }
1596 if (fp->f_type != DTYPE_PSXSHM) {
1597
1598 proc_fdunlock(p);
1599 return(EBADF);
1600 }
1601 fp->f_iocount++;
1602
1603 if (resultfp)
1604 *resultfp = fp;
1605 if (resultpshm)
1606 *resultpshm = (struct pshmnode *)fp->f_data;
1607 proc_fdunlock(p);
1608
1609 return (0);
1610 }
1611
1612
1613 int
1614 fp_getfpsem(p, fd, resultfp, resultpsem)
1615 struct proc *p;
1616 int fd;
1617 struct fileproc **resultfp;
1618 struct psemnode **resultpsem;
1619 {
1620 struct filedesc *fdp = p->p_fd;
1621 struct fileproc *fp;
1622
1623 proc_fdlock(p);
1624 if (fd < 0 || fd >= fdp->fd_nfiles ||
1625 (fp = fdp->fd_ofiles[fd]) == NULL ||
1626 (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
1627 proc_fdunlock(p);
1628 return (EBADF);
1629 }
1630 if (fp->f_type != DTYPE_PSXSEM) {
1631 proc_fdunlock(p);
1632 return(EBADF);
1633 }
1634 fp->f_iocount++;
1635
1636 if (resultfp)
1637 *resultfp = fp;
1638 if (resultpsem)
1639 *resultpsem = (struct psemnode *)fp->f_data;
1640 proc_fdunlock(p);
1641
1642 return (0);
1643 }
1644 int
1645 fp_lookup(p, fd, resultfp, locked)
1646 struct proc *p;
1647 int fd;
1648 struct fileproc **resultfp;
1649 int locked;
1650 {
1651 struct filedesc *fdp = p->p_fd;
1652 struct fileproc *fp;
1653
1654 if (!locked)
1655 proc_fdlock(p);
1656 if (fd < 0 || fd >= fdp->fd_nfiles ||
1657 (fp = fdp->fd_ofiles[fd]) == NULL ||
1658 (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
1659 if (!locked)
1660 proc_fdunlock(p);
1661 return (EBADF);
1662 }
1663 fp->f_iocount++;
1664
1665 if (resultfp)
1666 *resultfp = fp;
1667 if (!locked)
1668 proc_fdunlock(p);
1669
1670 return (0);
1671 }
1672
1673 int
1674 fp_drop_written(proc_t p, int fd, struct fileproc *fp)
1675 {
1676 int error;
1677
1678 proc_fdlock(p);
1679
1680 fp->f_flags |= FP_WRITTEN;
1681
1682 error = fp_drop(p, fd, fp, 1);
1683
1684 proc_fdunlock(p);
1685
1686 return (error);
1687 }
1688
1689
1690 int
1691 fp_drop_event(proc_t p, int fd, struct fileproc *fp)
1692 {
1693 int error;
1694
1695 proc_fdlock(p);
1696
1697 fp->f_flags |= FP_WAITEVENT;
1698
1699 error = fp_drop(p, fd, fp, 1);
1700
1701 proc_fdunlock(p);
1702
1703 return (error);
1704 }
1705
1706 int
1707 fp_drop(p, fd, fp, locked)
1708 struct proc *p;
1709 int fd;
1710 struct fileproc *fp;
1711 int locked;
1712 {
1713 struct filedesc *fdp = p->p_fd;
1714
1715 if (!locked)
1716 proc_fdlock(p);
1717 if ((fp == FILEPROC_NULL) && (fd < 0 || fd >= fdp->fd_nfiles ||
1718 (fp = fdp->fd_ofiles[fd]) == NULL ||
1719 ((fdp->fd_ofileflags[fd] & UF_RESERVED) &&
1720 !(fdp->fd_ofileflags[fd] & UF_CLOSING)))) {
1721 if (!locked)
1722 proc_fdunlock(p);
1723 return (EBADF);
1724 }
1725 fp->f_iocount--;
1726
1727 if (p->p_fpdrainwait && fp->f_iocount == 0) {
1728 p->p_fpdrainwait = 0;
1729 wakeup(&p->p_fpdrainwait);
1730 }
1731 if (!locked)
1732 proc_fdunlock(p);
1733
1734 return (0);
1735 }
1736
1737 int
1738 file_vnode(int fd, struct vnode **vpp)
1739 {
1740 struct proc * p = current_proc();
1741 struct fileproc *fp;
1742 int error;
1743
1744 proc_fdlock(p);
1745 if ( (error = fp_lookup(p, fd, &fp, 1)) ) {
1746 proc_fdunlock(p);
1747 return(error);
1748 }
1749 if (fp->f_type != DTYPE_VNODE) {
1750 fp_drop(p, fd, fp,1);
1751 proc_fdunlock(p);
1752 return(EINVAL);
1753 }
1754 *vpp = (struct vnode *)fp->f_data;
1755 proc_fdunlock(p);
1756
1757 return(0);
1758 }
1759
1760
1761 int
1762 file_socket(int fd, struct socket **sp)
1763 {
1764 struct proc * p = current_proc();
1765 struct fileproc *fp;
