]> 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-2050.22.13.tar.gz
[apple/xnu.git] / bsd / kern / kern_descrip.c
1 /*
2 * Copyright (c) 2000-2010 Apple Inc. All rights reserved.
3 *
4 * @APPLE_OSREFERENCE_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. The rights granted to you under the License
10 * may not be used to create, or enable the creation or redistribution of,
11 * unlawful or unlicensed copies of an Apple operating system, or to
12 * circumvent, violate, or enable the circumvention or violation of, any
13 * terms of an Apple operating system software license agreement.
14 *
15 * Please obtain a copy of the License at
16 * http://www.opensource.apple.com/apsl/ and read it before using this file.
17 *
18 * The Original Code and all software distributed under the License are
19 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
20 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
21 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
22 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
23 * Please see the License for the specific language governing rights and
24 * limitations under the License.
25 *
26 * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
27 */
28 /* Copyright (c) 1995, 1997 Apple Computer, Inc. All Rights Reserved */
29 /*
30 * Copyright (c) 1982, 1986, 1989, 1991, 1993
31 * The Regents of the University of California. All rights reserved.
32 * (c) UNIX System Laboratories, Inc.
33 * All or some portions of this file are derived from material licensed
34 * to the University of California by American Telephone and Telegraph
35 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
36 * the permission of UNIX System Laboratories, Inc.
37 *
38 * Redistribution and use in source and binary forms, with or without
39 * modification, are permitted provided that the following conditions
40 * are met:
41 * 1. Redistributions of source code must retain the above copyright
42 * notice, this list of conditions and the following disclaimer.
43 * 2. Redistributions in binary form must reproduce the above copyright
44 * notice, this list of conditions and the following disclaimer in the
45 * documentation and/or other materials provided with the distribution.
46 * 3. All advertising materials mentioning features or use of this software
47 * must display the following acknowledgement:
48 * This product includes software developed by the University of
49 * California, Berkeley and its contributors.
50 * 4. Neither the name of the University nor the names of its contributors
51 * may be used to endorse or promote products derived from this software
52 * without specific prior written permission.
53 *
54 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
64 * SUCH DAMAGE.
65 *
66 * @(#)kern_descrip.c 8.8 (Berkeley) 2/14/95
67 */
68 /*
69 * NOTICE: This file was modified by SPARTA, Inc. in 2006 to introduce
70 * support for mandatory and extensible security protections. This notice
71 * is included in support of clause 2.2 (b) of the Apple Public License,
72 * Version 2.0.
73 */
74
75 #include <sys/param.h>
76 #include <sys/systm.h>
77 #include <sys/filedesc.h>
78 #include <sys/kernel.h>
79 #include <sys/vnode_internal.h>
80 #include <sys/proc_internal.h>
81 #include <sys/kauth.h>
82 #include <sys/file_internal.h>
83 #include <sys/socket.h>
84 #include <sys/socketvar.h>
85 #include <sys/stat.h>
86 #include <sys/ioctl.h>
87 #include <sys/fcntl.h>
88 #include <sys/malloc.h>
89 #include <sys/mman.h>
90 #include <sys/syslog.h>
91 #include <sys/unistd.h>
92 #include <sys/resourcevar.h>
93 #include <sys/aio_kern.h>
94 #include <sys/ev.h>
95 #include <kern/lock.h>
96 #include <sys/uio_internal.h>
97
98 #include <security/audit/audit.h>
99
100 #include <sys/mount_internal.h>
101 #include <sys/kdebug.h>
102 #include <sys/sysproto.h>
103 #include <sys/pipe.h>
104 #include <sys/spawn.h>
105 #include <kern/kern_types.h>
106 #include <kern/kalloc.h>
107 #include <libkern/OSAtomic.h>
108
109 #include <sys/ubc_internal.h>
110
111 #include <kern/ipc_misc.h>
112 #include <vm/vm_protos.h>
113
114 #include <mach/mach_port.h>
115
116 #if CONFIG_PROTECT
117 #include <sys/cprotect.h>
118 #endif
119 #include <hfs/hfs.h>
120
121 kern_return_t ipc_object_copyin(ipc_space_t, mach_port_name_t,
122 mach_msg_type_name_t, ipc_port_t *);
123 void ipc_port_release_send(ipc_port_t);
124
125 struct psemnode;
126 struct pshmnode;
127
128 static int finishdup(proc_t p,
129 struct filedesc *fdp, int old, int new, int flags, int32_t *retval);
130
131 int falloc_locked(proc_t p, struct fileproc **resultfp, int *resultfd, vfs_context_t ctx, int locked);
132 void fg_drop(struct fileproc * fp);
133 void fg_free(struct fileglob *fg);
134 void fg_ref(struct fileproc * fp);
135 void fileport_releasefg(struct fileglob *fg);
136
137 /* flags for close_internal_locked */
138 #define FD_DUP2RESV 1
139 static int close_internal_locked(struct proc *p, int fd, struct fileproc *fp, int flags);
140
141 static int closef_finish(struct fileproc *fp, struct fileglob *fg, proc_t p, vfs_context_t ctx);
142
143 /* We don't want these exported */
144 __private_extern__
145 int open1(vfs_context_t, struct nameidata *, int, struct vnode_attr *, int32_t *);
146
147 __private_extern__
148 int unlink1(vfs_context_t, struct nameidata *, int);
149
150 static void _fdrelse(struct proc * p, int fd);
151
152
153 extern void file_lock_init(void) __attribute__((section("__TEXT, initcode")));
154 extern int kqueue_stat(struct fileproc *fp, void *ub, int isstat4, proc_t p);
155 #if SOCKETS
156 extern int soo_stat(struct socket *so, void *ub, int isstat64);
157 #endif /* SOCKETS */
158
159 extern kauth_scope_t kauth_scope_fileop;
160
161 extern int cs_debug;
162
163 /* Conflict wait queue for when selects collide (opaque type) */
164 extern struct wait_queue select_conflict_queue;
165
166 #define f_flag f_fglob->fg_flag
167 #define f_type f_fglob->fg_type
168 #define f_msgcount f_fglob->fg_msgcount
169 #define f_cred f_fglob->fg_cred
170 #define f_ops f_fglob->fg_ops
171 #define f_offset f_fglob->fg_offset
172 #define f_data f_fglob->fg_data
173 /*
174 * Descriptor management.
175 */
176 struct fmsglist fmsghead; /* head of list of open files */
177 struct fmsglist fmsg_ithead; /* head of list of open files */
178 int nfiles; /* actual number of open files */
179
180
181 lck_grp_attr_t * file_lck_grp_attr;
182 lck_grp_t * file_lck_grp;
183 lck_attr_t * file_lck_attr;
184
185 lck_mtx_t * uipc_lock;
186
187
188 /*
189 * file_lock_init
190 *
191 * Description: Initialize the file lock group and the uipc and flist locks
192 *
193 * Parameters: (void)
194 *
195 * Returns: void
196 *
197 * Notes: Called at system startup from bsd_init().
198 */
199 void
200 file_lock_init(void)
201 {
202 /* allocate file lock group attribute and group */
203 file_lck_grp_attr= lck_grp_attr_alloc_init();
204
205 file_lck_grp = lck_grp_alloc_init("file", file_lck_grp_attr);
206
207 /* Allocate file lock attribute */
208 file_lck_attr = lck_attr_alloc_init();
209
210 uipc_lock = lck_mtx_alloc_init(file_lck_grp, file_lck_attr);
211 }
212
213
214 /*
215 * proc_fdlock, proc_fdlock_spin
216 *
217 * Description: Lock to control access to the per process struct fileproc
218 * and struct filedesc
219 *
220 * Parameters: p Process to take the lock on
221 *
222 * Returns: void
223 *
224 * Notes: The lock is initialized in forkproc() and destroyed in
225 * reap_child_process().
226 */
227 void
228 proc_fdlock(proc_t p)
229 {
230 lck_mtx_lock(&p->p_fdmlock);
231 }
232
233 void
234 proc_fdlock_spin(proc_t p)
235 {
236 lck_mtx_lock_spin(&p->p_fdmlock);
237 }
238
239 void
240 proc_fdlock_assert(proc_t p, int assertflags)
241 {
242 lck_mtx_assert(&p->p_fdmlock, assertflags);
243 }
244
245
246 /*
247 * proc_fdunlock
248 *
249 * Description: Unlock the lock previously locked by a call to proc_fdlock()
250 *
251 * Parameters: p Process to drop the lock on
252 *
253 * Returns: void
254 */
255 void
256 proc_fdunlock(proc_t p)
257 {
258 lck_mtx_unlock(&p->p_fdmlock);
259 }
260
261
262 /*
263 * System calls on descriptors.
264 */
265
266
267 /*
268 * getdtablesize
269 *
270 * Description: Returns the per process maximum size of the descriptor table
271 *
272 * Parameters: p Process being queried
273 * retval Pointer to the call return area
274 *
275 * Returns: 0 Success
276 *
277 * Implicit returns:
278 * *retval (modified) Size of dtable
279 */
280 int
281 getdtablesize(proc_t p, __unused struct getdtablesize_args *uap, int32_t *retval)
282 {
283 proc_fdlock_spin(p);
284 *retval = min((int)p->p_rlimit[RLIMIT_NOFILE].rlim_cur, maxfiles);
285 proc_fdunlock(p);
286
287 return (0);
288 }
289
290
291 void
292 procfdtbl_reservefd(struct proc * p, int fd)
293 {
294 p->p_fd->fd_ofiles[fd] = NULL;
295 p->p_fd->fd_ofileflags[fd] |= UF_RESERVED;
296 }
297
298 void
299 procfdtbl_markclosefd(struct proc * p, int fd)
300 {
301 p->p_fd->fd_ofileflags[fd] |= (UF_RESERVED | UF_CLOSING);
302 }
303
304 void
305 procfdtbl_releasefd(struct proc * p, int fd, struct fileproc * fp)
306 {
307 if (fp != NULL)
308 p->p_fd->fd_ofiles[fd] = fp;
309 p->p_fd->fd_ofileflags[fd] &= ~UF_RESERVED;
310 if ((p->p_fd->fd_ofileflags[fd] & UF_RESVWAIT) == UF_RESVWAIT) {
311 p->p_fd->fd_ofileflags[fd] &= ~UF_RESVWAIT;
312 wakeup(&p->p_fd);
313 }
314 }
315
316 void
317 procfdtbl_waitfd(struct proc * p, int fd)
318 {
319 p->p_fd->fd_ofileflags[fd] |= UF_RESVWAIT;
320 msleep(&p->p_fd, &p->p_fdmlock, PRIBIO, "ftbl_waitfd", NULL);
321 }
322
323
324 void
325 procfdtbl_clearfd(struct proc * p, int fd)
326 {
327 int waiting;
328
329 waiting = (p->p_fd->fd_ofileflags[fd] & UF_RESVWAIT);
330 p->p_fd->fd_ofiles[fd] = NULL;
331 p->p_fd->fd_ofileflags[fd] = 0;
332 if ( waiting == UF_RESVWAIT) {
333 wakeup(&p->p_fd);
334 }
335 }
336
337 /*
338 * _fdrelse
339 *
340 * Description: Inline utility function to free an fd in a filedesc
341 *
342 * Parameters: fdp Pointer to filedesc fd lies in
343 * fd fd to free
344 * reserv fd should be reserved
345 *
346 * Returns: void
347 *
348 * Locks: Assumes proc_fdlock for process pointing to fdp is held by
349 * the caller
350 */
351 static void
352 _fdrelse(struct proc * p, int fd)
353 {
354 struct filedesc *fdp = p->p_fd;
355 int nfd = 0;
356
357 if (fd < fdp->fd_freefile)
358 fdp->fd_freefile = fd;
359 #if DIAGNOSTIC
360 if (fd > fdp->fd_lastfile)
361 panic("fdrelse: fd_lastfile inconsistent");
362 #endif
363 procfdtbl_clearfd(p, fd);
364
365 while ((nfd = fdp->fd_lastfile) > 0 &&
366 fdp->fd_ofiles[nfd] == NULL &&
367 !(fdp->fd_ofileflags[nfd] & UF_RESERVED))
368 fdp->fd_lastfile--;
369 }
370
371
372 int
373 fd_rdwr(
374 int fd,
375 enum uio_rw rw,
376 uint64_t base,
377 int64_t len,
378 enum uio_seg segflg,
379 off_t offset,
380 int io_flg,
381 int64_t *aresid)
382 {
383 struct fileproc *fp;
384 proc_t p;
385 int error = 0;
386 int flags = 0;
387 int spacetype;
388 uio_t auio = NULL;
389 char uio_buf[ UIO_SIZEOF(1) ];
390 struct vfs_context context = *(vfs_context_current());
391
392 p = current_proc();
393
394 error = fp_lookup(p, fd, &fp, 0);
395 if (error)
396 return(error);
397
398 if (fp->f_type != DTYPE_VNODE && fp->f_type != DTYPE_PIPE && fp->f_type != DTYPE_SOCKET) {
399 error = EINVAL;
400 goto out;
401 }
402 if (rw == UIO_WRITE && !(fp->f_flag & FWRITE)) {
403 error = EBADF;
404 goto out;
405 }
406
407 if (rw == UIO_READ && !(fp->f_flag & FREAD)) {
408 error = EBADF;
409 goto out;
410 }
411
412 context.vc_ucred = fp->f_fglob->fg_cred;
413
414 if (UIO_SEG_IS_USER_SPACE(segflg))
415 spacetype = proc_is64bit(p) ? UIO_USERSPACE64 : UIO_USERSPACE32;
416 else
417 spacetype = UIO_SYSSPACE;
418
419 auio = uio_createwithbuffer(1, offset, spacetype, rw, &uio_buf[0], sizeof(uio_buf));
420
421 uio_addiov(auio, base, len);
422
423 if ( !(io_flg & IO_APPEND))
424 flags = FOF_OFFSET;
425
426 if (rw == UIO_WRITE)
427 error = fo_write(fp, auio, flags, &context);
428 else
429 error = fo_read(fp, auio, flags, &context);
430
431 if (aresid)
432 *aresid = uio_resid(auio);
433 else {
434 if (uio_resid(auio) && error == 0)
435 error = EIO;
436 }
437 out:
438 if (rw == UIO_WRITE && error == 0)
439 fp_drop_written(p, fd, fp);
440 else
441 fp_drop(p, fd, fp, 0);
442
443 return error;
444 }
445
446
447
448 /*
449 * dup
450 *
451 * Description: Duplicate a file descriptor.
452 *
453 * Parameters: p Process performing the dup
454 * uap->fd The fd to dup
455 * retval Pointer to the call return area
456 *
457 * Returns: 0 Success
458 * !0 Errno
459 *
460 * Implicit returns:
461 * *retval (modified) The new descriptor
462 */
463 int
464 dup(proc_t p, struct dup_args *uap, int32_t *retval)
465 {
466 struct filedesc *fdp = p->p_fd;
467 int old = uap->fd;
468 int new, error;
469 struct fileproc *fp;
470
471 proc_fdlock(p);
472 if ( (error = fp_lookup(p, old, &fp, 1)) ) {
473 proc_fdunlock(p);
474 return(error);
475 }
476 if ( (error = fdalloc(p, 0, &new)) ) {
477 fp_drop(p, old, fp, 1);
478 proc_fdunlock(p);
479 return (error);
480 }
481 error = finishdup(p, fdp, old, new, 0, retval);
482 fp_drop(p, old, fp, 1);
483 proc_fdunlock(p);
484
485 return (error);
486 }
487
488 /*
489 * dup2
490 *
491 * Description: Duplicate a file descriptor to a particular value.
492 *
493 * Parameters: p Process performing the dup
494 * uap->from The fd to dup
495 * uap->to The fd to dup it to
496 * retval Pointer to the call return area
497 *
498 * Returns: 0 Success
499 * !0 Errno
500 *
501 * Implicit returns:
502 * *retval (modified) The new descriptor
503 */
504 int
505 dup2(proc_t p, struct dup2_args *uap, int32_t *retval)
506 {
507 struct filedesc *fdp = p->p_fd;
508 int old = uap->from, new = uap->to;
509 int i, error;
510 struct fileproc *fp, *nfp;
511
512 proc_fdlock(p);
513
514 startover:
515 if ( (error = fp_lookup(p, old, &fp, 1)) ) {
516 proc_fdunlock(p);
517 return(error);
518 }
519 if (new < 0 ||
520 (rlim_t)new >= p->p_rlimit[RLIMIT_NOFILE].rlim_cur ||
521 new >= maxfiles) {
522 fp_drop(p, old, fp, 1);
523 proc_fdunlock(p);
524 return (EBADF);
525 }
526 if (old == new) {
527 fp_drop(p, old, fp, 1);
528 *retval = new;
529 proc_fdunlock(p);
530 return (0);
531 }
532 if (new < 0 || new >= fdp->fd_nfiles) {
533 if ( (error = fdalloc(p, new, &i)) ) {
534 fp_drop(p, old, fp, 1);
535 proc_fdunlock(p);
536 return (error);
537 }
538 if (new != i) {
539 fdrelse(p, i);
540 goto closeit;
541 }
542 } else {
543 closeit:
544 while ((fdp->fd_ofileflags[new] & UF_RESERVED) == UF_RESERVED) {
545 fp_drop(p, old, fp, 1);
546 procfdtbl_waitfd(p, new);
547 #if DIAGNOSTIC
548 proc_fdlock_assert(p, LCK_MTX_ASSERT_OWNED);
549 #endif
550 goto startover;
551 }
552
553 if ((fdp->fd_ofiles[new] != NULL) &&
554 ((error = fp_lookup(p, new, &nfp, 1)) == 0)) {
555 fp_drop(p, old, fp, 1);
556 (void)close_internal_locked(p, new, nfp, FD_DUP2RESV);
557 #if DIAGNOSTIC
558 proc_fdlock_assert(p, LCK_MTX_ASSERT_OWNED);
559 #endif
560 procfdtbl_clearfd(p, new);
561 goto startover;
562 } else {
563 #if DIAGNOSTIC
564 if (fdp->fd_ofiles[new] != NULL)
565 panic("dup2: no ref on fileproc %d", new);
566 #endif
567 procfdtbl_reservefd(p, new);
568 }
569
570 #if DIAGNOSTIC
571 proc_fdlock_assert(p, LCK_MTX_ASSERT_OWNED);
572 #endif
573
574 }
575 #if DIAGNOSTIC
576 if (fdp->fd_ofiles[new] != 0)
577 panic("dup2: overwriting fd_ofiles with new %d", new);
578 if ((fdp->fd_ofileflags[new] & UF_RESERVED) == 0)
579 panic("dup2: unreserved fileflags with new %d", new);
580 #endif
581 error = finishdup(p, fdp, old, new, 0, retval);
582 fp_drop(p, old, fp, 1);
583 proc_fdunlock(p);
584
585 return(error);
586 }
587
588
589 /*
590 * fcntl
591 *
592 * Description: The file control system call.