1766 int error;
1767
1768 proc_fdlock(p);
1769 if ( (error = fp_lookup(p, fd, &fp, 1)) ) {
1770 proc_fdunlock(p);
1771 return(error);
1772 }
1773 if (fp->f_type != DTYPE_SOCKET) {
1774 fp_drop(p, fd, fp,1);
1775 proc_fdunlock(p);
1776 return(ENOTSOCK);
1777 }
1778 *sp = (struct socket *)fp->f_data;
1779 proc_fdunlock(p);
1780
1781 return(0);
1782 }
1783
1784 int
1785 file_flags(int fd, int * flags)
1786 {
1787
1788 struct proc * p = current_proc();
1789 struct fileproc *fp;
1790 int error;
1791
1792 proc_fdlock(p);
1793 if ( (error = fp_lookup(p, fd, &fp, 1)) ) {
1794 proc_fdunlock(p);
1795 return(error);
1796 }
1797 *flags = (int)fp->f_flag;
1798 fp_drop(p, fd, fp,1);
1799 proc_fdunlock(p);
1800
1801 return(0);
1802 }
1803
1804
1805 int
1806 file_drop(int fd)
1807 {
1808 struct fileproc *fp;
1809 struct proc *p = current_proc();
1810
1811 proc_fdlock(p);
1812 if (fd < 0 || fd >= p->p_fd->fd_nfiles ||
1813 (fp = p->p_fd->fd_ofiles[fd]) == NULL ||
1814 ((p->p_fd->fd_ofileflags[fd] & UF_RESERVED) &&
1815 !(p->p_fd->fd_ofileflags[fd] & UF_CLOSING))) {
1816 proc_fdunlock(p);
1817 return (EBADF);
1818 }
1819 fp->f_iocount --;
1820
1821 if (p->p_fpdrainwait && fp->f_iocount == 0) {
1822 p->p_fpdrainwait = 0;
1823 wakeup(&p->p_fpdrainwait);
1824 }
1825 proc_fdunlock(p);
1826 return(0);
1827
1828
1829 }
1830
1831 int
1832 falloc(p, resultfp, resultfd )
1833 struct proc *p;
1834 struct fileproc **resultfp;
1835 int *resultfd;
1836 {
1837 int error;
1838
1839 proc_fdlock(p);
1840 error = falloc_locked(p, resultfp, resultfd, 1);
1841 proc_fdunlock(p);
1842
1843 return(error);
1844 }
1845 /*
1846 * Create a new open file structure and allocate
1847 * a file decriptor for the process that refers to it.
1848 */
1849 int
1850 falloc_locked(p, resultfp, resultfd, locked)
1851 struct proc *p;
1852 struct fileproc **resultfp;
1853 int *resultfd;
1854 int locked;
1855 {
1856 struct fileproc *fp, *fq;
1857 struct fileglob *fg;
1858 int error, nfd;
1859
1860 if (!locked)
1861 proc_fdlock(p);
1862 if ( (error = fdalloc(p, 0, &nfd)) ) {
1863 if (!locked)
1864 proc_fdunlock(p);
1865 return (error);
1866 }
1867 if (nfiles >= maxfiles) {
1868 if (!locked)
1869 proc_fdunlock(p);
1870 tablefull("file");
1871 return (ENFILE);
1872 }
1873 /*
1874 * Allocate a new file descriptor.
1875 * If the process has file descriptor zero open, add to the list
1876 * of open files at that point, otherwise put it at the front of
1877 * the list of open files.
1878 */
1879 proc_fdunlock(p);
1880
1881 MALLOC_ZONE(fp, struct fileproc *, sizeof(struct fileproc), M_FILEPROC, M_WAITOK);
1882 MALLOC_ZONE(fg, struct fileglob *, sizeof(struct fileglob), M_FILEGLOB, M_WAITOK);
1883 bzero(fp, sizeof(struct fileproc));
1884 bzero(fg, sizeof(struct fileglob));
1885 lck_mtx_init(&fg->fg_lock, file_lck_grp, file_lck_attr);
1886
1887 fp->f_iocount = 1;
1888 fg->fg_count = 1;
1889 fp->f_fglob = fg;
1890
1891 proc_fdlock(p);
1892
1893 fp->f_cred = kauth_cred_proc_ref(p);
1894
1895 lck_mtx_lock(file_flist_lock);
1896
1897 nfiles++;
1898
1899 if ( (fq = p->p_fd->fd_ofiles[0]) ) {
1900 LIST_INSERT_AFTER(fq->f_fglob, fg, f_list);
1901 } else {
1902 LIST_INSERT_HEAD(&filehead, fg, f_list);
1903 }
1904 lck_mtx_unlock(file_flist_lock);
1905
1906 p->p_fd->fd_ofiles[nfd] = fp;
1907
1908 if (!locked)
1909 proc_fdunlock(p);
1910
1911 if (resultfp)
1912 *resultfp = fp;
1913 if (resultfd)
1914 *resultfd = nfd;
1915
1916 return (0);
1917 }
1918
1919 /*
1920 * Free a file structure.