593 *
594 * Parameters: p Process performing the fcntl
595 * uap->fd The fd to operate against
596 * uap->cmd The command to perform
597 * uap->arg Pointer to the command argument
598 * retval Pointer to the call return area
599 *
600 * Returns: 0 Success
601 * !0 Errno (see fcntl_nocancel)
602 *
603 * Implicit returns:
604 * *retval (modified) fcntl return value (if any)
605 *
606 * Notes: This system call differs from fcntl_nocancel() in that it
607 * tests for cancellation prior to performing a potentially
608 * blocking operation.
609 */
610 int
611 fcntl(proc_t p, struct fcntl_args *uap, int32_t *retval)
612 {
613 __pthread_testcancel(1);
614 return(fcntl_nocancel(p, (struct fcntl_nocancel_args *)uap, retval));
615 }
616
617
618 /*
619 * fcntl_nocancel
620 *
621 * Description: A non-cancel-testing file control system call.
622 *
623 * Parameters: p Process performing the fcntl
624 * uap->fd The fd to operate against
625 * uap->cmd The command to perform
626 * uap->arg Pointer to the command argument
627 * retval Pointer to the call return area
628 *
629 * Returns: 0 Success
630 * EINVAL
631 * fp_lookup:EBADF Bad file descriptor
632 * [F_DUPFD]
633 * fdalloc:EMFILE
634 * fdalloc:ENOMEM
635 * finishdup:EBADF
636 * finishdup:ENOMEM
637 * [F_SETOWN]
638 * ESRCH
639 * [F_SETLK]
640 * EBADF
641 * EOVERFLOW
642 * copyin:EFAULT
643 * vnode_getwithref:???
644 * VNOP_ADVLOCK:???
645 * [F_GETLK]
646 * EBADF
647 * EOVERFLOW
648 * copyin:EFAULT
649 * copyout:EFAULT
650 * vnode_getwithref:???
651 * VNOP_ADVLOCK:???
652 * [F_PREALLOCATE]
653 * EBADF
654 * EINVAL
655 * copyin:EFAULT
656 * copyout:EFAULT
657 * vnode_getwithref:???
658 * VNOP_ALLOCATE:???
659 * [F_SETSIZE,F_RDADVISE]
660 * EBADF
661 * copyin:EFAULT
662 * vnode_getwithref:???
663 * [F_RDAHEAD,F_NOCACHE]
664 * EBADF
665 * vnode_getwithref:???
666 * [???]
667 *
668 * Implicit returns:
669 * *retval (modified) fcntl return value (if any)
670 */
671 int
672 fcntl_nocancel(proc_t p, struct fcntl_nocancel_args *uap, int32_t *retval)
673 {
674 int fd = uap->fd;
675 struct filedesc *fdp = p->p_fd;
676 struct fileproc *fp;
677 char *pop;
678 struct vnode *vp = NULLVP; /* for AUDIT_ARG() at end */
679 int i, tmp, error, error2, flg = F_POSIX;
680 struct flock fl;
681 struct vfs_context context;
682 off_t offset;
683 int newmin;
684 daddr64_t lbn, bn;
685 unsigned int fflag;
686 user_addr_t argp;
687 boolean_t is64bit;
688
689 AUDIT_ARG(fd, uap->fd);
690 AUDIT_ARG(cmd, uap->cmd);
691
692 proc_fdlock(p);
693 if ( (error = fp_lookup(p, fd, &fp, 1)) ) {
694 proc_fdunlock(p);
695 return(error);
696 }
697 context.vc_thread = current_thread();
698 context.vc_ucred = fp->f_cred;
699
700 is64bit = proc_is64bit(p);
701 if (is64bit) {
702 argp = uap->arg;
703 }
704 else {
705 /*
706 * Since the arg parameter is defined as a long but may be
707 * either a long or a pointer we must take care to handle
708 * sign extension issues. Our sys call munger will sign
709 * extend a long when we are called from a 32-bit process.
710 * Since we can never have an address greater than 32-bits
711 * from a 32-bit process we lop off the top 32-bits to avoid
712 * getting the wrong address
713 */
714 argp = CAST_USER_ADDR_T((uint32_t)uap->arg);
715 }
716
717 pop = &fdp->fd_ofileflags[fd];
718
719 #if CONFIG_MACF
720 error = mac_file_check_fcntl(proc_ucred(p), fp->f_fglob, uap->cmd,
721 uap->arg);
722 if (error)
723 goto out;
724 #endif
725
726 switch (uap->cmd) {
727
728 case F_DUPFD:
729 case F_DUPFD_CLOEXEC:
730 newmin = CAST_DOWN_EXPLICIT(int, uap->arg); /* arg is an int, so we won't lose bits */
731 AUDIT_ARG(value32, newmin);
732 if ((u_int)newmin >= p->p_rlimit[RLIMIT_NOFILE].rlim_cur ||
733 newmin >= maxfiles) {
734 error = EINVAL;
735 goto out;
736 }
737 if ( (error = fdalloc(p, newmin, &i)) )
738 goto out;
739 error = finishdup(p, fdp, fd, i,
740 uap->cmd == F_DUPFD_CLOEXEC ? UF_EXCLOSE : 0, retval);
741 goto out;
742
743 case F_GETFD:
744 *retval = (*pop & UF_EXCLOSE)? 1 : 0;
745 error = 0;
746 goto out;
747
748 case F_SETFD:
749 AUDIT_ARG(value32, uap->arg);
750 *pop = (*pop &~ UF_EXCLOSE) |
751 (uap->arg & 1)? UF_EXCLOSE : 0;
752 error = 0;
753 goto out;
754
755 case F_GETFL:
756 *retval = OFLAGS(fp->f_flag);
757 error = 0;
758 goto out;
759
760 case F_SETFL:
761 fp->f_flag &= ~FCNTLFLAGS;
762 tmp = CAST_DOWN_EXPLICIT(int, uap->arg); /* arg is an int, so we won't lose bits */
763 AUDIT_ARG(value32, tmp);
764 fp->f_flag |= FFLAGS(tmp) & FCNTLFLAGS;
765 tmp = fp->f_flag & FNONBLOCK;
766 error = fo_ioctl(fp, FIONBIO, (caddr_t)&tmp, &context);
767 if (error)
768 goto out;
769 tmp = fp->f_flag & FASYNC;
770 error = fo_ioctl(fp, FIOASYNC, (caddr_t)&tmp, &context);
771 if (!error)
772 goto out;
773 fp->f_flag &= ~FNONBLOCK;
774 tmp = 0;
775 (void)fo_ioctl(fp, FIONBIO, (caddr_t)&tmp, &context);
776 goto out;
777
778 case F_GETOWN:
779 if (fp->f_type == DTYPE_SOCKET) {
780 *retval = ((struct socket *)fp->f_data)->so_pgid;
781 error = 0;
782 goto out;
783 }
784 error = fo_ioctl(fp, (int)TIOCGPGRP, (caddr_t)retval, &context);
785 *retval = -*retval;
786 goto out;
787
788 case F_SETOWN:
789 tmp = CAST_DOWN_EXPLICIT(pid_t, uap->arg); /* arg is an int, so we won't lose bits */
790 AUDIT_ARG(value32, tmp);
791 if (fp->f_type == DTYPE_SOCKET) {
792 ((struct socket *)fp->f_data)->so_pgid = tmp;
793 error =0;
794 goto out;
795 }
796 if (fp->f_type == DTYPE_PIPE) {
797 error = fo_ioctl(fp, TIOCSPGRP, (caddr_t)&tmp, &context);
798 goto out;
799 }
800
801 if (tmp <= 0) {
802 tmp = -tmp;
803 } else {
804 proc_t p1 = proc_find(tmp);
805 if (p1 == 0) {
806 error = ESRCH;
807 goto out;
808 }
809 tmp = (int)p1->p_pgrpid;
810 proc_rele(p1);
811 }
812 error = fo_ioctl(fp, (int)TIOCSPGRP, (caddr_t)&tmp, &context);
813 goto out;
814
815 case F_SETNOSIGPIPE:
816 tmp = CAST_DOWN_EXPLICIT(int, uap->arg);
817 if (fp->f_type == DTYPE_SOCKET) {
818 error = sock_setsockopt((struct socket *)fp->f_data,
819 SOL_SOCKET, SO_NOSIGPIPE, &tmp, sizeof (tmp));
820 } else {
821 struct fileglob *fg = fp->f_fglob;
822
823 lck_mtx_lock_spin(&fg->fg_lock);
824 if (tmp)
825 fg->fg_lflags |= FG_NOSIGPIPE;
826 else
827 fg->fg_lflags &= FG_NOSIGPIPE;
828 lck_mtx_unlock(&fg->fg_lock);
829 error = 0;
830 }
831 goto out;
832
833 case F_GETNOSIGPIPE:
834 if (fp->f_type == DTYPE_SOCKET) {
835 int retsize = sizeof (*retval);
836 error = sock_getsockopt((struct socket *)fp->f_data,
837 SOL_SOCKET, SO_NOSIGPIPE, retval, &retsize);
838 } else {
839 *retval = (fp->f_fglob->fg_lflags & FG_NOSIGPIPE) ?
840 1 : 0;
841 error = 0;
842 }
843 goto out;
844
845 case F_SETLKW:
846 flg |= F_WAIT;
847 /* Fall into F_SETLK */
848
849 case F_SETLK:
850 if (fp->f_type != DTYPE_VNODE) {
851 error = EBADF;
852 goto out;
853 }
854 vp = (struct vnode *)fp->f_data;
855
856 fflag = fp->f_flag;
857 offset = fp->f_offset;
858 proc_fdunlock(p);
859
860 /* Copy in the lock structure */
861 error = copyin(argp, (caddr_t)&fl, sizeof(fl));
862 if (error) {
863 goto outdrop;
864 }
865
866 volatile off_t affected_lock_area_set = 0;
867 affected_lock_area_set = fl.l_start + offset;
868 if ((fl.l_whence == SEEK_CUR) && (affected_lock_area_set < fl.l_start)) {
869 error = EOVERFLOW;
870 goto outdrop;
871 }
872
873 if ( (error = vnode_getwithref(vp)) ) {
874 goto outdrop;
875 }
876 if (fl.l_whence == SEEK_CUR)
877 fl.l_start += offset;
878
879 #if CONFIG_MACF
880 error = mac_file_check_lock(proc_ucred(p), fp->f_fglob,
881 F_SETLK, &fl);
882 if (error) {
883 (void)vnode_put(vp);
884 goto outdrop;
885 }
886 #endif
887 switch (fl.l_type) {
888
889 case F_RDLCK:
890 if ((fflag & FREAD) == 0) {
891 (void)vnode_put(vp);
892 error = EBADF;
893 goto outdrop;
894 }
895 // XXX UInt32 unsafe for LP64 kernel
896 OSBitOrAtomic(P_LADVLOCK, &p->p_ladvflag);
897 error = VNOP_ADVLOCK(vp, (caddr_t)p, F_SETLK, &fl, flg, &context);
898 (void)vnode_put(vp);
899 goto outdrop;
900
901 case F_WRLCK:
902 if ((fflag & FWRITE) == 0) {
903 (void)vnode_put(vp);
904 error = EBADF;
905 goto outdrop;
906 }
907 // XXX UInt32 unsafe for LP64 kernel
908 OSBitOrAtomic(P_LADVLOCK, &p->p_ladvflag);
909 error = VNOP_ADVLOCK(vp, (caddr_t)p, F_SETLK, &fl, flg, &context);
910 (void)vnode_put(vp);
911 goto outdrop;
912
913 case F_UNLCK:
914 error = VNOP_ADVLOCK(vp, (caddr_t)p, F_UNLCK, &fl,
915 F_POSIX, &context);
916 (void)vnode_put(vp);
917 goto outdrop;
918
919 default:
920 (void)vnode_put(vp);
921 error = EINVAL;
922 goto outdrop;
923 }
924
925 case F_GETLK:
926 #if CONFIG_EMBEDDED
927 case F_GETLKPID:
928 #endif
929 if (fp->f_type != DTYPE_VNODE) {
930 error = EBADF;
931 goto out;
932 }
933 vp = (struct vnode *)fp->f_data;
934
935 offset = fp->f_offset;
936 proc_fdunlock(p);
937
938 /* Copy in the lock structure */
939 error = copyin(argp, (caddr_t)&fl, sizeof(fl));
940 if (error)
941 goto outdrop;
942
943 volatile off_t affected_lock_area_end = 0;
944 affected_lock_area_end = fl.l_start + offset;
945 /* Check starting byte and ending byte for EOVERFLOW in SEEK_CUR */
946 /* and ending byte for EOVERFLOW in SEEK_SET */
947 if (((fl.l_whence == SEEK_CUR) &&
948 ((affected_lock_area_end < fl.l_start) ||
949 ((fl.l_len > 0) && (affected_lock_area_end + fl.l_len - 1 < affected_lock_area_end)))) ||
950 ((fl.l_whence == SEEK_SET) && (fl.l_len > 0) && (fl.l_start + fl.l_len - 1 < fl.l_start)))
951 {
952 /* lf_advlock doesn't check start/end for F_GETLK if file has no locks */
953 error = EOVERFLOW;
954 goto outdrop;
955 }
956
957 if ((fl.l_whence == SEEK_SET) && (fl.l_start < 0)) {
958 error = EINVAL;
959 goto outdrop;
960 }
961
962 switch (fl.l_type) {
963 case F_RDLCK:
964 case F_UNLCK:
965 case F_WRLCK:
966 break;
967 default:
968 error = EINVAL;
969 goto outdrop;
970 }
971
972 switch (fl.l_whence) {
973 case SEEK_CUR:
974 case SEEK_SET:
975 case SEEK_END:
976 break;
977 default:
978 error = EINVAL;
979 goto outdrop;
980 }
981
982 if ( (error = vnode_getwithref(vp)) == 0 ) {
983 if (fl.l_whence == SEEK_CUR)
984 fl.l_start += offset;
985
986 #if CONFIG_MACF
987 error = mac_file_check_lock(proc_ucred(p), fp->f_fglob,
988 uap->cmd, &fl);
989 if (error == 0)
990 #endif
991 error = VNOP_ADVLOCK(vp, (caddr_t)p, uap->cmd, &fl, F_POSIX, &context);
992
993 (void)vnode_put(vp);
994
995 if (error == 0)
996 error = copyout((caddr_t)&fl, argp, sizeof(fl));
997 }
998 goto outdrop;
999
1000 case F_PREALLOCATE: {
1001 fstore_t alloc_struct; /* structure for allocate command */
1002 u_int32_t alloc_flags = 0;
1003
1004 if (fp->f_type != DTYPE_VNODE) {
1005 error = EBADF;
1006 goto out;
1007 }
1008
1009 vp = (struct vnode *)fp->f_data;
1010 proc_fdunlock(p);
1011
1012 /* make sure that we have write permission */
1013 if ((fp->f_flag & FWRITE) == 0) {
1014 error = EBADF;
1015 goto outdrop;
1016 }
1017
1018 error = copyin(argp, (caddr_t)&alloc_struct, sizeof(alloc_struct));
1019 if (error)
1020 goto outdrop;
1021
1022 /* now set the space allocated to 0 */
1023 alloc_struct.fst_bytesalloc = 0;
1024
1025 /*
1026 * Do some simple parameter checking
1027 */
1028
1029 /* set up the flags */
1030
1031 alloc_flags |= PREALLOCATE;
1032
1033 if (alloc_struct.fst_flags & F_ALLOCATECONTIG)
1034 alloc_flags |= ALLOCATECONTIG;
1035
1036 if (alloc_struct.fst_flags & F_ALLOCATEALL)
1037 alloc_flags |= ALLOCATEALL;
1038
1039 /*
1040 * Do any position mode specific stuff. The only
1041 * position mode supported now is PEOFPOSMODE
1042 */
1043
1044 switch (alloc_struct.fst_posmode) {
1045
1046 case F_PEOFPOSMODE:
1047 if (alloc_struct.fst_offset != 0) {
1048 error = EINVAL;
1049 goto outdrop;
1050 }
1051
1052 alloc_flags |= ALLOCATEFROMPEOF;
1053 break;
1054
1055 case F_VOLPOSMODE:
1056 if (alloc_struct.fst_offset <= 0) {
1057 error = EINVAL;
1058 goto outdrop;
1059 }
1060
1061 alloc_flags |= ALLOCATEFROMVOL;
1062 break;
1063
1064 default: {
1065 error = EINVAL;
1066 goto outdrop;
1067 }
1068 }
1069 if ( (error = vnode_getwithref(vp)) == 0 ) {
1070 /*
1071 * call allocate to get the space
1072 */
1073 error = VNOP_ALLOCATE(vp,alloc_struct.fst_length,alloc_flags,
1074 &alloc_struct.fst_bytesalloc, alloc_struct.fst_offset,
1075 &context);
1076 (void)vnode_put(vp);
1077
1078 error2 = copyout((caddr_t)&alloc_struct, argp, sizeof(alloc_struct));
1079
1080 if (error == 0)
1081 error = error2;
1082 }
1083 goto outdrop;
1084
1085 }
1086 case F_SETSIZE:
1087 if (fp->f_type != DTYPE_VNODE) {
1088 error = EBADF;
1089 goto out;
1090 }
1091 vp = (struct vnode *)fp->f_data;
1092 proc_fdunlock(p);
1093
1094 error = copyin(argp, (caddr_t)&offset, sizeof (off_t));
1095 if (error)
1096 goto outdrop;
1097 AUDIT_ARG(value64, offset);
1098
1099 error = vnode_getwithref(vp);
1100 if (error)
1101 goto outdrop;
1102
1103 #if CONFIG_MACF
1104 error = mac_vnode_check_truncate(&context,
1105 fp->f_fglob->fg_cred, vp);
1106 if (error) {
1107 (void)vnode_put(vp);
1108 goto outdrop;
1109 }
1110 #endif
1111 /*
1112 * Make sure that we are root. Growing a file
1113 * without zero filling the data is a security hole
1114 * root would have access anyway so we'll allow it
1115 */
1116 if (!is_suser()) {
1117 error = EACCES;
1118 } else {
1119 /*
1120 * set the file size
1121 */
1122 error = vnode_setsize(vp, offset, IO_NOZEROFILL,
1123 &context);
1124 }
1125
1126 (void)vnode_put(vp);
1127 goto outdrop;
1128
1129 case F_RDAHEAD:
1130 if (fp->f_type != DTYPE_VNODE) {
1131 error = EBADF;
1132 goto out;
1133 }
1134 if (uap->arg)
1135 fp->f_fglob->fg_flag &= ~FNORDAHEAD;
1136 else
1137 fp->f_fglob->fg_flag |= FNORDAHEAD;
1138
1139 goto out;
1140
1141 case F_NOCACHE:
1142 if (fp->f_type != DTYPE_VNODE) {
1143 error = EBADF;
1144 goto out;
1145 }
1146 if (uap->arg)
1147 fp->f_fglob->fg_flag |= FNOCACHE;
1148 else
1149 fp->f_fglob->fg_flag &= ~FNOCACHE;
1150
1151 goto out;
1152
1153 case F_NODIRECT:
1154 if (fp->f_type != DTYPE_VNODE) {
1155 error = EBADF;
1156 goto out;
1157 }
1158 if (uap->arg)
1159 fp->f_fglob->fg_flag |= FNODIRECT;
1160 else
1161 fp->f_fglob->fg_flag &= ~FNODIRECT;
1162
1163 goto out;
1164
1165 case F_SINGLE_WRITER:
1166 if (fp->f_type != DTYPE_VNODE) {