1921 */
1922 void
1923 fg_free(fg)
1924 struct fileglob *fg;
1925 {
1926 kauth_cred_t cred;
1927
1928 lck_mtx_lock(file_flist_lock);
1929 LIST_REMOVE(fg, f_list);
1930 nfiles--;
1931 lck_mtx_unlock(file_flist_lock);
1932
1933 cred = fg->fg_cred;
1934 if (cred != NOCRED) {
1935 fg->fg_cred = NOCRED;
1936 kauth_cred_rele(cred);
1937 }
1938 lck_mtx_destroy(&fg->fg_lock, file_lck_grp);
1939
1940 FREE_ZONE(fg, sizeof *fg, M_FILEGLOB);
1941 }
1942
1943 void
1944 fdexec(p)
1945 struct proc *p;
1946 {
1947 struct filedesc *fdp = p->p_fd;
1948 int i = fdp->fd_lastfile;
1949 struct fileproc **fpp = &fdp->fd_ofiles[i];
1950 char *flags = &fdp->fd_ofileflags[i];
1951 int funnel_state;
1952
1953 funnel_state = thread_funnel_set(kernel_flock, FALSE);
1954 proc_fdlock(p);
1955
1956 while (i >= 0) {
1957 if ((*flags & (UF_RESERVED|UF_EXCLOSE)) == UF_EXCLOSE) {
1958 struct fileproc *fp = *fpp;
1959
1960 if (i < fdp->fd_knlistsize)
1961 knote_fdclose(p, i);
1962
1963 *fpp = NULL; *flags = 0;
1964 if (i == fdp->fd_lastfile && i > 0)
1965 fdp->fd_lastfile--;
1966 closef_locked(fp, fp->f_fglob, p);
1967 FREE_ZONE(fp, sizeof *fp, M_FILEPROC);
1968 }
1969
1970 i--; fpp--; flags--;
1971 }
1972 proc_fdunlock(p);
1973 thread_funnel_set(kernel_flock, funnel_state);
1974 }
1975
1976 /*
1977 * Copy a filedesc structure.
1978 */
1979 struct filedesc *
1980 fdcopy(p)
1981 struct proc *p;
1982 {
1983 struct filedesc *newfdp, *fdp = p->p_fd;
1984 int i;
1985 struct fileproc *ofp, *fp;
1986 vnode_t v_dir;
1987
1988 MALLOC_ZONE(newfdp, struct filedesc *,
1989 sizeof *newfdp, M_FILEDESC, M_WAITOK);
1990 if (newfdp == NULL)
1991 return(NULL);
1992
1993 proc_fdlock(p);
1994
1995 /*
1996 * the FD_CHROOT flag will be inherited via this copy
1997 */
1998 (void) memcpy(newfdp, fdp, sizeof *newfdp);
1999
2000 /*
2001 * for both fd_cdir and fd_rdir make sure we get
2002 * a valid reference... if we can't, than set
2003 * set the pointer(s) to NULL in the child... this
2004 * will keep us from using a non-referenced vp
2005 * and allows us to do the vnode_rele only on
2006 * a properly referenced vp
2007 */
2008 if ( (v_dir = newfdp->fd_cdir) ) {
2009 if (vnode_getwithref(v_dir) == 0) {
2010 if ( (vnode_ref(v_dir)) )
2011 newfdp->fd_cdir = NULL;
2012 vnode_put(v_dir);
2013 } else
2014 newfdp->fd_cdir = NULL;
2015 }
2016 if (newfdp->fd_cdir == NULL && fdp->fd_cdir) {
2017 /*
2018 * we couldn't get a new reference on
2019 * the current working directory being
2020 * inherited... we might as well drop
2021 * our reference from the parent also
2022 * since the vnode has gone DEAD making
2023 * it useless... by dropping it we'll
2024 * be that much closer to recyling it
2025 */
2026 vnode_rele(fdp->fd_cdir);
2027 fdp->fd_cdir = NULL;
2028 }
2029
2030 if ( (v_dir = newfdp->fd_rdir) ) {
2031 if (vnode_getwithref(v_dir) == 0) {
2032 if ( (vnode_ref(v_dir)) )
2033 newfdp->fd_rdir = NULL;
2034 vnode_put(v_dir);
2035 } else
2036 newfdp->fd_rdir = NULL;
2037 }
2038 if (newfdp->fd_rdir == NULL && fdp->fd_rdir) {
2039 /*
2040 * we couldn't get a new reference on
2041 * the root directory being
2042 * inherited... we might as well drop
2043 * our reference from the parent also
2044 * since the vnode has gone DEAD making
2045 * it useless... by dropping it we'll
2046 * be that much closer to recyling it
2047 */
2048 vnode_rele(fdp->fd_rdir);
2049 fdp->fd_rdir = NULL;
2050 }
2051 newfdp->fd_refcnt = 1;
2052
2053 /*
2054 * If the number of open files fits in the internal arrays
2055 * of the open file structure, use them, otherwise allocate
2056 * additional memory for the number of descriptors currently
2057 * in use.