1167 error = EBADF;
1168 goto out;
1169 }
1170 if (uap->arg)
1171 fp->f_fglob->fg_flag |= FSINGLE_WRITER;
1172 else
1173 fp->f_fglob->fg_flag &= ~FSINGLE_WRITER;
1174
1175 goto out;
1176
1177 case F_GLOBAL_NOCACHE:
1178 if (fp->f_type != DTYPE_VNODE) {
1179 error = EBADF;
1180 goto out;
1181 }
1182 vp = (struct vnode *)fp->f_data;
1183 proc_fdunlock(p);
1184
1185 if ( (error = vnode_getwithref(vp)) == 0 ) {
1186
1187 *retval = vnode_isnocache(vp);
1188
1189 if (uap->arg)
1190 vnode_setnocache(vp);
1191 else
1192 vnode_clearnocache(vp);
1193
1194 (void)vnode_put(vp);
1195 }
1196 goto outdrop;
1197
1198 case F_CHECK_OPENEVT:
1199 if (fp->f_type != DTYPE_VNODE) {
1200 error = EBADF;
1201 goto out;
1202 }
1203 vp = (struct vnode *)fp->f_data;
1204 proc_fdunlock(p);
1205
1206 if ( (error = vnode_getwithref(vp)) == 0 ) {
1207
1208 *retval = vnode_is_openevt(vp);
1209
1210 if (uap->arg)
1211 vnode_set_openevt(vp);
1212 else
1213 vnode_clear_openevt(vp);
1214
1215 (void)vnode_put(vp);
1216 }
1217 goto outdrop;
1218
1219 case F_RDADVISE: {
1220 struct radvisory ra_struct;
1221
1222 if (fp->f_type != DTYPE_VNODE) {
1223 error = EBADF;
1224 goto out;
1225 }
1226 vp = (struct vnode *)fp->f_data;
1227 proc_fdunlock(p);
1228
1229 if ( (error = copyin(argp, (caddr_t)&ra_struct, sizeof(ra_struct))) )
1230 goto outdrop;
1231 if ( (error = vnode_getwithref(vp)) == 0 ) {
1232 error = VNOP_IOCTL(vp, F_RDADVISE, (caddr_t)&ra_struct, 0, &context);
1233
1234 (void)vnode_put(vp);
1235 }
1236 goto outdrop;
1237 }
1238
1239 case F_FLUSH_DATA:
1240
1241 if (fp->f_type != DTYPE_VNODE) {
1242 error = EBADF;
1243 goto out;
1244 }
1245 vp = (struct vnode *)fp->f_data;
1246 proc_fdunlock(p);
1247
1248 if ( (error = vnode_getwithref(vp)) == 0 ) {
1249 error = cluster_push(vp, 0);
1250
1251 (void)vnode_put(vp);
1252 }
1253 goto outdrop;
1254
1255 case F_LOG2PHYS:
1256 case F_LOG2PHYS_EXT: {
1257 struct log2phys l2p_struct; /* structure for allocate command */
1258 int devBlockSize;
1259
1260 off_t file_offset = 0;
1261 size_t a_size = 0;
1262 size_t run = 0;
1263
1264 if (uap->cmd == F_LOG2PHYS_EXT) {
1265 error = copyin(argp, (caddr_t)&l2p_struct, sizeof(l2p_struct));
1266 if (error)
1267 goto out;
1268 file_offset = l2p_struct.l2p_devoffset;
1269 } else {
1270 file_offset = fp->f_offset;
1271 }
1272 if (fp->f_type != DTYPE_VNODE) {
1273 error = EBADF;
1274 goto out;
1275 }
1276 vp = (struct vnode *)fp->f_data;
1277 proc_fdunlock(p);
1278 if ( (error = vnode_getwithref(vp)) ) {
1279 goto outdrop;
1280 }
1281 error = VNOP_OFFTOBLK(vp, file_offset, &lbn);
1282 if (error) {
1283 (void)vnode_put(vp);
1284 goto outdrop;
1285 }
1286 error = VNOP_BLKTOOFF(vp, lbn, &offset);
1287 if (error) {
1288 (void)vnode_put(vp);
1289 goto outdrop;
1290 }
1291 devBlockSize = vfs_devblocksize(vnode_mount(vp));
1292 if (uap->cmd == F_LOG2PHYS_EXT) {
1293 a_size = l2p_struct.l2p_contigbytes;
1294 } else {
1295 a_size = devBlockSize;
1296 }
1297
1298 error = VNOP_BLOCKMAP(vp, offset, a_size, &bn, &run, NULL, 0, &context);
1299
1300 (void)vnode_put(vp);
1301
1302 if (!error) {
1303 l2p_struct.l2p_flags = 0; /* for now */
1304 if (uap->cmd == F_LOG2PHYS_EXT) {
1305 l2p_struct.l2p_contigbytes = run - (file_offset - offset);
1306 } else {
1307 l2p_struct.l2p_contigbytes = 0; /* for now */
1308 }
1309 l2p_struct.l2p_devoffset = bn * devBlockSize;
1310 l2p_struct.l2p_devoffset += file_offset - offset;
1311 error = copyout((caddr_t)&l2p_struct, argp, sizeof(l2p_struct));
1312 }
1313 goto outdrop;
1314 }
1315 case F_GETPATH: {
1316 char *pathbufp;
1317 int pathlen;
1318
1319 if (fp->f_type != DTYPE_VNODE) {
1320 error = EBADF;
1321 goto out;
1322 }
1323 vp = (struct vnode *)fp->f_data;
1324 proc_fdunlock(p);
1325
1326 pathlen = MAXPATHLEN;
1327 MALLOC(pathbufp, char *, pathlen, M_TEMP, M_WAITOK);
1328 if (pathbufp == NULL) {
1329 error = ENOMEM;
1330 goto outdrop;
1331 }
1332 if ( (error = vnode_getwithref(vp)) == 0 ) {
1333 error = vn_getpath(vp, pathbufp, &pathlen);
1334 (void)vnode_put(vp);
1335
1336 if (error == 0)
1337 error = copyout((caddr_t)pathbufp, argp, pathlen);
1338 }
1339 FREE(pathbufp, M_TEMP);
1340 goto outdrop;
1341 }
1342
1343 case F_PATHPKG_CHECK: {
1344 char *pathbufp;
1345 size_t pathlen;
1346
1347 if (fp->f_type != DTYPE_VNODE) {
1348 error = EBADF;
1349 goto out;
1350 }
1351 vp = (struct vnode *)fp->f_data;
1352 proc_fdunlock(p);
1353
1354 pathlen = MAXPATHLEN;
1355 pathbufp = kalloc(MAXPATHLEN);
1356
1357 if ( (error = copyinstr(argp, pathbufp, MAXPATHLEN, &pathlen)) == 0 ) {
1358 if ( (error = vnode_getwithref(vp)) == 0 ) {
1359 AUDIT_ARG(text, pathbufp);
1360 error = vn_path_package_check(vp, pathbufp, pathlen, retval);
1361
1362 (void)vnode_put(vp);
1363 }
1364 }
1365 kfree(pathbufp, MAXPATHLEN);
1366 goto outdrop;
1367 }
1368
1369 case F_CHKCLEAN: // used by regression tests to see if all dirty pages got cleaned by fsync()
1370 case F_FULLFSYNC: // fsync + flush the journal + DKIOCSYNCHRONIZECACHE
1371 case F_FREEZE_FS: // freeze all other fs operations for the fs of this fd
1372 case F_THAW_FS: { // thaw all frozen fs operations for the fs of this fd
1373 if (fp->f_type != DTYPE_VNODE) {
1374 error = EBADF;
1375 goto out;
1376 }
1377 vp = (struct vnode *)fp->f_data;
1378 proc_fdunlock(p);
1379
1380 if ( (error = vnode_getwithref(vp)) == 0 ) {
1381 error = VNOP_IOCTL(vp, uap->cmd, (caddr_t)NULL, 0, &context);
1382
1383 (void)vnode_put(vp);
1384 }
1385 break;
1386 }
1387
1388 /*
1389 * SPI (private) for opening a file starting from a dir fd
1390 */
1391 case F_OPENFROM: {
1392 struct user_fopenfrom fopen;
1393 struct vnode_attr va;
1394 struct nameidata nd;
1395 int cmode;
1396
1397 /* Check if this isn't a valid file descriptor */
1398 if ((fp->f_type != DTYPE_VNODE) ||
1399 (fp->f_flag & FREAD) == 0) {
1400 error = EBADF;
1401 goto out;
1402 }
1403 vp = (struct vnode *)fp->f_data;
1404 proc_fdunlock(p);
1405
1406 if (vnode_getwithref(vp)) {
1407 error = ENOENT;
1408 goto outdrop;
1409 }
1410
1411 /* Only valid for directories */
1412 if (vp->v_type != VDIR) {
1413 vnode_put(vp);
1414 error = ENOTDIR;
1415 goto outdrop;
1416 }
1417
1418 /* Get flags, mode and pathname arguments. */
1419 if (IS_64BIT_PROCESS(p)) {
1420 error = copyin(argp, &fopen, sizeof(fopen));
1421 } else {
1422 struct user32_fopenfrom fopen32;
1423
1424 error = copyin(argp, &fopen32, sizeof(fopen32));
1425 fopen.o_flags = fopen32.o_flags;
1426 fopen.o_mode = fopen32.o_mode;
1427 fopen.o_pathname = CAST_USER_ADDR_T(fopen32.o_pathname);
1428 }
1429 if (error) {
1430 vnode_put(vp);
1431 goto outdrop;
1432 }
1433 AUDIT_ARG(fflags, fopen.o_flags);
1434 AUDIT_ARG(mode, fopen.o_mode);
1435 VATTR_INIT(&va);
1436 /* Mask off all but regular access permissions */
1437 cmode = ((fopen.o_mode &~ fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT;
1438 VATTR_SET(&va, va_mode, cmode & ACCESSPERMS);
1439
1440 /* Start the lookup relative to the file descriptor's vnode. */
1441 NDINIT(&nd, LOOKUP, OP_OPEN, USEDVP | FOLLOW | AUDITVNPATH1, UIO_USERSPACE,
1442 fopen.o_pathname, &context);
1443 nd.ni_dvp = vp;
1444
1445 error = open1(&context, &nd, fopen.o_flags, &va, retval);
1446
1447 vnode_put(vp);
1448 break;
1449 }
1450 /*
1451 * SPI (private) for unlinking a file starting from a dir fd
1452 */
1453 case F_UNLINKFROM: {
1454 struct nameidata nd;
1455 user_addr_t pathname;
1456
1457 /* Check if this isn't a valid file descriptor */
1458 if ((fp->f_type != DTYPE_VNODE) ||
1459 (fp->f_flag & FREAD) == 0) {
1460 error = EBADF;
1461 goto out;
1462 }
1463 vp = (struct vnode *)fp->f_data;
1464 proc_fdunlock(p);
1465
1466 if (vnode_getwithref(vp)) {
1467 error = ENOENT;
1468 goto outdrop;
1469 }
1470
1471 /* Only valid for directories */
1472 if (vp->v_type != VDIR) {
1473 vnode_put(vp);
1474 error = ENOTDIR;
1475 goto outdrop;
1476 }
1477
1478 /* Get flags, mode and pathname arguments. */
1479 if (IS_64BIT_PROCESS(p)) {
1480 pathname = (user_addr_t)argp;
1481 } else {
1482 pathname = CAST_USER_ADDR_T(argp);
1483 }
1484
1485 /* Start the lookup relative to the file descriptor's vnode. */
1486 NDINIT(&nd, DELETE, OP_UNLINK, USEDVP | AUDITVNPATH1, UIO_USERSPACE,
1487 pathname, &context);
1488 nd.ni_dvp = vp;
1489
1490 error = unlink1(&context, &nd, 0);
1491
1492 vnode_put(vp);
1493 break;
1494
1495 }
1496
1497 case F_ADDSIGS:
1498 case F_ADDFILESIGS:
1499 {
1500 struct user_fsignatures fs;
1501 kern_return_t kr;
1502 vm_offset_t kernel_blob_addr;
1503 vm_size_t kernel_blob_size;
1504
1505 if (fp->f_type != DTYPE_VNODE) {
1506 error = EBADF;
1507 goto out;
1508 }
1509 vp = (struct vnode *)fp->f_data;
1510 proc_fdunlock(p);
1511 error = vnode_getwithref(vp);
1512 if (error)
1513 goto outdrop;
1514
1515 if (IS_64BIT_PROCESS(p)) {
1516 error = copyin(argp, &fs, sizeof (fs));
1517 } else {
1518 struct user32_fsignatures fs32;
1519
1520 error = copyin(argp, &fs32, sizeof (fs32));
1521 fs.fs_file_start = fs32.fs_file_start;
1522 fs.fs_blob_start = CAST_USER_ADDR_T(fs32.fs_blob_start);
1523 fs.fs_blob_size = fs32.fs_blob_size;
1524 }
1525
1526 if (error) {
1527 vnode_put(vp);
1528 goto outdrop;
1529 }
1530
1531 if(ubc_cs_blob_get(vp, CPU_TYPE_ANY, fs.fs_file_start))
1532 {
1533 /*
1534 if(cs_debug)
1535 printf("CODE SIGNING: resident blob offered for: %s\n", vp->v_name);
1536 */
1537 vnode_put(vp);
1538 goto outdrop;
1539 }
1540
1541 #define CS_MAX_BLOB_SIZE (1280ULL * 1024) /* max shared cache file XXX ? */
1542 if (fs.fs_blob_size > CS_MAX_BLOB_SIZE) {
1543 error = E2BIG;
1544 vnode_put(vp);
1545 goto outdrop;
1546 }
1547
1548 kernel_blob_size = CAST_DOWN(vm_size_t, fs.fs_blob_size);
1549 kr = ubc_cs_blob_allocate(&kernel_blob_addr, &kernel_blob_size);
1550 if (kr != KERN_SUCCESS) {
1551 error = ENOMEM;
1552 vnode_put(vp);
1553 goto outdrop;
1554 }
1555
1556 if(uap->cmd == F_ADDSIGS) {
1557 error = copyin(fs.fs_blob_start,
1558 (void *) kernel_blob_addr,
1559 kernel_blob_size);
1560 } else /* F_ADDFILESIGS */ {
1561 error = vn_rdwr(UIO_READ,
1562 vp,
1563 (caddr_t) kernel_blob_addr,
1564 kernel_blob_size,
1565 fs.fs_file_start + fs.fs_blob_start,
1566 UIO_SYSSPACE,
1567 0,
1568 kauth_cred_get(),
1569 0,
1570 p);
1571 }
1572
1573 if (error) {
1574 ubc_cs_blob_deallocate(kernel_blob_addr,
1575 kernel_blob_size);
1576 vnode_put(vp);
1577 goto outdrop;
1578 }
1579
1580 error = ubc_cs_blob_add(
1581 vp,
1582 CPU_TYPE_ANY, /* not for a specific architecture */
1583 fs.fs_file_start,
1584 kernel_blob_addr,
1585 kernel_blob_size);
1586 if (error) {
1587 ubc_cs_blob_deallocate(kernel_blob_addr,
1588 kernel_blob_size);
1589 } else {
1590 /* ubc_blob_add() has consumed "kernel_blob_addr" */
1591 #if CHECK_CS_VALIDATION_BITMAP
1592 ubc_cs_validation_bitmap_allocate( vp );
1593 #endif
1594 }
1595
1596 (void) vnode_put(vp);
1597 break;
1598 }
1599
1600 case F_MARKDEPENDENCY: {
1601 struct vnode_attr va;
1602 vfs_context_t ctx = vfs_context_current();
1603 kauth_cred_t cred;
1604
1605 if ((current_proc()->p_flag & P_DEPENDENCY_CAPABLE) == 0) {
1606 error = EPERM;
1607 goto out;
1608 }
1609
1610 if (fp->f_type != DTYPE_VNODE) {
1611 error = EBADF;
1612 goto out;
1613 }
1614
1615 vp = (struct vnode *)fp->f_data;
1616 proc_fdunlock(p);
1617
1618 if (vnode_getwithref(vp)) {
1619 error = ENOENT;
1620 goto outdrop;
1621 }
1622
1623 if (!vnode_isvroot(vp)) {
1624 error = EINVAL;
1625 vnode_put(vp);
1626 goto outdrop;
1627 }
1628
1629 // get the owner of the root dir
1630 VATTR_INIT(&va);
1631 VATTR_WANTED(&va, va_uid);
1632 if (vnode_getattr(vp, &va, ctx) != 0) {
1633 error = EINVAL;
1634 vnode_put(vp);
1635 goto outdrop;
1636 }
1637
1638 // and last, check that the caller is the super user or
1639 // the owner of the mount point
1640 cred = vfs_context_ucred(ctx);
1641 if (!is_suser() && va.va_uid != kauth_cred_getuid(cred)) {
1642 error = EACCES;
1643 vnode_put(vp);
1644 goto outdrop;
1645 }
1646
1647 // if all those checks pass then we can mark the dependency
1648 vfs_markdependency(vp->v_mount);
1649 error = 0;
1650
1651 vnode_put(vp);
1652
1653 break;
1654 }
1655
1656 #if CONFIG_PROTECT
1657 case F_GETPROTECTIONCLASS: {
1658 int class = 0;
1659
1660 if (fp->f_type != DTYPE_VNODE) {
1661 error = EBADF;
1662 goto out;
1663 }
1664 vp = (struct vnode *)fp->f_data;
1665
1666 proc_fdunlock(p);
1667
1668 if (vnode_getwithref(vp)) {
1669 error = ENOENT;
1670 goto outdrop;
1671 }
1672
1673 error = cp_vnode_getclass (vp, &class);
1674 if (error == 0) {
1675 *retval = class;
1676 }
1677
1678 vnode_put(vp);
1679 break;
1680 }
1681
1682 case F_SETPROTECTIONCLASS: {
1683 /* tmp must be a valid PROTECTION_CLASS_* */
1684 tmp = CAST_DOWN_EXPLICIT(uint32_t, uap->arg);
1685
1686 if (fp->f_type != DTYPE_VNODE) {
1687 error = EBADF;
1688 goto out;
1689 }
1690 vp = (struct vnode *)fp->f_data;
1691
1692 proc_fdunlock(p);
1693
1694 if (vnode_getwithref(vp)) {
1695 error = ENOENT;
1696 goto outdrop;
1697 }
1698
1699 /* Only go forward if you have write access */
1700 vfs_context_t ctx = vfs_context_current();
1701 if(vnode_authorize(vp, NULLVP, (KAUTH_VNODE_ACCESS | KAUTH_VNODE_WRITE_DATA), ctx) != 0) {
1702 vnode_put(vp);
1703 error = EBADF;
1704 goto outdrop;
1705 }
1706 error = cp_vnode_setclass (vp, tmp);
1707 vnode_put(vp);
1708 break;
1709 }
1710
1711 case F_TRANSCODEKEY: {
1712
1713 if (fp->f_type != DTYPE_VNODE) {
1714 error = EBADF;
1715 goto out;
1716 }
1717
1718 vp = (struct vnode *)fp->f_data;
1719 proc_fdunlock(p);
1720
1721 if (vnode_getwithref(vp)) {
1722 error = ENOENT;
1723 goto outdrop;
1724 }
1725
1726 error = cp_vnode_transcode (vp);
1727 vnode_put(vp);
1728 break;
1729 }
1730
1731 case F_GETPROTECTIONLEVEL: {
1732 uint32_t cp_version = 0;
1733
1734 if (fp->f_type != DTYPE_VNODE) {
1735 error = EBADF;
1736 goto out;
1737 }
1738
1739 vp = (struct vnode*) fp->f_data;
1740 proc_fdunlock (p);
1741
1742 if (vnode_getwithref(vp)) {
1743 error = ENOENT;
1744 goto outdrop;
1745 }
1746
1747 /*
1748 * if cp_get_major_vers fails, error will be set to proper errno
1749 * and cp_version will still be 0.
1750 */
1751
1752 error = cp_get_root_major_vers (vp, &cp_version);
1753 *retval = cp_version;
1754
1755 vnode_put (vp);
1756 break;
1757 }
1758
1759 #endif /* CONFIG_PROTECT */
1760
1761 case F_MOVEDATAEXTENTS: {
1762 struct fileproc *fp2 = NULL;
1763 struct vnode *src_vp = NULLVP;
1764 struct vnode *dst_vp = NULLVP;
1765 /* We need to grab the 2nd FD out of the argments before moving on. */
1766 int fd2 = CAST_DOWN_EXPLICIT(int32_t, uap->arg);
1767
1768 if (fp->f_type != DTYPE_VNODE) {
1769 error = EBADF;
1770 goto out;
1771 }
1772
1773 /* For now, special case HFS+ only, since this is SPI. */
1774 src_vp = (struct vnode *)fp->f_data;
1775 if (src_vp->v_tag != VT_HFS) {
1776 error = EINVAL;
1777 goto out;
1778 }
1779
1780 /*
1781 * Get the references before we start acquiring iocounts on the vnodes,
1782 * while we still hold the proc fd lock
1783 */
1784 if ( (error = fp_lookup(p, fd2, &fp2, 1)) ) {
1785 error = EBADF;
1786 goto out;
1787 }
1788 if (fp2->f_type != DTYPE_VNODE) {
1789 fp_drop(p, fd2, fp2, 1);
1790 error = EBADF;
1791 goto out;
1792 }
1793 dst_vp = (struct vnode *)fp2->f_data;
1794 if (dst_vp->v_tag != VT_HFS) {
1795 fp_drop(p, fd2, fp2, 1);
1796 error = EINVAL;
1797 goto out;
1798 }
1799
1800 #if CONFIG_MACF
1801 /* Re-do MAC checks against the new FD, pass in a fake argument */
1802 error = mac_file_check_fcntl(proc_ucred(p), fp2->f_fglob, uap->cmd, 0);
1803 if (error) {
1804 fp_drop(p, fd2, fp2, 1);
1805 goto out;
1806 }
1807 #endif
1808 /* Audit the 2nd FD */
1809 AUDIT_ARG(fd, fd2);
1810
1811 proc_fdunlock(p);
1812
1813 if (vnode_getwithref(src_vp)) {
1814 fp_drop(p, fd2, fp2, 0);
1815 error = ENOENT;
1816 goto outdrop;
1817 }
1818 if (vnode_getwithref(dst_vp)) {
1819 vnode_put (src_vp);
1820 fp_drop(p, fd2, fp2, 0);
1821 error = ENOENT;
1822 goto outdrop;
1823 }
1824
1825 /*
1826 * Basic asserts; validate they are not the same and that
1827 * both live on the same filesystem.
1828 */
1829 if (dst_vp == src_vp) {
1830 vnode_put (src_vp);
1831 vnode_put (dst_vp);
1832 fp_drop (p, fd2, fp2, 0);
1833 error = EINVAL;
1834 goto outdrop;
1835 }
1836
1837 if (dst_vp->v_mount != src_vp->v_mount) {
1838 vnode_put (src_vp);
1839 vnode_put (dst_vp);
1840 fp_drop (p, fd2, fp2, 0);
1841 error = EXDEV;
1842 goto outdrop;
1843 }
1844
1845 /* Now we have a legit pair of FDs. Go to work */
1846
1847 /* Now check for write access to the target files */
1848 if(vnode_authorize(src_vp, NULLVP,
1849 (KAUTH_VNODE_ACCESS | KAUTH_VNODE_WRITE_DATA), &context) != 0) {
1850 vnode_put(src_vp);
1851 vnode_put(dst_vp);
1852 fp_drop(p, fd2, fp2, 0);
1853 error = EBADF;
1854 goto outdrop;
1855 }
1856
1857 if(vnode_authorize(dst_vp, NULLVP,
1858 (KAUTH_VNODE_ACCESS | KAUTH_VNODE_WRITE_DATA), &context) != 0) {
1859 vnode_put(src_vp);
1860 vnode_put(dst_vp);
1861 fp_drop(p, fd2, fp2, 0);
1862 error = EBADF;
1863 goto outdrop;
1864 }
1865
1866 /* Verify that both vps point to files and not directories */
1867 if ( !vnode_isreg(src_vp) || !vnode_isreg(dst_vp)) {
1868 error = EINVAL;
1869 vnode_put (src_vp);
1870 vnode_put (dst_vp);
1871 fp_drop (p, fd2, fp2, 0);
1872 goto outdrop;
1873 }
1874
1875 /*
1876 * The exchangedata syscall handler passes in 0 for the flags to VNOP_EXCHANGE.
1877 * We'll pass in our special bit indicating that the new behavior is expected
1878 */
1879
1880 error = VNOP_EXCHANGE(src_vp, dst_vp, FSOPT_EXCHANGE_DATA_ONLY, &context);
1881
1882 vnode_put (src_vp);
1883 vnode_put (dst_vp);
1884 fp_drop(p, fd2, fp2, 0);
1885 break;
1886 }
1887
1888
1889 /*
1890 * SPI (private) for indicating to a filesystem that subsequent writes to
1891 * the open FD will represent static content.
1892 */
1893 case F_SETSTATICCONTENT: {
1894 caddr_t ioctl_arg = NULL;
1895
1896 if (uap->arg) {
1897 ioctl_arg = (caddr_t) 1;
1898 }
1899
1900 if (fp->f_type != DTYPE_VNODE) {
1901 error = EBADF;
1902 goto out;
1903 }
1904 vp = (struct vnode *)fp->f_data;
1905 proc_fdunlock(p);
1906
1907 error = vnode_getwithref(vp);
1908 if (error) {
1909 error = ENOENT;
1910 goto outdrop;
1911 }
1912
1913 /* Only go forward if you have write access */
1914 vfs_context_t ctx = vfs_context_current();
1915 if(vnode_authorize(vp, NULLVP, (KAUTH_VNODE_ACCESS | KAUTH_VNODE_WRITE_DATA), ctx) != 0) {
1916 vnode_put(vp);
1917 error = EBADF;
1918 goto outdrop;
1919 }
1920
1921 error = VNOP_IOCTL(vp, uap->cmd, ioctl_arg, 0, &context);
1922 (void)vnode_put(vp);
1923
1924 break;
1925 }
1926
1927 /*
1928 * Set the vnode pointed to by 'fd'
1929 * and tag it as the (potentially future) backing store
1930 * for another filesystem
1931 */
1932 case F_SETBACKINGSTORE: {
1933 if (fp->f_type != DTYPE_VNODE) {
1934 error = EBADF;
1935 goto out;
1936 }
1937
1938 vp = (struct vnode *)fp->f_data;
1939
1940 if (vp->v_tag != VT_HFS) {
1941 error = EINVAL;
1942 goto out;
1943 }
1944 proc_fdunlock(p);
1945
1946 if (vnode_getwithref(vp)) {
1947 error = ENOENT;
1948 goto outdrop;
1949 }
1950
1951 /* only proceed if you have write access */
1952 vfs_context_t ctx = vfs_context_current();
1953 if(vnode_authorize(vp, NULLVP, (KAUTH_VNODE_ACCESS | KAUTH_VNODE_WRITE_DATA), ctx) != 0) {
1954 vnode_put(vp);
1955 error = EBADF;
1956 goto outdrop;
1957 }
1958
1959
1960 /* If arg != 0, set, otherwise unset */
1961 if (uap->arg) {
1962 error = VNOP_IOCTL (vp, uap->cmd, (caddr_t)1, 0, &context);
1963 }
1964 else {
1965 error = VNOP_IOCTL (vp, uap->cmd, (caddr_t)NULL, 0, &context);
1966 }
1967
1968 vnode_put(vp);
1969 break;
1970 }
1971
1972 /*
1973 * like F_GETPATH, but special semantics for
1974 * the mobile time machine handler.
1975 */
1976 case F_GETPATH_MTMINFO: {
1977 char *pathbufp;
1978 int pathlen;
1979
1980 if (fp->f_type != DTYPE_VNODE) {
1981 error = EBADF;
1982 goto out;
1983 }
1984 vp = (struct vnode *)fp->f_data;
1985 proc_fdunlock(p);
1986
1987 pathlen = MAXPATHLEN;
1988 MALLOC(pathbufp, char *, pathlen, M_TEMP, M_WAITOK);
1989 if (pathbufp == NULL) {
1990 error = ENOMEM;
1991 goto outdrop;
1992 }
1993 if ( (error = vnode_getwithref(vp)) == 0 ) {
1994 int backingstore = 0;
1995
1996 /* Check for error from vn_getpath before moving on */
1997 if ((error = vn_getpath(vp, pathbufp, &pathlen)) == 0) {
1998 if (vp->v_tag == VT_HFS) {
1999 error = VNOP_IOCTL (vp, uap->cmd, (caddr_t) &backingstore, 0, &context);
2000 }
2001 (void)vnode_put(vp);
2002
2003 if (error == 0) {
2004 error = copyout((caddr_t)pathbufp, argp, pathlen);
2005 }
2006 if (error == 0) {
2007 /*
2008 * If the copyout was successful, now check to ensure
2009 * that this vnode is not a BACKINGSTORE vnode. mtmd
2010 * wants the path regardless.
2011 */
2012 if (backingstore) {
2013 error = EBUSY;
2014 }
2015 }
2016 } else
2017 (void)vnode_put(vp);
2018 }
2019 FREE(pathbufp, M_TEMP);
2020 goto outdrop;
2021 }
2022
2023 default:
2024 /*
2025 * This is an fcntl() that we d not recognize at this level;
2026 * if this is a vnode, we send it down into the VNOP_IOCTL
2027 * for this vnode; this can include special devices, and will
2028 * effectively overload fcntl() to send ioctl()'s.
2029 */
2030 if((uap->cmd & IOC_VOID) && (uap->cmd & IOC_INOUT)){
2031 error = EINVAL;
2032 goto out;
2033 }
2034
2035 if (fp->f_type != DTYPE_VNODE) {
2036 error = EBADF;
2037 goto out;
2038 }
2039 vp = (struct vnode *)fp->f_data;
2040 proc_fdunlock(p);
2041
2042 if ( (error = vnode_getwithref(vp)) == 0 ) {
2043 #define STK_PARAMS 128
2044 char stkbuf[STK_PARAMS];
2045 unsigned int size;
2046 caddr_t data, memp;
2047 /*
2048 * For this to work properly, we have to copy in the
2049 * ioctl() cmd argument if there is one; we must also
2050 * check that a command parameter, if present, does
2051 * not exceed the maximum command length dictated by
2052 * the number of bits we have available in the command
2053 * to represent a structure length. Finally, we have
2054 * to copy the results back out, if it is that type of
2055 * ioctl().
2056 */
2057 size = IOCPARM_LEN(uap->cmd);
2058 if (size > IOCPARM_MAX) {
2059 (void)vnode_put(vp);
2060 error = EINVAL;
2061 break;
2062 }
2063
2064 memp = NULL;
2065 if (size > sizeof (stkbuf)) {
2066 if ((memp = (caddr_t)kalloc(size)) == 0) {
2067 (void)vnode_put(vp);
2068 error = ENOMEM;
2069 goto outdrop;
2070 }
2071 data = memp;
2072 } else {
2073 data = &stkbuf[0];
2074 }
2075
2076 if (uap->cmd & IOC_IN) {
2077 if (size) {
2078 /* structure */
2079 error = copyin(argp, data, size);
2080 if (error) {
2081 (void)vnode_put(vp);
2082 if (memp)
2083 kfree(memp, size);
2084 goto outdrop;
2085 }
2086 } else {
2087 /* int */
2088 if (is64bit) {
2089 *(user_addr_t *)data = argp;
2090 } else {
2091 *(uint32_t *)data = (uint32_t)argp;
2092 }
2093 };
2094 } else if ((uap->cmd & IOC_OUT) && size) {
2095 /*
2096 * Zero the buffer so the user always
2097 * gets back something deterministic.
2098 */
2099 bzero(data, size);
2100 } else if (uap->cmd & IOC_VOID) {
2101 if (is64bit) {
2102 *(user_addr_t *)data = argp;
2103 } else {
2104 *(uint32_t *)data = (uint32_t)argp;
2105 }
2106 }
2107
2108 error = VNOP_IOCTL(vp, uap->cmd, CAST_DOWN(caddr_t, data), 0, &context);
2109
2110 (void)vnode_put(vp);
2111
2112 /* Copy any output data to user */
2113 if (error == 0 && (uap->cmd & IOC_OUT) && size)
2114 error = copyout(data, argp, size);
2115 if (memp)
2116 kfree(memp, size);
2117 }
2118 break;
2119 }
2120
2121 outdrop:
2122 AUDIT_ARG(vnpath_withref, vp, ARG_VNODE1);
2123 fp_drop(p, fd, fp, 0);
2124 return(error);
2125 out:
2126 fp_drop(p, fd, fp, 1);
2127 proc_fdunlock(p);
2128 return(error);
2129 }
2130
2131
2132 /*
2133 * finishdup
2134 *
2135 * Description: Common code for dup, dup2, and fcntl(F_DUPFD).
2136 *
2137 * Parameters: p Process performing the dup
2138 * old The fd to dup
2139 * new The fd to dup it to
2140 * fd_flags Flags to augment the new fd
2141 * retval Pointer to the call return area
2142 *
2143 * Returns: 0 Success
2144 * EBADF
2145 * ENOMEM
2146 *
2147 * Implicit returns:
2148 * *retval (modified) The new descriptor
2149 *
2150 * Locks: Assumes proc_fdlock for process pointing to fdp is held by
2151 * the caller
2152 *
2153 * Notes: This function may drop and reacquire this lock; it is unsafe
2154 * for a caller to assume that other state protected by the lock
2155 * has not been subsequently changed out from under it.