2058 */
2059 if (newfdp->fd_lastfile < NDFILE)
2060 i = NDFILE;
2061 else {
2062 /*
2063 * Compute the smallest multiple of NDEXTENT needed
2064 * for the file descriptors currently in use,
2065 * allowing the table to shrink.
2066 */
2067 i = newfdp->fd_nfiles;
2068 while (i > 2 * NDEXTENT && i > newfdp->fd_lastfile * 2)
2069 i /= 2;
2070 }
2071 proc_fdunlock(p);
2072
2073 MALLOC_ZONE(newfdp->fd_ofiles, struct fileproc **,
2074 i * OFILESIZE, M_OFILETABL, M_WAITOK);
2075 if (newfdp->fd_ofiles == NULL) {
2076 if (newfdp->fd_cdir)
2077 vnode_rele(newfdp->fd_cdir);
2078 if (newfdp->fd_rdir)
2079 vnode_rele(newfdp->fd_rdir);
2080
2081 FREE_ZONE(newfdp, sizeof *newfdp, M_FILEDESC);
2082 return(NULL);
2083 }
2084 proc_fdlock(p);
2085
2086 newfdp->fd_ofileflags = (char *) &newfdp->fd_ofiles[i];
2087 newfdp->fd_nfiles = i;
2088
2089 if (fdp->fd_nfiles > 0) {
2090 struct fileproc **fpp;
2091 char *flags;
2092
2093 (void) memcpy(newfdp->fd_ofiles, fdp->fd_ofiles,
2094 i * sizeof *fdp->fd_ofiles);
2095 (void) memcpy(newfdp->fd_ofileflags, fdp->fd_ofileflags,
2096 i * sizeof *fdp->fd_ofileflags);
2097
2098 /*
2099 * kq descriptors cannot be copied.
2100 */
2101 if (newfdp->fd_knlistsize != -1) {
2102 fpp = &newfdp->fd_ofiles[newfdp->fd_lastfile];
2103 for (i = newfdp->fd_lastfile; i >= 0; i--, fpp--) {
2104 if (*fpp != NULL && (*fpp)->f_type == DTYPE_KQUEUE) {
2105 *fpp = NULL;
2106 if (i < newfdp->fd_freefile)
2107 newfdp->fd_freefile = i;
2108 }
2109 if (*fpp == NULL && i == newfdp->fd_lastfile && i > 0)
2110 newfdp->fd_lastfile--;
2111 }
2112 newfdp->fd_knlist = NULL;
2113 newfdp->fd_knlistsize = -1;
2114 newfdp->fd_knhash = NULL;
2115 newfdp->fd_knhashmask = 0;
2116 }
2117 fpp = newfdp->fd_ofiles;
2118 flags = newfdp->fd_ofileflags;
2119
2120 for (i = newfdp->fd_lastfile; i-- >= 0; fpp++, flags++)
2121 if ((ofp = *fpp) != NULL && !(*flags & UF_RESERVED)) {
2122 MALLOC_ZONE(fp, struct fileproc *, sizeof(struct fileproc), M_FILEPROC, M_WAITOK);
2123 bzero(fp, sizeof(struct fileproc));
2124 fp->f_flags = ofp->f_flags;
2125 //fp->f_iocount = ofp->f_iocount;
2126 fp->f_iocount = 0;
2127 fp->f_fglob = ofp->f_fglob;
2128 (void)fg_ref(fp);
2129 *fpp = fp;
2130 } else {
2131 *fpp = NULL;
2132 *flags = 0;
2133 }
2134 } else
2135 (void) memset(newfdp->fd_ofiles, 0, i * OFILESIZE);
2136
2137 proc_fdunlock(p);
2138 return (newfdp);
2139 }
2140
2141 /*
2142 * Release a filedesc structure.