2156 */
2157 int
2158 finishdup(proc_t p,
2159 struct filedesc *fdp, int old, int new, int fd_flags, int32_t *retval)
2160 {
2161 struct fileproc *nfp;
2162 struct fileproc *ofp;
2163 #if CONFIG_MACF
2164 int error;
2165 #endif
2166
2167 #if DIAGNOSTIC
2168 proc_fdlock_assert(p, LCK_MTX_ASSERT_OWNED);
2169 #endif
2170 if ((ofp = fdp->fd_ofiles[old]) == NULL ||
2171 (fdp->fd_ofileflags[old] & UF_RESERVED)) {
2172 fdrelse(p, new);
2173 return (EBADF);
2174 }
2175 fg_ref(ofp);
2176
2177 #if CONFIG_MACF
2178 error = mac_file_check_dup(proc_ucred(p), ofp->f_fglob, new);
2179 if (error) {
2180 fg_drop(ofp);
2181 fdrelse(p, new);
2182 return (error);
2183 }
2184 #endif
2185
2186 proc_fdunlock(p);
2187
2188 MALLOC_ZONE(nfp, struct fileproc *, sizeof(struct fileproc), M_FILEPROC, M_WAITOK);
2189 /* Failure check follows proc_fdlock() due to handling requirements */
2190
2191 proc_fdlock(p);
2192
2193 if (nfp == NULL) {
2194 fg_drop(ofp);
2195 fdrelse(p, new);
2196 return (ENOMEM);
2197 }
2198
2199 bzero(nfp, sizeof(struct fileproc));
2200
2201 nfp->f_flags = 0;
2202 nfp->f_fglob = ofp->f_fglob;
2203 nfp->f_iocount = 0;
2204
2205 #if DIAGNOSTIC
2206 if (fdp->fd_ofiles[new] != 0)
2207 panic("finishdup: overwriting fd_ofiles with new %d", new);
2208 if ((fdp->fd_ofileflags[new] & UF_RESERVED) == 0)
2209 panic("finishdup: unreserved fileflags with new %d", new);
2210 #endif
2211
2212 if (new > fdp->fd_lastfile)
2213 fdp->fd_lastfile = new;
2214 *fdflags(p, new) |= fd_flags;
2215 procfdtbl_releasefd(p, new, nfp);
2216 *retval = new;
2217 return (0);
2218 }
2219
2220
2221 /*
2222 * close
2223 *
2224 * Description: The implementation of the close(2) system call
2225 *
2226 * Parameters: p Process in whose per process file table
2227 * the close is to occur
2228 * uap->fd fd to be closed
2229 * retval <unused>
2230 *
2231 * Returns: 0 Success
2232 * fp_lookup:EBADF Bad file descriptor
2233 * close_internal:EBADF
2234 * close_internal:??? Anything returnable by a per-fileops
2235 * close function
2236 */
2237 int
2238 close(proc_t p, struct close_args *uap, int32_t *retval)
2239 {
2240 __pthread_testcancel(1);
2241 return(close_nocancel(p, (struct close_nocancel_args *)uap, retval));
2242 }
2243
2244
2245 int
2246 close_nocancel(proc_t p, struct close_nocancel_args *uap, __unused int32_t *retval)
2247 {
2248 struct fileproc *fp;
2249 int fd = uap->fd;
2250 int error =0;
2251
2252 AUDIT_SYSCLOSE(p, fd);
2253
2254 proc_fdlock(p);
2255
2256 if ( (error = fp_lookup(p,fd,&fp, 1)) ) {
2257 proc_fdunlock(p);
2258 return(error);
2259 }
2260
2261 error = close_internal_locked(p, fd, fp, 0);
2262
2263 proc_fdunlock(p);
2264
2265 return(error);
2266 }
2267
2268
2269 /*
2270 * close_internal_locked
2271 *
2272 * Close a file descriptor.
2273 *
2274 * Parameters: p Process in whose per process file table
2275 * the close is to occur
2276 * fd fd to be closed
2277 * fp fileproc associated with the fd
2278 *
2279 * Returns: 0 Success
2280 * EBADF fd already in close wait state
2281 * closef_locked:??? Anything returnable by a per-fileops
2282 * close function
2283 *
2284 * Locks: Assumes proc_fdlock for process is held by the caller and returns
2285 * with lock held
2286 *
2287 * Notes: This function may drop and reacquire this lock; it is unsafe
2288 * for a caller to assume that other state protected by the lock
2289 * has not been subsequently changes out from under it, if the
2290 * caller made the call with the lock held.
2291 */
2292 static int
2293 close_internal_locked(proc_t p, int fd, struct fileproc *fp, int flags)
2294 {
2295 struct filedesc *fdp = p->p_fd;
2296 int error =0;
2297 int resvfd = flags & FD_DUP2RESV;
2298
2299
2300 #if DIAGNOSTIC
2301 proc_fdlock_assert(p, LCK_MTX_ASSERT_OWNED);
2302 #endif
2303
2304 /* Keep people from using the filedesc while we are closing it */
2305 procfdtbl_markclosefd(p, fd);
2306
2307
2308 if ((fp->f_flags & FP_CLOSING) == FP_CLOSING) {
2309 panic("close_internal_locked: being called on already closing fd");
2310 }
2311
2312
2313 #if DIAGNOSTIC
2314 if ((fdp->fd_ofileflags[fd] & UF_RESERVED) == 0)
2315 panic("close_internal: unreserved fileflags with fd %d", fd);
2316 #endif
2317
2318 fp->f_flags |= FP_CLOSING;
2319
2320 if ( (fp->f_flags & FP_AIOISSUED) || kauth_authorize_fileop_has_listeners() ) {
2321
2322 proc_fdunlock(p);
2323
2324 if ( (fp->f_type == DTYPE_VNODE) && kauth_authorize_fileop_has_listeners() ) {
2325 /*
2326 * call out to allow 3rd party notification of close.
2327 * Ignore result of kauth_authorize_fileop call.
2328 */
2329 if (vnode_getwithref((vnode_t)fp->f_data) == 0) {
2330 u_int fileop_flags = 0;
2331 if ((fp->f_flags & FP_WRITTEN) != 0)
2332 fileop_flags |= KAUTH_FILEOP_CLOSE_MODIFIED;
2333 kauth_authorize_fileop(fp->f_fglob->fg_cred, KAUTH_FILEOP_CLOSE,
2334 (uintptr_t)fp->f_data, (uintptr_t)fileop_flags);
2335 vnode_put((vnode_t)fp->f_data);
2336 }
2337 }
2338 if (fp->f_flags & FP_AIOISSUED)
2339 /*
2340 * cancel all async IO requests that can be cancelled.
2341 */
2342 _aio_close( p, fd );
2343
2344 proc_fdlock(p);
2345 }
2346
2347 if (fd < fdp->fd_knlistsize)
2348 knote_fdclose(p, fd);
2349
2350 if (fp->f_flags & FP_WAITEVENT)
2351 (void)waitevent_close(p, fp);
2352
2353 if ((fp->f_flags & FP_INCHRREAD) == 0)
2354 fileproc_drain(p, fp);
2355
2356 if (resvfd == 0) {
2357 _fdrelse(p, fd);
2358 } else {
2359 procfdtbl_reservefd(p, fd);
2360 }
2361
2362 error = closef_locked(fp, fp->f_fglob, p);
2363 if ((fp->f_flags & FP_WAITCLOSE) == FP_WAITCLOSE)
2364 wakeup(&fp->f_flags);
2365 fp->f_flags &= ~(FP_WAITCLOSE | FP_CLOSING);
2366
2367 proc_fdunlock(p);
2368
2369 FREE_ZONE(fp, sizeof(*fp), M_FILEPROC);
2370
2371 proc_fdlock(p);
2372
2373 #if DIAGNOSTIC
2374 if (resvfd != 0) {
2375 if ((fdp->fd_ofileflags[fd] & UF_RESERVED) == 0)
2376 panic("close with reserved fd returns with freed fd:%d: proc: %p", fd, p);
2377 }
2378 #endif
2379
2380 return(error);
2381 }
2382
2383
2384 /*
2385 * fstat1
2386 *
2387 * Description: Return status information about a file descriptor.
2388 *
2389 * Parameters: p The process doing the fstat
2390 * fd The fd to stat
2391 * ub The user stat buffer
2392 * xsecurity The user extended security
2393 * buffer, or 0 if none
2394 * xsecurity_size The size of xsecurity, or 0
2395 * if no xsecurity
2396 * isstat64 Flag to indicate 64 bit version
2397 * for inode size, etc.
2398 *
2399 * Returns: 0 Success
2400 * EBADF
2401 * EFAULT
2402 * fp_lookup:EBADF Bad file descriptor
2403 * vnode_getwithref:???
2404 * copyout:EFAULT
2405 * vnode_getwithref:???
2406 * vn_stat:???
2407 * soo_stat:???
2408 * pipe_stat:???
2409 * pshm_stat:???
2410 * kqueue_stat:???
2411 *
2412 * Notes: Internal implementation for all other fstat() related
2413 * functions
2414 *
2415 * XXX switch on node type is bogus; need a stat in struct
2416 * XXX fileops instead.
2417 */
2418 static int
2419 fstat1(proc_t p, int fd, user_addr_t ub, user_addr_t xsecurity, user_addr_t xsecurity_size, int isstat64)
2420 {
2421 struct fileproc *fp;
2422 union {
2423 struct stat sb;
2424 struct stat64 sb64;
2425 } source;
2426 union {
2427 struct user64_stat user64_sb;
2428 struct user32_stat user32_sb;
2429 struct user64_stat64 user64_sb64;
2430 struct user32_stat64 user32_sb64;
2431 } dest;
2432 int error, my_size;
2433 int funnel_state;
2434 file_type_t type;
2435 caddr_t data;
2436 kauth_filesec_t fsec;
2437 user_size_t xsecurity_bufsize;
2438 vfs_context_t ctx = vfs_context_current();
2439 void * sbptr;
2440
2441
2442 AUDIT_ARG(fd, fd);
2443
2444 if ((error = fp_lookup(p, fd, &fp, 0)) != 0) {
2445 return(error);
2446 }
2447 type = fp->f_type;
2448 data = fp->f_data;
2449 fsec = KAUTH_FILESEC_NONE;
2450
2451 sbptr = (void *)&source;
2452
2453 switch (type) {
2454
2455 case DTYPE_VNODE:
2456 if ((error = vnode_getwithref((vnode_t)data)) == 0) {
2457 /*
2458 * If the caller has the file open, and is not
2459 * requesting extended security information, we are
2460 * going to let them get the basic stat information.
2461 */
2462 if (xsecurity == USER_ADDR_NULL) {
2463 error = vn_stat_noauth((vnode_t)data, sbptr, NULL, isstat64, ctx);
2464 } else {
2465 error = vn_stat((vnode_t)data, sbptr, &fsec, isstat64, ctx);
2466 }
2467
2468 AUDIT_ARG(vnpath, (struct vnode *)data, ARG_VNODE1);
2469 (void)vnode_put((vnode_t)data);
2470 }
2471 break;
2472
2473 #if SOCKETS
2474 case DTYPE_SOCKET:
2475 error = soo_stat((struct socket *)data, sbptr, isstat64);
2476 break;
2477 #endif /* SOCKETS */
2478
2479 case DTYPE_PIPE:
2480 error = pipe_stat((void *)data, sbptr, isstat64);
2481 break;
2482
2483 case DTYPE_PSXSHM:
2484 error = pshm_stat((void *)data, sbptr, isstat64);
2485 break;
2486
2487 case DTYPE_KQUEUE:
2488 funnel_state = thread_funnel_set(kernel_flock, TRUE);
2489 error = kqueue_stat(fp, sbptr, isstat64, p);
2490 thread_funnel_set(kernel_flock, funnel_state);
2491 break;
2492
2493 default:
2494 error = EBADF;
2495 goto out;
2496 }
2497 if (error == 0) {
2498 caddr_t sbp;
2499
2500 if (isstat64 != 0) {
2501 source.sb64.st_lspare = 0;
2502 source.sb64.st_qspare[0] = 0LL;
2503 source.sb64.st_qspare[1] = 0LL;
2504
2505 if (IS_64BIT_PROCESS(current_proc())) {
2506 munge_user64_stat64(&source.sb64, &dest.user64_sb64);
2507 my_size = sizeof(dest.user64_sb64);
2508 sbp = (caddr_t)&dest.user64_sb64;
2509 } else {
2510 munge_user32_stat64(&source.sb64, &dest.user32_sb64);
2511 my_size = sizeof(dest.user32_sb64);
2512 sbp = (caddr_t)&dest.user32_sb64;
2513 }
2514 } else {
2515 source.sb.st_lspare = 0;
2516 source.sb.st_qspare[0] = 0LL;
2517 source.sb.st_qspare[1] = 0LL;
2518 if (IS_64BIT_PROCESS(current_proc())) {
2519 munge_user64_stat(&source.sb, &dest.user64_sb);
2520 my_size = sizeof(dest.user64_sb);
2521 sbp = (caddr_t)&dest.user64_sb;
2522 } else {
2523 munge_user32_stat(&source.sb, &dest.user32_sb);
2524 my_size = sizeof(dest.user32_sb);
2525 sbp = (caddr_t)&dest.user32_sb;
2526 }
2527 }
2528
2529 error = copyout(sbp, ub, my_size);
2530 }
2531
2532 /* caller wants extended security information? */
2533 if (xsecurity != USER_ADDR_NULL) {
2534
2535 /* did we get any? */
2536 if (fsec == KAUTH_FILESEC_NONE) {
2537 if (susize(xsecurity_size, 0) != 0) {
2538 error = EFAULT;
2539 goto out;
2540 }
2541 } else {
2542 /* find the user buffer size */
2543 xsecurity_bufsize = fusize(xsecurity_size);
2544
2545 /* copy out the actual data size */
2546 if (susize(xsecurity_size, KAUTH_FILESEC_COPYSIZE(fsec)) != 0) {
2547 error = EFAULT;
2548 goto out;
2549 }
2550
2551 /* if the caller supplied enough room, copy out to it */
2552 if (xsecurity_bufsize >= KAUTH_FILESEC_COPYSIZE(fsec))
2553 error = copyout(fsec, xsecurity, KAUTH_FILESEC_COPYSIZE(fsec));
2554 }
2555 }
2556 out:
2557 fp_drop(p, fd, fp, 0);
2558 if (fsec != NULL)
2559 kauth_filesec_free(fsec);
2560 return (error);
2561 }
2562
2563
2564 /*
2565 * fstat_extended
2566 *
2567 * Description: Extended version of fstat supporting returning extended
2568 * security information
2569 *
2570 * Parameters: p The process doing the fstat
2571 * uap->fd The fd to stat
2572 * uap->ub The user stat buffer
2573 * uap->xsecurity The user extended security
2574 * buffer, or 0 if none
2575 * uap->xsecurity_size The size of xsecurity, or 0
2576 *
2577 * Returns: 0 Success
2578 * !0 Errno (see fstat1)
2579 */
2580 int
2581 fstat_extended(proc_t p, struct fstat_extended_args *uap, __unused int32_t *retval)
2582 {
2583 return(fstat1(p, uap->fd, uap->ub, uap->xsecurity, uap->xsecurity_size, 0));
2584 }
2585
2586
2587 /*
2588 * fstat
2589 *
2590 * Description: Get file status for the file associated with fd
2591 *
2592 * Parameters: p The process doing the fstat
2593 * uap->fd The fd to stat
2594 * uap->ub The user stat buffer
2595 *
2596 * Returns: 0 Success
2597 * !0 Errno (see fstat1)
2598 */
2599 int
2600 fstat(proc_t p, register struct fstat_args *uap, __unused int32_t *retval)
2601 {
2602 return(fstat1(p, uap->fd, uap->ub, 0, 0, 0));
2603 }
2604
2605
2606 /*
2607 * fstat64_extended
2608 *
2609 * Description: Extended version of fstat64 supporting returning extended
2610 * security information
2611 *
2612 * Parameters: p The process doing the fstat
2613 * uap->fd The fd to stat
2614 * uap->ub The user stat buffer
2615 * uap->xsecurity The user extended security
2616 * buffer, or 0 if none
2617 * uap->xsecurity_size The size of xsecurity, or 0
2618 *
2619 * Returns: 0 Success
2620 * !0 Errno (see fstat1)
2621 */
2622 int
2623 fstat64_extended(proc_t p, struct fstat64_extended_args *uap, __unused int32_t *retval)
2624 {
2625 return(fstat1(p, uap->fd, uap->ub, uap->xsecurity, uap->xsecurity_size, 1));
2626 }
2627
2628
2629 /*
2630 * fstat64
2631 *
2632 * Description: Get 64 bit version of the file status for the file associated
2633 * with fd
2634 *
2635 * Parameters: p The process doing the fstat
2636 * uap->fd The fd to stat
2637 * uap->ub The user stat buffer
2638 *
2639 * Returns: 0 Success
2640 * !0 Errno (see fstat1)
2641 */
2642 int
2643 fstat64(proc_t p, register struct fstat64_args *uap, __unused int32_t *retval)
2644 {
2645 return(fstat1(p, uap->fd, uap->ub, 0, 0, 1));
2646 }
2647
2648
2649 /*
2650 * fpathconf
2651 *
2652 * Description: Return pathconf information about a file descriptor.
2653 *
2654 * Parameters: p Process making the request
2655 * uap->fd fd to get information about
2656 * uap->name Name of information desired
2657 * retval Pointer to the call return area
2658 *
2659 * Returns: 0 Success
2660 * EINVAL
2661 * fp_lookup:EBADF Bad file descriptor
2662 * vnode_getwithref:???
2663 * vn_pathconf:???
2664 *
2665 * Implicit returns:
2666 * *retval (modified) Returned information (numeric)
2667 */
2668 int
2669 fpathconf(proc_t p, struct fpathconf_args *uap, int32_t *retval)
2670 {
2671 int fd = uap->fd;
2672 struct fileproc *fp;
2673 struct vnode *vp;
2674 int error = 0;
2675 file_type_t type;
2676 caddr_t data;
2677
2678
2679 AUDIT_ARG(fd, uap->fd);
2680 if ( (error = fp_lookup(p, fd, &fp, 0)) )
2681 return(error);
2682 type = fp->f_type;
2683 data = fp->f_data;
2684
2685 switch (type) {
2686
2687 case DTYPE_SOCKET:
2688 if (uap->name != _PC_PIPE_BUF) {
2689 error = EINVAL;
2690 goto out;
2691 }
2692 *retval = PIPE_BUF;
2693 error = 0;
2694 goto out;
2695
2696 case DTYPE_PIPE:
2697 if (uap->name != _PC_PIPE_BUF) {
2698 error = EINVAL;
2699 goto out;
2700 }
2701 *retval = PIPE_BUF;
2702 error = 0;
2703 goto out;
2704
2705 case DTYPE_VNODE:
2706 vp = (struct vnode *)data;
2707
2708 if ( (error = vnode_getwithref(vp)) == 0) {
2709 AUDIT_ARG(vnpath, vp, ARG_VNODE1);
2710
2711 error = vn_pathconf(vp, uap->name, retval, vfs_context_current());
2712
2713 (void)vnode_put(vp);
2714 }
2715 goto out;
2716
2717 case DTYPE_PSXSHM:
2718 case DTYPE_PSXSEM:
2719 case DTYPE_KQUEUE:
2720 case DTYPE_FSEVENTS:
2721 error = EINVAL;
2722 goto out;
2723
2724 }
2725 /*NOTREACHED*/
2726 out:
2727 fp_drop(p, fd, fp, 0);
2728 return(error);
2729 }
2730
2731 /*
2732 * Statistics counter for the number of times a process calling fdalloc()
2733 * has resulted in an expansion of the per process open file table.