2143 */
2144 void
2145 fdfree(p)
2146 struct proc *p;
2147 {
2148 struct filedesc *fdp;
2149 struct fileproc *fp;
2150 int i;
2151
2152 proc_fdlock(p);
2153
2154 /* Certain daemons might not have file descriptors */
2155 fdp = p->p_fd;
2156
2157 if ((fdp == NULL) || (--fdp->fd_refcnt > 0)) {
2158 proc_fdunlock(p);
2159 return;
2160 }
2161 if (fdp->fd_refcnt == 0xffff)
2162 panic("fdfree: bad fd_refcnt");
2163
2164 /* Last reference: the structure can't change out from under us */
2165
2166 if (fdp->fd_nfiles > 0 && fdp->fd_ofiles) {
2167 for (i = fdp->fd_lastfile; i >= 0; i--) {
2168 if ((fp = fdp->fd_ofiles[i]) != NULL) {
2169
2170 if (fdp->fd_ofileflags[i] & UF_RESERVED)
2171 panic("fdfree: found fp with UF_RESERVED\n");
2172
2173 /* closef drops the iocount ... */
2174 if ((fp->f_flags & FP_INCHRREAD) != 0)
2175 fp->f_iocount++;
2176 fdp->fd_ofiles[i] = NULL;
2177 fdp->fd_ofileflags[i] |= UF_RESERVED;
2178
2179 if (i < fdp->fd_knlistsize)
2180 knote_fdclose(p, i);
2181 if (fp->f_flags & FP_WAITEVENT)
2182 (void)waitevent_close(p, fp);
2183 (void) closef_locked(fp, fp->f_fglob, p);
2184 FREE_ZONE(fp, sizeof *fp, M_FILEPROC);
2185 }
2186 }
2187 FREE_ZONE(fdp->fd_ofiles, fdp->fd_nfiles * OFILESIZE, M_OFILETABL);
2188 fdp->fd_ofiles = NULL;
2189 fdp->fd_nfiles = 0;
2190 }
2191
2192 proc_fdunlock(p);
2193
2194 if (fdp->fd_cdir)
2195 vnode_rele(fdp->fd_cdir);
2196 if (fdp->fd_rdir)
2197 vnode_rele(fdp->fd_rdir);
2198
2199 proc_fdlock(p);
2200 p->p_fd = NULL;
2201 proc_fdunlock(p);
2202
2203 if (fdp->fd_knlist)
2204 FREE(fdp->fd_knlist, M_KQUEUE);
2205 if (fdp->fd_knhash)
2206 FREE(fdp->fd_knhash, M_KQUEUE);
2207
2208 FREE_ZONE(fdp, sizeof *fdp, M_FILEDESC);
2209 }
2210
2211 static int
2212 closef_finish(fp, fg, p)
2213 struct fileproc *fp;
2214 struct fileglob *fg;
2215 struct proc *p;
2216 {
2217 struct vnode *vp;
2218 struct flock lf;
2219 int error;
2220 struct vfs_context context;
2221
2222 if ((fg->fg_flag & FHASLOCK) && fg->fg_type == DTYPE_VNODE) {
2223 lf.l_whence = SEEK_SET;
2224 lf.l_start = 0;
2225 lf.l_len = 0;
2226 lf.l_type = F_UNLCK;
2227 vp = (struct vnode *)fg->fg_data;
2228 context.vc_proc = p;
2229 context.vc_ucred = fg->fg_cred;
2230
2231 (void) VNOP_ADVLOCK(vp, (caddr_t)fg, F_UNLCK, &lf, F_FLOCK, &context);
2232 }
2233 if (fg->fg_ops)
2234 error = fo_close(fg, p);
2235 else
2236 error = 0;
2237
2238 if (((fp != (struct fileproc *)0) && ((fp->f_flags & FP_INCHRREAD) != 0))) {
2239 proc_fdlock(p);
2240 if ( ((fp->f_flags & FP_INCHRREAD) != 0) ) {
2241 fileproc_drain(p, fp);
2242 }
2243 proc_fdunlock(p);
2244 }
2245 fg_free(fg);
2246
2247 return (error);
2248 }
2249
2250 int
2251 closef(fg, p)
2252 struct fileglob *fg;
2253 struct proc *p;
2254 {
2255 int error;
2256
2257 proc_fdlock(p);
2258 error = closef_locked((struct fileproc *)0, fg, p);
2259 proc_fdunlock(p);
2260
2261 return(error);
2262 }
2263 /*
2264 * Internal form of close.
2265 * Decrement reference count on file structure.
2266 * Note: p may be NULL when closing a file
2267 * that was being passed in a message.
2268 */
2269 int
2270 closef_locked(fp, fg, p)
2271 struct fileproc *fp;
2272 struct fileglob *fg;
2273 struct proc *p;
2274 {
2275 struct vnode *vp;
2276 struct flock lf;
2277 struct vfs_context context;
2278 int error;
2279
2280 if (fg == NULL) {
2281 return (0);
2282 }
2283 /*
2284 * POSIX record locking dictates that any close releases ALL
2285 * locks owned by this process. This is handled by setting
2286 * a flag in the unlock to free ONLY locks obeying POSIX
2287 * semantics, and not to free BSD-style file locks.