2734 *
2735 * XXX This would likely be of more use if it were per process
2736 */
2737 int fdexpand;
2738
2739
2740 /*
2741 * fdalloc
2742 *
2743 * Description: Allocate a file descriptor for the process.
2744 *
2745 * Parameters: p Process to allocate the fd in
2746 * want The fd we would prefer to get
2747 * result Pointer to fd we got
2748 *
2749 * Returns: 0 Success
2750 * EMFILE
2751 * ENOMEM
2752 *
2753 * Implicit returns:
2754 * *result (modified) The fd which was allocated
2755 */
2756 int
2757 fdalloc(proc_t p, int want, int *result)
2758 {
2759 struct filedesc *fdp = p->p_fd;
2760 int i;
2761 int lim, last, numfiles, oldnfiles;
2762 struct fileproc **newofiles, **ofiles;
2763 char *newofileflags;
2764
2765 /*
2766 * Search for a free descriptor starting at the higher
2767 * of want or fd_freefile. If that fails, consider
2768 * expanding the ofile array.
2769 */
2770 #if DIAGNOSTIC
2771 proc_fdlock_assert(p, LCK_MTX_ASSERT_OWNED);
2772 #endif
2773
2774 lim = min((int)p->p_rlimit[RLIMIT_NOFILE].rlim_cur, maxfiles);
2775 for (;;) {
2776 last = min(fdp->fd_nfiles, lim);
2777 if ((i = want) < fdp->fd_freefile)
2778 i = fdp->fd_freefile;
2779 for (; i < last; i++) {
2780 if (fdp->fd_ofiles[i] == NULL && !(fdp->fd_ofileflags[i] & UF_RESERVED)) {
2781 procfdtbl_reservefd(p, i);
2782 if (i > fdp->fd_lastfile)
2783 fdp->fd_lastfile = i;
2784 if (want <= fdp->fd_freefile)
2785 fdp->fd_freefile = i;
2786 *result = i;
2787 return (0);
2788 }
2789 }
2790
2791 /*
2792 * No space in current array. Expand?
2793 */
2794 if (fdp->fd_nfiles >= lim)
2795 return (EMFILE);
2796 if (fdp->fd_nfiles < NDEXTENT)
2797 numfiles = NDEXTENT;
2798 else
2799 numfiles = 2 * fdp->fd_nfiles;
2800 /* Enforce lim */
2801 if (numfiles > lim)
2802 numfiles = lim;
2803 proc_fdunlock(p);
2804 MALLOC_ZONE(newofiles, struct fileproc **,
2805 numfiles * OFILESIZE, M_OFILETABL, M_WAITOK);
2806 proc_fdlock(p);
2807 if (newofiles == NULL) {
2808 return (ENOMEM);
2809 }
2810 if (fdp->fd_nfiles >= numfiles) {
2811 FREE_ZONE(newofiles, numfiles * OFILESIZE, M_OFILETABL);
2812 continue;
2813 }
2814 newofileflags = (char *) &newofiles[numfiles];
2815 /*
2816 * Copy the existing ofile and ofileflags arrays
2817 * and zero the new portion of each array.
2818 */
2819 oldnfiles = fdp->fd_nfiles;
2820 (void) memcpy(newofiles, fdp->fd_ofiles,
2821 oldnfiles * sizeof(*fdp->fd_ofiles));
2822 (void) memset(&newofiles[oldnfiles], 0,
2823 (numfiles - oldnfiles) * sizeof(*fdp->fd_ofiles));
2824
2825 (void) memcpy(newofileflags, fdp->fd_ofileflags,
2826 oldnfiles * sizeof(*fdp->fd_ofileflags));
2827 (void) memset(&newofileflags[oldnfiles], 0,
2828 (numfiles - oldnfiles) *
2829 sizeof(*fdp->fd_ofileflags));
2830 ofiles = fdp->fd_ofiles;
2831 fdp->fd_ofiles = newofiles;
2832 fdp->fd_ofileflags = newofileflags;
2833 fdp->fd_nfiles = numfiles;
2834 FREE_ZONE(ofiles, oldnfiles * OFILESIZE, M_OFILETABL);
2835 fdexpand++;
2836 }
2837 }
2838
2839
2840 /*
2841 * fdavail
2842 *
2843 * Description: Check to see whether n user file descriptors are available
2844 * to the process p.
2845 *
2846 * Parameters: p Process to check in
2847 * n The number of fd's desired
2848 *
2849 * Returns: 0 No
2850 * 1 Yes
2851 *
2852 * Locks: Assumes proc_fdlock for process is held by the caller
2853 *
2854 * Notes: The answer only remains valid so long as the proc_fdlock is
2855 * held by the caller.
2856 */
2857 int
2858 fdavail(proc_t p, int n)
2859 {
2860 struct filedesc *fdp = p->p_fd;
2861 struct fileproc **fpp;
2862 char *flags;
2863 int i, lim;
2864
2865 lim = min((int)p->p_rlimit[RLIMIT_NOFILE].rlim_cur, maxfiles);
2866 if ((i = lim - fdp->fd_nfiles) > 0 && (n -= i) <= 0)
2867 return (1);
2868 fpp = &fdp->fd_ofiles[fdp->fd_freefile];
2869 flags = &fdp->fd_ofileflags[fdp->fd_freefile];
2870 for (i = fdp->fd_nfiles - fdp->fd_freefile; --i >= 0; fpp++, flags++)
2871 if (*fpp == NULL && !(*flags & UF_RESERVED) && --n <= 0)
2872 return (1);
2873 return (0);
2874 }
2875
2876
2877 /*
2878 * fdrelse
2879 *
2880 * Description: Legacy KPI wrapper function for _fdrelse
2881 *
2882 * Parameters: p Process in which fd lives
2883 * fd fd to free
2884 *
2885 * Returns: void
2886 *
2887 * Locks: Assumes proc_fdlock for process is held by the caller
2888 */
2889 void
2890 fdrelse(proc_t p, int fd)
2891 {
2892 _fdrelse(p, fd);
2893 }
2894
2895
2896 /*
2897 * fdgetf_noref
2898 *
2899 * Description: Get the fileproc pointer for the given fd from the per process
2900 * open file table without taking an explicit reference on it.
2901 *
2902 * Parameters: p Process containing fd
2903 * fd fd to obtain fileproc for
2904 * resultfp Pointer to pointer return area
2905 *
2906 * Returns: 0 Success
2907 * EBADF
2908 *
2909 * Implicit returns:
2910 * *resultfp (modified) Pointer to fileproc pointer
2911 *
2912 * Locks: Assumes proc_fdlock for process is held by the caller
2913 *
2914 * Notes: Because there is no reference explicitly taken, the returned
2915 * fileproc pointer is only valid so long as the proc_fdlock
2916 * remains held by the caller.
2917 */
2918 int
2919 fdgetf_noref(proc_t p, int fd, struct fileproc **resultfp)
2920 {
2921 struct filedesc *fdp = p->p_fd;
2922 struct fileproc *fp;
2923
2924 if (fd < 0 || fd >= fdp->fd_nfiles ||
2925 (fp = fdp->fd_ofiles[fd]) == NULL ||
2926 (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
2927 return (EBADF);
2928 }
2929 if (resultfp)
2930 *resultfp = fp;
2931 return (0);
2932 }
2933
2934
2935 /*
2936 * fp_getfvp
2937 *
2938 * Description: Get fileproc and vnode pointer for a given fd from the per
2939 * process open file table of the specified process, and if
2940 * successful, increment the f_iocount
2941 *
2942 * Parameters: p Process in which fd lives
2943 * fd fd to get information for
2944 * resultfp Pointer to result fileproc
2945 * pointer area, or 0 if none
2946 * resultvp Pointer to result vnode pointer
2947 * area, or 0 if none
2948 *
2949 * Returns: 0 Success
2950 * EBADF Bad file descriptor
2951 * ENOTSUP fd does not refer to a vnode
2952 *
2953 * Implicit returns:
2954 * *resultfp (modified) Fileproc pointer
2955 * *resultvp (modified) vnode pointer
2956 *
2957 * Notes: The resultfp and resultvp fields are optional, and may be
2958 * independently specified as NULL to skip returning information
2959 *
2960 * Locks: Internally takes and releases proc_fdlock
2961 */
2962 int
2963 fp_getfvp(proc_t p, int fd, struct fileproc **resultfp, struct vnode **resultvp)
2964 {
2965 struct filedesc *fdp = p->p_fd;
2966 struct fileproc *fp;
2967
2968 proc_fdlock_spin(p);
2969 if (fd < 0 || fd >= fdp->fd_nfiles ||
2970 (fp = fdp->fd_ofiles[fd]) == NULL ||
2971 (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
2972 proc_fdunlock(p);
2973 return (EBADF);
2974 }
2975 if (fp->f_type != DTYPE_VNODE) {
2976 proc_fdunlock(p);
2977 return(ENOTSUP);
2978 }
2979 fp->f_iocount++;
2980
2981 if (resultfp)
2982 *resultfp = fp;
2983 if (resultvp)
2984 *resultvp = (struct vnode *)fp->f_data;
2985 proc_fdunlock(p);
2986
2987 return (0);
2988 }
2989
2990
2991 /*
2992 * fp_getfvpandvid
2993 *
2994 * Description: Get fileproc, vnode pointer, and vid for a given fd from the
2995 * per process open file table of the specified process, and if
2996 * successful, increment the f_iocount
2997 *
2998 * Parameters: p Process in which fd lives
2999 * fd fd to get information for
3000 * resultfp Pointer to result fileproc
3001 * pointer area, or 0 if none
3002 * resultvp Pointer to result vnode pointer
3003 * area, or 0 if none
3004 * vidp Pointer to resuld vid area
3005 *
3006 * Returns: 0 Success
3007 * EBADF Bad file descriptor
3008 * ENOTSUP fd does not refer to a vnode
3009 *
3010 * Implicit returns:
3011 * *resultfp (modified) Fileproc pointer
3012 * *resultvp (modified) vnode pointer
3013 * *vidp vid value
3014 *
3015 * Notes: The resultfp and resultvp fields are optional, and may be
3016 * independently specified as NULL to skip returning information
3017 *
3018 * Locks: Internally takes and releases proc_fdlock
3019 */
3020 int
3021 fp_getfvpandvid(proc_t p, int fd, struct fileproc **resultfp,
3022 struct vnode **resultvp, uint32_t *vidp)
3023 {
3024 struct filedesc *fdp = p->p_fd;
3025 struct fileproc *fp;
3026
3027 proc_fdlock_spin(p);
3028 if (fd < 0 || fd >= fdp->fd_nfiles ||
3029 (fp = fdp->fd_ofiles[fd]) == NULL ||
3030 (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
3031 proc_fdunlock(p);
3032 return (EBADF);
3033 }
3034 if (fp->f_type != DTYPE_VNODE) {
3035 proc_fdunlock(p);
3036 return(ENOTSUP);
3037 }
3038 fp->f_iocount++;
3039
3040 if (resultfp)
3041 *resultfp = fp;
3042 if (resultvp)
3043 *resultvp = (struct vnode *)fp->f_data;
3044 if (vidp)
3045 *vidp = (uint32_t)vnode_vid((struct vnode *)fp->f_data);
3046 proc_fdunlock(p);
3047
3048 return (0);
3049 }
3050
3051
3052 /*
3053 * fp_getfsock
3054 *
3055 * Description: Get fileproc and socket pointer for a given fd from the
3056 * per process open file table of the specified process, and if
3057 * successful, increment the f_iocount
3058 *
3059 * Parameters: p Process in which fd lives
3060 * fd fd to get information for
3061 * resultfp Pointer to result fileproc
3062 * pointer area, or 0 if none
3063 * results Pointer to result socket
3064 * pointer area, or 0 if none
3065 *
3066 * Returns: EBADF The file descriptor is invalid
3067 * EOPNOTSUPP The file descriptor is not a socket
3068 * 0 Success
3069 *
3070 * Implicit returns:
3071 * *resultfp (modified) Fileproc pointer
3072 * *results (modified) socket pointer
3073 *
3074 * Notes: EOPNOTSUPP should probably be ENOTSOCK; this function is only
3075 * ever called from accept1().
3076 */
3077 int
3078 fp_getfsock(proc_t p, int fd, struct fileproc **resultfp,
3079 struct socket **results)
3080 {
3081 struct filedesc *fdp = p->p_fd;
3082 struct fileproc *fp;
3083
3084 proc_fdlock_spin(p);
3085 if (fd < 0 || fd >= fdp->fd_nfiles ||
3086 (fp = fdp->fd_ofiles[fd]) == NULL ||
3087 (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
3088 proc_fdunlock(p);
3089 return (EBADF);
3090 }
3091 if (fp->f_type != DTYPE_SOCKET) {
3092 proc_fdunlock(p);
3093 return(EOPNOTSUPP);
3094 }
3095 fp->f_iocount++;
3096
3097 if (resultfp)
3098 *resultfp = fp;
3099 if (results)
3100 *results = (struct socket *)fp->f_data;
3101 proc_fdunlock(p);
3102
3103 return (0);
3104 }
3105
3106
3107 /*
3108 * fp_getfkq
3109 *
3110 * Description: Get fileproc and kqueue pointer for a given fd from the
3111 * per process open file table of the specified process, and if
3112 * successful, increment the f_iocount
3113 *
3114 * Parameters: p Process in which fd lives
3115 * fd fd to get information for
3116 * resultfp Pointer to result fileproc
3117 * pointer area, or 0 if none
3118 * resultkq Pointer to result kqueue
3119 * pointer area, or 0 if none
3120 *
3121 * Returns: EBADF The file descriptor is invalid
3122 * EBADF The file descriptor is not a socket
3123 * 0 Success
3124 *
3125 * Implicit returns:
3126 * *resultfp (modified) Fileproc pointer
3127 * *resultkq (modified) kqueue pointer
3128 *
3129 * Notes: The second EBADF should probably be something else to make
3130 * the error condition distinct.
3131 */
3132 int
3133 fp_getfkq(proc_t p, int fd, struct fileproc **resultfp,
3134 struct kqueue **resultkq)
3135 {
3136 struct filedesc *fdp = p->p_fd;
3137 struct fileproc *fp;
3138
3139 proc_fdlock_spin(p);
3140 if ( fd < 0 || fd >= fdp->fd_nfiles ||
3141 (fp = fdp->fd_ofiles[fd]) == NULL ||
3142 (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
3143 proc_fdunlock(p);
3144 return (EBADF);
3145 }
3146 if (fp->f_type != DTYPE_KQUEUE) {
3147 proc_fdunlock(p);
3148 return(EBADF);
3149 }
3150 fp->f_iocount++;
3151
3152 if (resultfp)
3153 *resultfp = fp;
3154 if (resultkq)
3155 *resultkq = (struct kqueue *)fp->f_data;
3156 proc_fdunlock(p);
3157
3158 return (0);
3159 }
3160
3161
3162 /*
3163 * fp_getfpshm
3164 *
3165 * Description: Get fileproc and POSIX shared memory pointer for a given fd
3166 * from the per process open file table of the specified process
3167 * and if successful, increment the f_iocount
3168 *
3169 * Parameters: p Process in which fd lives
3170 * fd fd to get information for
3171 * resultfp Pointer to result fileproc
3172 * pointer area, or 0 if none
3173 * resultpshm Pointer to result POSIX
3174 * shared memory pointer
3175 * pointer area, or 0 if none
3176 *
3177 * Returns: EBADF The file descriptor is invalid
3178 * EBADF The file descriptor is not a POSIX
3179 * shared memory area
3180 * 0 Success
3181 *
3182 * Implicit returns:
3183 * *resultfp (modified) Fileproc pointer
3184 * *resultpshm (modified) POSIX shared memory pointer
3185 *
3186 * Notes: The second EBADF should probably be something else to make
3187 * the error condition distinct.
3188 */
3189 int
3190 fp_getfpshm(proc_t p, int fd, struct fileproc **resultfp,
3191 struct pshmnode **resultpshm)
3192 {
3193 struct filedesc *fdp = p->p_fd;
3194 struct fileproc *fp;
3195
3196 proc_fdlock_spin(p);
3197 if (fd < 0 || fd >= fdp->fd_nfiles ||
3198 (fp = fdp->fd_ofiles[fd]) == NULL ||
3199 (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
3200 proc_fdunlock(p);
3201 return (EBADF);
3202 }
3203 if (fp->f_type != DTYPE_PSXSHM) {
3204
3205 proc_fdunlock(p);
3206 return(EBADF);
3207 }
3208 fp->f_iocount++;
3209
3210 if (resultfp)
3211 *resultfp = fp;
3212 if (resultpshm)
3213 *resultpshm = (struct pshmnode *)fp->f_data;
3214 proc_fdunlock(p);
3215
3216 return (0);
3217 }
3218
3219
3220 /*
3221 * fp_getfsem
3222 *
3223 * Description: Get fileproc and POSIX semaphore pointer for a given fd from
3224 * the per process open file table of the specified process
3225 * and if successful, increment the f_iocount
3226 *
3227 * Parameters: p Process in which fd lives
3228 * fd fd to get information for
3229 * resultfp Pointer to result fileproc
3230 * pointer area, or 0 if none
3231 * resultpsem Pointer to result POSIX
3232 * semaphore pointer area, or
3233 * 0 if none
3234 *
3235 * Returns: EBADF The file descriptor is invalid
3236 * EBADF The file descriptor is not a POSIX
3237 * semaphore
3238 * 0 Success
3239 *
3240 * Implicit returns:
3241 * *resultfp (modified) Fileproc pointer
3242 * *resultpsem (modified) POSIX semaphore pointer
3243 *
3244 * Notes: The second EBADF should probably be something else to make
3245 * the error condition distinct.
3246 *
3247 * In order to support unnamed POSIX semaphores, the named
3248 * POSIX semaphores will have to move out of the per-process
3249 * open filetable, and into a global table that is shared with
3250 * unnamed POSIX semaphores, since unnamed POSIX semaphores
3251 * are typically used by declaring instances in shared memory,
3252 * and there's no other way to do this without changing the
3253 * underlying type, which would introduce binary compatibility
3254 * issues.
3255 */
3256 int
3257 fp_getfpsem(proc_t p, int fd, struct fileproc **resultfp,
3258 struct psemnode **resultpsem)
3259 {
3260 struct filedesc *fdp = p->p_fd;
3261 struct fileproc *fp;
3262
3263 proc_fdlock_spin(p);
3264 if (fd < 0 || fd >= fdp->fd_nfiles ||
3265 (fp = fdp->fd_ofiles[fd]) == NULL ||
3266 (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
3267 proc_fdunlock(p);
3268 return (EBADF);
3269 }
3270 if (fp->f_type != DTYPE_PSXSEM) {
3271 proc_fdunlock(p);
3272 return(EBADF);
3273 }
3274 fp->f_iocount++;
3275
3276 if (resultfp)
3277 *resultfp = fp;
3278 if (resultpsem)
3279 *resultpsem = (struct psemnode *)fp->f_data;
3280 proc_fdunlock(p);
3281
3282 return (0);
3283 }
3284
3285
3286 /*
3287 * fp_getfpipe
3288 *
3289 * Description: Get fileproc and pipe pointer for a given fd from the
3290 * per process open file table of the specified process
3291 * and if successful, increment the f_iocount
3292 *
3293 * Parameters: p Process in which fd lives
3294 * fd fd to get information for
3295 * resultfp Pointer to result fileproc
3296 * pointer area, or 0 if none
3297 * resultpipe Pointer to result pipe
3298 * pointer area, or 0 if none
3299 *
3300 * Returns: EBADF The file descriptor is invalid
3301 * EBADF The file descriptor is not a socket
3302 * 0 Success
3303 *
3304 * Implicit returns:
3305 * *resultfp (modified) Fileproc pointer
3306 * *resultpipe (modified) pipe pointer
3307 *
3308 * Notes: The second EBADF should probably be something else to make
3309 * the error condition distinct.
3310 */
3311 int
3312 fp_getfpipe(proc_t p, int fd, struct fileproc **resultfp,
3313 struct pipe **resultpipe)
3314 {
3315 struct filedesc *fdp = p->p_fd;
3316 struct fileproc *fp;
3317
3318 proc_fdlock_spin(p);
3319 if (fd < 0 || fd >= fdp->fd_nfiles ||
3320 (fp = fdp->fd_ofiles[fd]) == NULL ||
3321 (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
3322 proc_fdunlock(p);
3323 return (EBADF);
3324 }
3325 if (fp->f_type != DTYPE_PIPE) {
3326 proc_fdunlock(p);
3327 return(EBADF);
3328 }
3329 fp->f_iocount++;
3330
3331 if (resultfp)
3332 *resultfp = fp;
3333 if (resultpipe)
3334 *resultpipe = (struct pipe *)fp->f_data;
3335 proc_fdunlock(p);
3336
3337 return (0);
3338 }
3339
3340 #if NETAT
3341 #define DTYPE_ATALK -1 /* XXX This does not belong here */
3342
3343
3344 /*
3345 * fp_getfatalk
3346 *
3347 * Description: Get fileproc and atalk pointer for a given fd from the
3348 * per process open file table of the specified process
3349 * and if successful, increment the f_iocount
3350 *
3351 * Parameters: p Process in which fd lives
3352 * fd fd to get information for
3353 * resultfp Pointer to result fileproc
3354 * pointer area, or 0 if none
3355 * resultatalk Pointer to result atalk
3356 * pointer area, or 0 if none
3357 * Returns: EBADF The file descriptor is invalid
3358 * EBADF The file descriptor is not a socket
3359 * 0 Success
3360 *
3361 * Implicit returns:
3362 * *resultfp (modified) Fileproc pointer
3363 * *resultatalk (modified) atalk pointer
3364 *
3365 * Notes: The second EBADF should probably be something else to make
3366 * the error condition distinct.
3367 *
3368 * XXX This code is specific to AppleTalk protocol support, and
3369 * XXX should be conditionally compiled
3370 */
3371 int
3372 fp_getfatalk(proc_t p, int fd, struct fileproc **resultfp,
3373 struct atalk **resultatalk)
3374 {
3375 struct filedesc *fdp = p->p_fd;
3376 struct fileproc *fp;
3377
3378 proc_fdlock_spin(p);
3379 if (fd < 0 || fd >= fdp->fd_nfiles ||
3380 (fp = fdp->fd_ofiles[fd]) == NULL ||
3381 (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
3382 proc_fdunlock(p);
3383 return (EBADF);
3384 }
3385 if (fp->f_type != (DTYPE_ATALK+1)) {
3386 proc_fdunlock(p);
3387 return(EBADF);
3388 }
3389 fp->f_iocount++;
3390
3391 if (resultfp)
3392 *resultfp = fp;
3393 if (resultatalk)
3394 *resultatalk = (struct atalk *)fp->f_data;
3395 proc_fdunlock(p);
3396
3397 return (0);
3398 }
3399
3400 #endif /* NETAT */
3401
3402 /*
3403 * fp_lookup
3404 *
3405 * Description: Get fileproc pointer for a given fd from the per process
3406 * open file table of the specified process and if successful,
3407 * increment the f_iocount
3408 *
3409 * Parameters: p Process in which fd lives
3410 * fd fd to get information for
3411 * resultfp Pointer to result fileproc
3412 * pointer area, or 0 if none
3413 * locked !0 if the caller holds the
3414 * proc_fdlock, 0 otherwise
3415 *
3416 * Returns: 0 Success
3417 * EBADF Bad file descriptor
3418 *
3419 * Implicit returns:
3420 * *resultfp (modified) Fileproc pointer
3421 *
3422 * Locks: If the argument 'locked' is non-zero, then the caller is
3423 * expected to have taken and held the proc_fdlock; if it is
3424 * zero, than this routine internally takes and drops this lock.
3425 */
3426 int
3427 fp_lookup(proc_t p, int fd, struct fileproc **resultfp, int locked)
3428 {
3429 struct filedesc *fdp = p->p_fd;
3430 struct fileproc *fp;
3431
3432 if (!locked)
3433 proc_fdlock_spin(p);
3434 if (fd < 0 || fdp == NULL || fd >= fdp->fd_nfiles ||
3435 (fp = fdp->fd_ofiles[fd]) == NULL ||
3436 (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
3437 if (!locked)
3438 proc_fdunlock(p);
3439 return (EBADF);
3440 }
3441 fp->f_iocount++;
3442
3443 if (resultfp)
3444 *resultfp = fp;
3445 if (!locked)
3446 proc_fdunlock(p);
3447
3448 return (0);
3449 }
3450
3451
3452 /*
3453 * fp_drop_written
3454 *
3455 * Description: Set the FP_WRITTEN flag on the fileproc and drop the I/O
3456 * reference previously taken by calling fp_lookup et. al.
3457 *
3458 * Parameters: p Process in which the fd lives
3459 * fd fd associated with the fileproc
3460 * fp fileproc on which to set the
3461 * flag and drop the reference
3462 *
3463 * Returns: 0 Success
3464 * fp_drop:EBADF Bad file descriptor
3465 *
3466 * Locks: This function internally takes and drops the proc_fdlock for
3467 * the supplied process
3468 *
3469 * Notes: The fileproc must correspond to the fd in the supplied proc
3470 */
3471 int
3472 fp_drop_written(proc_t p, int fd, struct fileproc *fp)
3473 {
3474 int error;
3475
3476 proc_fdlock_spin(p);
3477
3478 fp->f_flags |= FP_WRITTEN;
3479
3480 error = fp_drop(p, fd, fp, 1);
3481
3482 proc_fdunlock(p);
3483
3484 return (error);
3485 }
3486
3487
3488 /*
3489 * fp_drop_event
3490 *
3491 * Description: Set the FP_WAITEVENT flag on the fileproc and drop the I/O
3492 * reference previously taken by calling fp_lookup et. al.
3493 *
3494 * Parameters: p Process in which the fd lives
3495 * fd fd associated with the fileproc
3496 * fp fileproc on which to set the
3497 * flag and drop the reference
3498 *
3499 * Returns: 0 Success
3500 * fp_drop:EBADF Bad file descriptor
3501 *
3502 * Locks: This function internally takes and drops the proc_fdlock for
3503 * the supplied process
3504 *
3505 * Notes: The fileproc must correspond to the fd in the supplied proc
3506 */
3507 int
3508 fp_drop_event(proc_t p, int fd, struct fileproc *fp)
3509 {
3510 int error;
3511
3512 proc_fdlock_spin(p);
3513
3514 fp->f_flags |= FP_WAITEVENT;
3515
3516 error = fp_drop(p, fd, fp, 1);
3517
3518 proc_fdunlock(p);
3519
3520 return (error);
3521 }
3522
3523
3524 /*
3525 * fp_drop
3526 *
3527 * Description: Drop the I/O reference previously taken by calling fp_lookup
3528 * et. al.
3529 *
3530 * Parameters: p Process in which the fd lives
3531 * fd fd associated with the fileproc
3532 * fp fileproc on which to set the
3533 * flag and drop the reference
3534 * locked flag to internally take and
3535 * drop proc_fdlock if it is not
3536 * already held by the caller
3537 *
3538 * Returns: 0 Success
3539 * EBADF Bad file descriptor
3540 *
3541 * Locks: This function internally takes and drops the proc_fdlock for
3542 * the supplied process if 'locked' is non-zero, and assumes that
3543 * the caller already holds this lock if 'locked' is non-zero.
3544 *
3545 * Notes: The fileproc must correspond to the fd in the supplied proc
3546 */
3547 int
3548 fp_drop(proc_t p, int fd, struct fileproc *fp, int locked)
3549 {
3550 struct filedesc *fdp = p->p_fd;
3551 int needwakeup = 0;
3552
3553 if (!locked)
3554 proc_fdlock_spin(p);
3555 if ((fp == FILEPROC_NULL) && (fd < 0 || fd >= fdp->fd_nfiles ||
3556 (fp = fdp->fd_ofiles[fd]) == NULL ||
3557 ((fdp->fd_ofileflags[fd] & UF_RESERVED) &&
3558 !(fdp->fd_ofileflags[fd] & UF_CLOSING)))) {
3559 if (!locked)
3560 proc_fdunlock(p);
3561 return (EBADF);
3562 }
3563 fp->f_iocount--;
3564
3565 if (fp->f_iocount == 0) {
3566 if (fp->f_flags & FP_SELCONFLICT)
3567 fp->f_flags &= ~FP_SELCONFLICT;
3568
3569 if (p->p_fpdrainwait) {
3570 p->p_fpdrainwait = 0;
3571 needwakeup = 1;
3572 }
3573 }
3574 if (!locked)
3575 proc_fdunlock(p);
3576 if (needwakeup)
3577 wakeup(&p->p_fpdrainwait);
3578
3579 return (0);
3580 }
3581
3582
3583 /*
3584 * file_vnode
3585 *
3586 * Description: Given an fd, look it up in the current process's per process
3587 * open file table, and return its internal vnode pointer.
3588 *
3589 * Parameters: fd fd to obtain vnode from
3590 * vpp pointer to vnode return area
3591 *
3592 * Returns: 0 Success
3593 * EINVAL The fd does not refer to a
3594 * vnode fileproc entry
3595 * fp_lookup:EBADF Bad file descriptor
3596 *
3597 * Implicit returns:
3598 * *vpp (modified) Returned vnode pointer
3599 *
3600 * Locks: This function internally takes and drops the proc_fdlock for
3601 * the current process
3602 *
3603 * Notes: If successful, this function increments the f_iocount on the
3604 * fd's corresponding fileproc.
3605 *
3606 * The fileproc referenced is not returned; because of this, care
3607 * must be taken to not drop the last reference (e.g. by closing
3608 * the file). This is inherently unsafe, since the reference may
3609 * not be recoverable from the vnode, if there is a subsequent
3610 * close that destroys the associate fileproc. The caller should
3611 * therefore retain their own reference on the fileproc so that
3612 * the f_iocount can be dropped subsequently. Failure to do this
3613 * can result in the returned pointer immediately becoming invalid
3614 * following the call.
3615 *
3616 * Use of this function is discouraged.
3617 */
3618 int
3619 file_vnode(int fd, struct vnode **vpp)
3620 {
3621 proc_t p = current_proc();
3622 struct fileproc *fp;
3623 int error;
3624
3625 proc_fdlock_spin(p);
3626 if ( (error = fp_lookup(p, fd, &fp, 1)) ) {
3627 proc_fdunlock(p);
3628 return(error);
3629 }
3630 if (fp->f_type != DTYPE_VNODE) {
3631 fp_drop(p, fd, fp,1);
3632 proc_fdunlock(p);
3633 return(EINVAL);
3634 }
3635 if (vpp != NULL)
3636 *vpp = (struct vnode *)fp->f_data;
3637 proc_fdunlock(p);
3638
3639 return(0);
3640 }
3641
3642
3643 /*
3644 * file_vnode_withvid
3645 *
3646 * Description: Given an fd, look it up in the current process's per process
3647 * open file table, and return its internal vnode pointer.
3648 *
3649 * Parameters: fd fd to obtain vnode from
3650 * vpp pointer to vnode return area
3651 * vidp pointer to vid of the returned vnode
3652 *
3653 * Returns: 0 Success
3654 * EINVAL The fd does not refer to a
3655 * vnode fileproc entry
3656 * fp_lookup:EBADF Bad file descriptor
3657 *
3658 * Implicit returns:
3659 * *vpp (modified) Returned vnode pointer
3660 *
3661 * Locks: This function internally takes and drops the proc_fdlock for
3662 * the current process
3663 *
3664 * Notes: If successful, this function increments the f_iocount on the
3665 * fd's corresponding fileproc.
3666 *
3667 * The fileproc referenced is not returned; because of this, care
3668 * must be taken to not drop the last reference (e.g. by closing
3669 * the file). This is inherently unsafe, since the reference may
3670 * not be recoverable from the vnode, if there is a subsequent
3671 * close that destroys the associate fileproc. The caller should
3672 * therefore retain their own reference on the fileproc so that
3673 * the f_iocount can be dropped subsequently. Failure to do this
3674 * can result in the returned pointer immediately becoming invalid
3675 * following the call.
3676 *
3677 * Use of this function is discouraged.
3678 */
3679 int
3680 file_vnode_withvid(int fd, struct vnode **vpp, uint32_t * vidp)
3681 {
3682 proc_t p = current_proc();
3683 struct fileproc *fp;
3684 vnode_t vp;
3685 int error;
3686
3687 proc_fdlock_spin(p);
3688 if ( (error = fp_lookup(p, fd, &fp, 1)) ) {
3689 proc_fdunlock(p);
3690 return(error);
3691 }
3692 if (fp->f_type != DTYPE_VNODE) {
3693 fp_drop(p, fd, fp,1);
3694 proc_fdunlock(p);
3695 return(EINVAL);
3696 }
3697 vp = (struct vnode *)fp->f_data;
3698 if (vpp != NULL)
3699 *vpp = vp;
3700
3701 if ((vidp != NULL) && (vp != NULLVP))
3702 *vidp = (uint32_t)vp->v_id;
3703
3704 proc_fdunlock(p);
3705
3706 return(0);
3707 }
3708
3709
3710 /*
3711 * file_socket
3712 *
3713 * Description: Given an fd, look it up in the current process's per process
3714 * open file table, and return its internal socket pointer.
3715 *
3716 * Parameters: fd fd to obtain vnode from
3717 * sp pointer to socket return area
3718 *
3719 * Returns: 0 Success
3720 * ENOTSOCK Not a socket
3721 * fp_lookup:EBADF Bad file descriptor
3722 *
3723 * Implicit returns:
3724 * *sp (modified) Returned socket pointer
3725 *
3726 * Locks: This function internally takes and drops the proc_fdlock for
3727 * the current process
3728 *
3729 * Notes: If successful, this function increments the f_iocount on the
3730 * fd's corresponding fileproc.
3731 *
3732 * The fileproc referenced is not returned; because of this, care
3733 * must be taken to not drop the last reference (e.g. by closing
3734 * the file). This is inherently unsafe, since the reference may
3735 * not be recoverable from the socket, if there is a subsequent
3736 * close that destroys the associate fileproc. The caller should
3737 * therefore retain their own reference on the fileproc so that
3738 * the f_iocount can be dropped subsequently. Failure to do this
3739 * can result in the returned pointer immediately becoming invalid
3740 * following the call.
3741 *
3742 * Use of this function is discouraged.
3743 */
3744 int
3745 file_socket(int fd, struct socket **sp)
3746 {
3747 proc_t p = current_proc();
3748 struct fileproc *fp;
3749 int error;
3750
3751 proc_fdlock_spin(p);
3752 if ( (error = fp_lookup(p, fd, &fp, 1)) ) {
3753 proc_fdunlock(p);
3754 return(error);
3755 }
3756 if (fp->f_type != DTYPE_SOCKET) {
3757 fp_drop(p, fd, fp,1);
3758 proc_fdunlock(p);
3759 return(ENOTSOCK);
3760 }
3761 *sp = (struct socket *)fp->f_data;
3762 proc_fdunlock(p);
3763
3764 return(0);
3765 }
3766
3767
3768 /*
3769 * file_flags
3770 *
3771 * Description: Given an fd, look it up in the current process's per process
3772 * open file table, and return its fileproc's flags field.
3773 *
3774 * Parameters: fd fd whose flags are to be
3775 * retrieved
3776 * flags pointer to flags data area
3777 *
3778 * Returns: 0 Success
3779 * ENOTSOCK Not a socket
3780 * fp_lookup:EBADF Bad file descriptor
3781 *
3782 * Implicit returns:
3783 * *flags (modified) Returned flags field
3784 *
3785 * Locks: This function internally takes and drops the proc_fdlock for
3786 * the current process
3787 *
3788 * Notes: This function will internally increment and decrement the
3789 * f_iocount of the fileproc as part of its operation.
3790 */
3791 int
3792 file_flags(int fd, int *flags)
3793 {
3794
3795 proc_t p = current_proc();
3796 struct fileproc *fp;
3797 int error;
3798
3799 proc_fdlock_spin(p);
3800 if ( (error = fp_lookup(p, fd, &fp, 1)) ) {
3801 proc_fdunlock(p);
3802 return(error);
3803 }
3804 *flags = (int)fp->f_flag;
3805 fp_drop(p, fd, fp,1);
3806 proc_fdunlock(p);
3807
3808 return(0);
3809 }
3810
3811
3812 /*
3813 * file_drop
3814 *
3815 * Description: Drop an iocount reference on an fd, and wake up any waiters
3816 * for draining (i.e. blocked in fileproc_drain() called during
3817 * the last attempt to close a file).