2288 * If the descriptor was in a message, POSIX-style locks
2289 * aren't passed with the descriptor.
2290 */
2291 if (p && (p->p_ladvflag & P_LADVLOCK) && fg->fg_type == DTYPE_VNODE) {
2292 proc_fdunlock(p);
2293
2294 lf.l_whence = SEEK_SET;
2295 lf.l_start = 0;
2296 lf.l_len = 0;
2297 lf.l_type = F_UNLCK;
2298 vp = (struct vnode *)fg->fg_data;
2299
2300 if ( (error = vnode_getwithref(vp)) == 0 ) {
2301 context.vc_proc = p;
2302 context.vc_ucred = fg->fg_cred;
2303 (void) VNOP_ADVLOCK(vp, (caddr_t)p, F_UNLCK, &lf, F_POSIX, &context);
2304
2305 (void)vnode_put(vp);
2306 }
2307 proc_fdlock(p);
2308 }
2309 lck_mtx_lock(&fg->fg_lock);
2310 fg->fg_count--;
2311
2312 if (fg->fg_count > 0) {
2313 lck_mtx_unlock(&fg->fg_lock);
2314 return (0);
2315 }
2316 if (fg->fg_count != 0)
2317 panic("fg: being freed with bad fg_count (%d)", fg, fg->fg_count);
2318
2319 if (fp && (fp->f_flags & FP_WRITTEN))
2320 fg->fg_flag |= FWASWRITTEN;
2321
2322 fg->fg_lflags |= FG_TERM;
2323 lck_mtx_unlock(&fg->fg_lock);
2324
2325 proc_fdunlock(p);
2326 error = closef_finish(fp, fg, p);
2327 proc_fdlock(p);
2328
2329 return(error);
2330 }
2331
2332
2333 extern int selwait;
2334 void
2335 fileproc_drain(struct proc *p, struct fileproc * fp)
2336 {
2337 fp->f_iocount-- ; /* (the one the close holds) */
2338
2339 while (fp->f_iocount) {
2340 if (((fp->f_flags & FP_INSELECT)== FP_INSELECT)) {
2341 wait_queue_wakeup_all((wait_queue_t)fp->f_waddr, &selwait, THREAD_INTERRUPTED);
2342 } else {
2343 if (fp->f_fglob->fg_ops->fo_drain) {
2344 (*fp->f_fglob->fg_ops->fo_drain)(fp, p);
2345 }
2346 }
2347 p->p_fpdrainwait = 1;
2348
2349 msleep(&p->p_fpdrainwait, &p->p_fdmlock, PRIBIO, "fpdrain",0);
2350
2351 //panic("successful wait after drain\n");
2352 }
2353 }
2354
2355 int
2356 fp_free(struct proc * p, int fd, struct fileproc * fp)
2357 {
2358 proc_fdlock(p);
2359 fdrelse(p, fd);
2360 proc_fdunlock(p);
2361
2362 fg_free(fp->f_fglob);
2363 FREE_ZONE(fp, sizeof *fp, M_FILEPROC);
2364 }
2365
2366
2367 /*
2368 * Apply an advisory lock on a file descriptor.
2369 *
2370 * Just attempt to get a record lock of the requested type on
2371 * the entire file (l_whence = SEEK_SET, l_start = 0, l_len = 0).
2372 */
2373 int
2374 flock(struct proc *p, register struct flock_args *uap, __unused register_t *retval)
2375 {
2376 int fd = uap->fd;
2377 int how = uap->how;
2378 struct fileproc *fp;
2379 struct vnode *vp;
2380 struct flock lf;
2381 struct vfs_context context;
2382 int error=0;
2383
2384 AUDIT_ARG(fd, uap->fd);
2385 if ( (error = fp_getfvp(p, fd, &fp, &vp)) ) {
2386 return(error);
2387 }
2388 if ( (error = vnode_getwithref(vp)) ) {
2389 goto out1;
2390 }
2391 AUDIT_ARG(vnpath, vp, ARG_VNODE1);
2392
2393 context.vc_proc = p;
2394 context.vc_ucred = fp->f_cred;
2395
2396 lf.l_whence = SEEK_SET;
2397 lf.l_start = 0;
2398 lf.l_len = 0;
2399 if (how & LOCK_UN) {
2400 lf.l_type = F_UNLCK;
2401 fp->f_flag &= ~FHASLOCK;
2402 error = VNOP_ADVLOCK(vp, (caddr_t)fp->f_fglob, F_UNLCK, &lf, F_FLOCK, &context);
2403 goto out;
2404 }
2405 if (how & LOCK_EX)
2406 lf.l_type = F_WRLCK;
2407 else if (how & LOCK_SH)
2408 lf.l_type = F_RDLCK;
2409 else {
2410 error = EBADF;
2411 goto out;
2412 }
2413 fp->f_flag |= FHASLOCK;
2414 if (how & LOCK_NB) {
2415 error = VNOP_ADVLOCK(vp, (caddr_t)fp->f_fglob, F_SETLK, &lf, F_FLOCK, &context);
2416 goto out;
2417 }
2418 error = VNOP_ADVLOCK(vp, (caddr_t)fp->f_fglob, F_SETLK, &lf, F_FLOCK|F_WAIT, &context);
2419 out:
2420 (void)vnode_put(vp);
2421 out1:
2422 fp_drop(p, fd, fp, 0);
2423 return(error);
2424
2425 }
2426
2427 /*
2428 * File Descriptor pseudo-device driver (/dev/fd/).
2429 *
2430 * Opening minor device N dup()s the file (if any) connected to file
2431 * descriptor N belonging to the calling process. Note that this driver
2432 * consists of only the ``open()'' routine, because all subsequent
2433 * references to this file will be direct to the other driver.
2434 */
2435 int
2436 fdopen(dev_t dev, __unused int mode, __unused int type, struct proc *p)
2437 {
2438
2439 /*
2440 * XXX Kludge: set curproc->p_dupfd to contain the value of the
2441 * the file descriptor being sought for duplication. The error
2442 * return ensures that the vnode for this device will be released
2443 * by vn_open. Open will detect this special error and take the
2444 * actions in dupfdopen below. Other callers of vn_open or vnop_open
2445 * will simply report the error.
2446 */
2447 p->p_dupfd = minor(dev);
2448 return (ENODEV);
2449 }
2450
2451 /*
2452 * Duplicate the specified descriptor to a free descriptor.
2453 */
2454 int
2455 dupfdopen(fdp, indx, dfd, mode, error)
2456 register struct filedesc *fdp;
2457 register int indx, dfd;
2458 int mode;
2459 int error;
2460 {
2461 struct fileproc *wfp;
2462 struct fileproc *fp;
2463 struct proc * p = current_proc();
2464
2465 /*
2466 * If the to-be-dup'd fd number is greater than the allowed number
2467 * of file descriptors, or the fd to be dup'd has already been
2468 * closed, reject. Note, check for new == old is necessary as
2469 * falloc could allocate an already closed to-be-dup'd descriptor
2470 * as the new descriptor.
2471 */
2472 proc_fdlock(p);
2473
2474 fp = fdp->fd_ofiles[indx];
2475 if (dfd < 0 || dfd >= fdp->fd_nfiles ||
2476 (wfp = fdp->fd_ofiles[dfd]) == NULL || wfp == fp ||
2477 (fdp->fd_ofileflags[dfd] & UF_RESERVED)) {
2478
2479 proc_fdunlock(p);
2480 return (EBADF);
2481 }
2482 /*
2483 * There are two cases of interest here.
2484 *
2485 * For ENODEV simply dup (dfd) to file descriptor
2486 * (indx) and return.
2487 *
2488 * For ENXIO steal away the file structure from (dfd) and
2489 * store it in (indx). (dfd) is effectively closed by
2490 * this operation.
2491 *
2492 * Any other error code is just returned.
2493 */
2494 switch (error) {
2495 case ENODEV:
2496 /*
2497 * Check that the mode the file is being opened for is a
2498 * subset of the mode of the existing descriptor.
2499 */
2500 if (((mode & (FREAD|FWRITE)) | wfp->f_flag) != wfp->f_flag) {
2501 proc_fdunlock(p);
2502 return (EACCES);
2503 }
2504 if (indx > fdp->fd_lastfile)
2505 fdp->fd_lastfile = indx;
2506 (void)fg_ref(wfp);
2507
2508 if (fp->f_fglob)
2509 fg_free(fp->f_fglob);
2510 fp->f_fglob = wfp->f_fglob;
2511
2512 fdp->fd_ofileflags[indx] = fdp->fd_ofileflags[dfd];
2513
2514 proc_fdunlock(p);
2515 return (0);
2516
2517 case ENXIO:
2518 /*
2519 * Steal away the file pointer from dfd, and stuff it into indx.
2520 */
2521 if (indx > fdp->fd_lastfile)
2522 fdp->fd_lastfile = indx;
2523
2524 if (fp->f_fglob)
2525 fg_free(fp->f_fglob);
2526 fp->f_fglob = wfp->f_fglob;
2527
2528 fdp->fd_ofileflags[indx] = fdp->fd_ofileflags[dfd];
2529 _fdrelse(fdp, dfd);
2530
2531 proc_fdunlock(p);
2532
2533 FREE_ZONE(wfp, sizeof *fp, M_FILEPROC);
2534
2535 return (0);
2536
2537 default:
2538 proc_fdunlock(p);
2539 return (error);
2540 }
2541 /* NOTREACHED */
2542 }
2543
2544 void
2545 fg_ref(struct fileproc * fp)
2546 {
2547 struct fileglob *fg;
2548
2549 fg = fp->f_fglob;
2550
2551 lck_mtx_lock(&fg->fg_lock);
2552 fg->fg_count++;
2553 lck_mtx_unlock(&fg->fg_lock);
2554 }
2555
2556 void
2557 fg_drop(struct fileproc * fp)
2558 {
2559 struct fileglob *fg;
2560
2561 fg = fp->f_fglob;
2562 lck_mtx_lock(&fg->fg_lock);
2563 fg->fg_count--;
2564 lck_mtx_unlock(&fg->fg_lock);
2565 }
2566
2567
2568 void
2569 fg_insertuipc(struct fileglob * fg)
2570 {
2571 int insertque = 0;
2572
2573 lck_mtx_lock(&fg->fg_lock);
2574
2575 while (fg->fg_lflags & FG_RMMSGQ) {
2576 fg->fg_lflags |= FG_WRMMSGQ;
2577 msleep(&fg->fg_lflags, &fg->fg_lock, 0, "fg_insertuipc", 0);
2578 }
2579
2580 fg->fg_count++;
2581 fg->fg_msgcount++;
2582 if (fg->fg_msgcount == 1) {
2583 fg->fg_lflags |= FG_INSMSGQ;
2584 insertque=1;
2585 }
2586 lck_mtx_unlock(&fg->fg_lock);
2587
2588 if (insertque) {
2589 lck_mtx_lock(uipc_lock);
2590 LIST_INSERT_HEAD(&fmsghead, fg, f_msglist);
2591 lck_mtx_unlock(uipc_lock);
2592 lck_mtx_lock(&fg->fg_lock);
2593 fg->fg_lflags &= ~FG_INSMSGQ;
2594 if (fg->fg_lflags & FG_WINSMSGQ) {
2595 fg->fg_lflags &= ~FG_WINSMSGQ;
2596 wakeup(&fg->fg_lflags);
2597 }
2598 lck_mtx_unlock(&fg->fg_lock);
2599 }
2600
2601 }
2602
2603 void
2604 fg_removeuipc(struct fileglob * fg)
2605 {
2606 int removeque = 0;
2607
2608 lck_mtx_lock(&fg->fg_lock);
2609 while (fg->fg_lflags & FG_INSMSGQ) {
2610 fg->fg_lflags |= FG_WINSMSGQ;
2611 msleep(&fg->fg_lflags, &fg->fg_lock, 0, "fg_removeuipc", 0);
2612 }
2613 fg->fg_msgcount--;
2614 if (fg->fg_msgcount == 0) {
2615 fg->fg_lflags |= FG_RMMSGQ;
2616 removeque=1;
2617 }
2618 lck_mtx_unlock(&fg->fg_lock);
2619
2620 if (removeque) {
2621 lck_mtx_lock(uipc_lock);
2622 LIST_REMOVE(fg, f_msglist);
2623 lck_mtx_unlock(uipc_lock);
2624 lck_mtx_lock(&fg->fg_lock);
2625 fg->fg_lflags &= ~FG_RMMSGQ;
2626 if (fg->fg_lflags & FG_WRMMSGQ) {
2627 fg->fg_lflags &= ~FG_WRMMSGQ;
2628 wakeup(&fg->fg_lflags);
2629 }
2630 lck_mtx_unlock(&fg->fg_lock);
2631 }
2632 }
2633
2634
2635 int
2636 fo_read(struct fileproc *fp, struct uio *uio, kauth_cred_t cred, int flags, struct proc *p)
2637 {
2638 return ((*fp->f_ops->fo_read)(fp, uio, cred, flags, p));
2639 }
2640
2641 int
2642 fo_write(struct fileproc *fp, struct uio *uio, kauth_cred_t cred, int flags, struct proc *p)
2643 {
2644 return((*fp->f_ops->fo_write)(fp, uio, cred, flags, p));
2645 }
2646
2647 int
2648 fo_ioctl(struct fileproc *fp, u_long com, caddr_t data, struct proc *p)
2649 {
2650 int error;
2651
2652 proc_fdunlock(p);
2653 error = (*fp->f_ops->fo_ioctl)(fp, com, data, p);
2654 proc_fdlock(p);
2655 return(error);
2656 }
2657
2658 int
2659 fo_select(struct fileproc *fp, int which, void *wql, struct proc *p)
2660 {
2661 return((*fp->f_ops->fo_select)(fp, which, wql, p));
2662 }
2663
2664 int
2665 fo_close(struct fileglob *fg, struct proc *p)
2666 {
2667 return((*fg->fg_ops->fo_close)(fg, p));
2668 }
2669
2670 int
2671 fo_kqfilter(struct fileproc *fp, struct knote *kn, struct proc *p)
2672 {
2673 return ((*fp->f_ops->fo_kqfilter)(fp, kn, p));
2674 }
2675
mirror of XNU