3818 *
3819 * Parameters: fd fd on which an ioreference is
3820 * to be dropped
3821 *
3822 * Returns: 0 Success
3823 * EBADF Bad file descriptor
3824 *
3825 * Description: Given an fd, look it up in the current process's per process
3826 * open file table, and drop it's fileproc's f_iocount by one
3827 *
3828 * Notes: This is intended as a corresponding operation to the functions
3829 * file_vnode() and file_socket() operations.
3830 *
3831 * Technically, the close reference is supposed to be protected
3832 * by a fileproc_drain(), however, a drain will only block if
3833 * the fd refers to a character device, and that device has had
3834 * preparefileread() called on it. If it refers to something
3835 * other than a character device, then the drain will occur and
3836 * block each close attempt, rather than merely the last close.
3837 *
3838 * Since it's possible for an fd that refers to a character
3839 * device to have an intermediate close followed by an open to
3840 * cause a different file to correspond to that descriptor,
3841 * unless there was a cautionary reference taken on the fileproc,
3842 * this is an inherently unsafe function. This happens in the
3843 * case where multiple fd's in a process refer to the same
3844 * character device (e.g. stdin/out/err pointing to a tty, etc.).
3845 *
3846 * Use of this function is discouraged.
3847 */
3848 int
3849 file_drop(int fd)
3850 {
3851 struct fileproc *fp;
3852 proc_t p = current_proc();
3853 int needwakeup = 0;
3854
3855 proc_fdlock_spin(p);
3856 if (fd < 0 || fd >= p->p_fd->fd_nfiles ||
3857 (fp = p->p_fd->fd_ofiles[fd]) == NULL ||
3858 ((p->p_fd->fd_ofileflags[fd] & UF_RESERVED) &&
3859 !(p->p_fd->fd_ofileflags[fd] & UF_CLOSING))) {
3860 proc_fdunlock(p);
3861 return (EBADF);
3862 }
3863 fp->f_iocount --;
3864
3865 if (fp->f_iocount == 0) {
3866 if (fp->f_flags & FP_SELCONFLICT)
3867 fp->f_flags &= ~FP_SELCONFLICT;
3868
3869 if (p->p_fpdrainwait) {
3870 p->p_fpdrainwait = 0;
3871 needwakeup = 1;
3872 }
3873 }
3874 proc_fdunlock(p);
3875
3876 if (needwakeup)
3877 wakeup(&p->p_fpdrainwait);
3878 return(0);
3879 }
3880
3881
3882 /*
3883 * falloc
3884 *
3885 * Description: Allocate an entry in the per process open file table and
3886 * return the corresponding fileproc and fd.
3887 *
3888 * Parameters: p The process in whose open file
3889 * table the fd is to be allocated
3890 * resultfp Pointer to fileproc pointer
3891 * return area
3892 * resultfd Pointer to fd return area
3893 * ctx VFS context
3894 *
3895 * Returns: 0 Success
3896 * falloc:ENFILE Too many open files in system
3897 * falloc:EMFILE Too many open files in process
3898 * falloc:ENOMEM M_FILEPROC or M_FILEGLOB zone
3899 * exhausted
3900 *
3901 * Implicit returns:
3902 * *resultfd (modified) Returned fileproc pointer
3903 * *resultfd (modified) Returned fd
3904 *
3905 * Locks: This function takes and drops the proc_fdlock; if this lock
3906 * is already held, use falloc_locked() instead.
3907 *
3908 * Notes: This function takes separate process and context arguments
3909 * solely to support kern_exec.c; otherwise, it would take
3910 * neither, and expect falloc_locked() to use the
3911 * vfs_context_current() routine internally.
3912 */
3913 int
3914 falloc(proc_t p, struct fileproc **resultfp, int *resultfd, vfs_context_t ctx)
3915 {
3916 int error;
3917
3918 proc_fdlock(p);
3919 error = falloc_locked(p, resultfp, resultfd, ctx, 1);
3920 proc_fdunlock(p);
3921
3922 return(error);
3923 }
3924
3925
3926 /*
3927 * falloc_locked
3928 *
3929 * Create a new open file structure and allocate
3930 * a file descriptor for the process that refers to it.
3931 *
3932 * Returns: 0 Success
3933 *
3934 * Description: Allocate an entry in the per process open file table and
3935 * return the corresponding fileproc and fd.
3936 *
3937 * Parameters: p The process in whose open file
3938 * table the fd is to be allocated
3939 * resultfp Pointer to fileproc pointer
3940 * return area
3941 * resultfd Pointer to fd return area
3942 * ctx VFS context
3943 * locked Flag to indicate whether the
3944 * caller holds proc_fdlock
3945 *
3946 * Returns: 0 Success
3947 * ENFILE Too many open files in system
3948 * fdalloc:EMFILE Too many open files in process
3949 * ENOMEM M_FILEPROC or M_FILEGLOB zone
3950 * exhausted
3951 * fdalloc:ENOMEM
3952 *
3953 * Implicit returns:
3954 * *resultfd (modified) Returned fileproc pointer
3955 * *resultfd (modified) Returned fd
3956 *
3957 * Locks: If the parameter 'locked' is zero, this function takes and
3958 * drops the proc_fdlock; if non-zero, the caller must hold the
3959 * lock.
3960 *
3961 * Notes: If you intend to use a non-zero 'locked' parameter, use the
3962 * utility function falloc() instead.
3963 *
3964 * This function takes separate process and context arguments
3965 * solely to support kern_exec.c; otherwise, it would take
3966 * neither, and use the vfs_context_current() routine internally.
3967 */
3968 int
3969 falloc_locked(proc_t p, struct fileproc **resultfp, int *resultfd,
3970 vfs_context_t ctx, int locked)
3971 {
3972 struct fileproc *fp;
3973 struct fileglob *fg;
3974 int error, nfd;
3975
3976 if (!locked)
3977 proc_fdlock(p);
3978 if ( (error = fdalloc(p, 0, &nfd)) ) {
3979 if (!locked)
3980 proc_fdunlock(p);
3981 return (error);
3982 }
3983 if (nfiles >= maxfiles) {
3984 if (!locked)
3985 proc_fdunlock(p);
3986 tablefull("file");
3987 return (ENFILE);
3988 }
3989 #if CONFIG_MACF
3990 error = mac_file_check_create(proc_ucred(p));
3991 if (error) {
3992 if (!locked)
3993 proc_fdunlock(p);
3994 return (error);
3995 }
3996 #endif
3997
3998 /*
3999 * Allocate a new file descriptor.
4000 * If the process has file descriptor zero open, add to the list
4001 * of open files at that point, otherwise put it at the front of
4002 * the list of open files.
4003 */
4004 proc_fdunlock(p);
4005
4006 MALLOC_ZONE(fp, struct fileproc *, sizeof(struct fileproc), M_FILEPROC, M_WAITOK);
4007 if (fp == NULL) {
4008 if (locked)
4009 proc_fdlock(p);
4010 return (ENOMEM);
4011 }
4012 MALLOC_ZONE(fg, struct fileglob *, sizeof(struct fileglob), M_FILEGLOB, M_WAITOK);
4013 if (fg == NULL) {
4014 FREE_ZONE(fp, sizeof(*fp), M_FILEPROC);
4015 if (locked)
4016 proc_fdlock(p);
4017 return (ENOMEM);
4018 }
4019 bzero(fp, sizeof(struct fileproc));
4020 bzero(fg, sizeof(struct fileglob));
4021 lck_mtx_init(&fg->fg_lock, file_lck_grp, file_lck_attr);
4022
4023 fp->f_iocount = 1;
4024 fg->fg_count = 1;
4025 fp->f_fglob = fg;
4026 #if CONFIG_MACF
4027 mac_file_label_init(fg);
4028 #endif
4029
4030 kauth_cred_ref(ctx->vc_ucred);
4031
4032 proc_fdlock(p);
4033
4034 fp->f_cred = ctx->vc_ucred;
4035
4036 #if CONFIG_MACF
4037 mac_file_label_associate(fp->f_cred, fg);
4038 #endif
4039
4040 OSAddAtomic(1, &nfiles);
4041
4042 p->p_fd->fd_ofiles[nfd] = fp;
4043
4044 if (!locked)
4045 proc_fdunlock(p);
4046
4047 if (resultfp)
4048 *resultfp = fp;
4049 if (resultfd)
4050 *resultfd = nfd;
4051
4052 return (0);
4053 }
4054
4055
4056 /*
4057 * fg_free
4058 *
4059 * Description: Free a file structure; drop the global open file count, and
4060 * drop the credential reference, if the fileglob has one, and
4061 * destroy the instance mutex before freeing
4062 *
4063 * Parameters: fg Pointer to fileglob to be
4064 * freed
4065 *
4066 * Returns: void
4067 */
4068 void
4069 fg_free(struct fileglob *fg)
4070 {
4071 OSAddAtomic(-1, &nfiles);
4072
4073 if (IS_VALID_CRED(fg->fg_cred)) {
4074 kauth_cred_unref(&fg->fg_cred);
4075 }
4076 lck_mtx_destroy(&fg->fg_lock, file_lck_grp);
4077
4078 #if CONFIG_MACF
4079 mac_file_label_destroy(fg);
4080 #endif
4081 FREE_ZONE(fg, sizeof *fg, M_FILEGLOB);
4082 }
4083
4084
4085 /*
4086 * fdexec
4087 *
4088 * Description: Perform close-on-exec processing for all files in a process
4089 * that are either marked as close-on-exec, or which were in the
4090 * process of being opened at the time of the execve
4091 *
4092 * Also handles the case (via posix_spawn()) where -all-
4093 * files except those marked with "inherit" as treated as
4094 * close-on-exec.
4095 *
4096 * Parameters: p Pointer to process calling
4097 * execve
4098 *
4099 * Returns: void
4100 *
4101 * Locks: This function internally takes and drops proc_fdlock()
4102 *
4103 * Notes: This function drops and retakes the kernel funnel; this is
4104 * inherently unsafe, since another thread may have the
4105 * proc_fdlock.
4106 *
4107 * XXX: We should likely reverse the lock and funnel drop/acquire
4108 * order to avoid the small race window; it's also possible that
4109 * if the program doing the exec has an outstanding listen socket
4110 * and a network connection is completed asynchronously that we
4111 * will end up with a "ghost" socket reference in the new process.
4112 *
4113 * This needs reworking to make it safe to remove the funnel from
4114 * the execve and posix_spawn system calls.
4115 */
4116 void
4117 fdexec(proc_t p, short flags)
4118 {
4119 struct filedesc *fdp = p->p_fd;
4120 int i;
4121 boolean_t cloexec_default = (flags & POSIX_SPAWN_CLOEXEC_DEFAULT) != 0;
4122
4123 proc_fdlock(p);
4124 for (i = fdp->fd_lastfile; i >= 0; i--) {
4125
4126 struct fileproc *fp = fdp->fd_ofiles[i];
4127 char *flagp = &fdp->fd_ofileflags[i];
4128
4129 if (fp && cloexec_default) {
4130 /*
4131 * Reverse the usual semantics of file descriptor
4132 * inheritance - all of them should be closed
4133 * except files marked explicitly as "inherit" and
4134 * not marked close-on-exec.
4135 */
4136 if ((*flagp & (UF_EXCLOSE|UF_INHERIT)) != UF_INHERIT)
4137 *flagp |= UF_EXCLOSE;
4138 *flagp &= ~UF_INHERIT;
4139 }
4140
4141 if (
4142 ((*flagp & (UF_RESERVED|UF_EXCLOSE)) == UF_EXCLOSE)
4143 #if CONFIG_MACF
4144 || (fp && mac_file_check_inherit(proc_ucred(p), fp->f_fglob))
4145 #endif
4146 ) {
4147 if (i < fdp->fd_knlistsize)
4148 knote_fdclose(p, i);
4149 procfdtbl_clearfd(p, i);
4150 if (i == fdp->fd_lastfile && i > 0)
4151 fdp->fd_lastfile--;
4152 if (i < fdp->fd_freefile)
4153 fdp->fd_freefile = i;
4154
4155 /*
4156 * Wait for any third party viewers (e.g., lsof)
4157 * to release their references to this fileproc.
4158 */
4159 while (fp->f_iocount > 0) {
4160 p->p_fpdrainwait = 1;
4161 msleep(&p->p_fpdrainwait, &p->p_fdmlock, PRIBIO,
4162 "fpdrain", NULL);
4163 }
4164
4165 closef_locked(fp, fp->f_fglob, p);
4166
4167 FREE_ZONE(fp, sizeof(*fp), M_FILEPROC);
4168 }
4169 }
4170 proc_fdunlock(p);
4171 }
4172
4173
4174 /*
4175 * fdcopy
4176 *
4177 * Description: Copy a filedesc structure. This is normally used as part of
4178 * forkproc() when forking a new process, to copy the per process
4179 * open file table over to the new process.
4180 *
4181 * Parameters: p Process whose open file table
4182 * is to be copied (parent)
4183 * uth_cdir Per thread current working
4184 * cirectory, or NULL
4185 *
4186 * Returns: NULL Copy failed
4187 * !NULL Pointer to new struct filedesc
4188 *
4189 * Locks: This function internally takes and drops proc_fdlock()
4190 *
4191 * Notes: Files are copied directly, ignoring the new resource limits
4192 * for the process that's being copied into. Since the descriptor
4193 * references are just additional references, this does not count
4194 * against the number of open files on the system.
4195 *
4196 * The struct filedesc includes the current working directory,
4197 * and the current root directory, if the process is chroot'ed.
4198 *
4199 * If the exec was called by a thread using a per thread current
4200 * working directory, we inherit the working directory from the
4201 * thread making the call, rather than from the process.
4202 *
4203 * In the case of a failure to obtain a reference, for most cases,
4204 * the file entry will be silently dropped. There's an exception
4205 * for the case of a chroot dir, since a failure to to obtain a
4206 * reference there would constitute an "escape" from the chroot
4207 * environment, which must not be allowed. In that case, we will
4208 * deny the execve() operation, rather than allowing the escape.
4209 */
4210 struct filedesc *
4211 fdcopy(proc_t p, vnode_t uth_cdir)
4212 {
4213 struct filedesc *newfdp, *fdp = p->p_fd;
4214 int i;
4215 struct fileproc *ofp, *fp;
4216 vnode_t v_dir;
4217
4218 MALLOC_ZONE(newfdp, struct filedesc *,
4219 sizeof(*newfdp), M_FILEDESC, M_WAITOK);
4220 if (newfdp == NULL)
4221 return(NULL);
4222
4223 proc_fdlock(p);
4224
4225 /*
4226 * the FD_CHROOT flag will be inherited via this copy
4227 */
4228 (void) memcpy(newfdp, fdp, sizeof(*newfdp));
4229
4230 /*
4231 * If we are running with per-thread current working directories,
4232 * inherit the new current working directory from the current thread
4233 * instead, before we take our references.
4234 */
4235 if (uth_cdir != NULLVP)
4236 newfdp->fd_cdir = uth_cdir;
4237
4238 /*
4239 * For both fd_cdir and fd_rdir make sure we get
4240 * a valid reference... if we can't, than set
4241 * set the pointer(s) to NULL in the child... this
4242 * will keep us from using a non-referenced vp
4243 * and allows us to do the vnode_rele only on
4244 * a properly referenced vp
4245 */
4246 if ( (v_dir = newfdp->fd_cdir) ) {
4247 if (vnode_getwithref(v_dir) == 0) {
4248 if ( (vnode_ref(v_dir)) )
4249 newfdp->fd_cdir = NULL;
4250 vnode_put(v_dir);
4251 } else
4252 newfdp->fd_cdir = NULL;
4253 }
4254 if (newfdp->fd_cdir == NULL && fdp->fd_cdir) {
4255 /*
4256 * we couldn't get a new reference on
4257 * the current working directory being
4258 * inherited... we might as well drop
4259 * our reference from the parent also
4260 * since the vnode has gone DEAD making
4261 * it useless... by dropping it we'll
4262 * be that much closer to recycling it
4263 */
4264 vnode_rele(fdp->fd_cdir);
4265 fdp->fd_cdir = NULL;
4266 }
4267
4268 if ( (v_dir = newfdp->fd_rdir) ) {
4269 if (vnode_getwithref(v_dir) == 0) {
4270 if ( (vnode_ref(v_dir)) )
4271 newfdp->fd_rdir = NULL;
4272 vnode_put(v_dir);
4273 } else {
4274 newfdp->fd_rdir = NULL;
4275 }
4276 }
4277 /* Coming from a chroot environment and unable to get a reference... */
4278 if (newfdp->fd_rdir == NULL && fdp->fd_rdir) {
4279 /*
4280 * We couldn't get a new reference on
4281 * the chroot directory being
4282 * inherited... this is fatal, since
4283 * otherwise it would constitute an
4284 * escape from a chroot environment by
4285 * the new process.
4286 */
4287 if (newfdp->fd_cdir)
4288 vnode_rele(newfdp->fd_cdir);
4289 FREE_ZONE(newfdp, sizeof *newfdp, M_FILEDESC);
4290 return(NULL);
4291 }
4292 newfdp->fd_refcnt = 1;
4293
4294 /*
4295 * If the number of open files fits in the internal arrays
4296 * of the open file structure, use them, otherwise allocate
4297 * additional memory for the number of descriptors currently
4298 * in use.
4299 */
4300 if (newfdp->fd_lastfile < NDFILE)
4301 i = NDFILE;
4302 else {
4303 /*
4304 * Compute the smallest multiple of NDEXTENT needed
4305 * for the file descriptors currently in use,
4306 * allowing the table to shrink.
4307 */
4308 i = newfdp->fd_nfiles;
4309 while (i > 2 * NDEXTENT && i > newfdp->fd_lastfile * 2)
4310 i /= 2;
4311 }
4312 proc_fdunlock(p);
4313
4314 MALLOC_ZONE(newfdp->fd_ofiles, struct fileproc **,
4315 i * OFILESIZE, M_OFILETABL, M_WAITOK);
4316 if (newfdp->fd_ofiles == NULL) {
4317 if (newfdp->fd_cdir)
4318 vnode_rele(newfdp->fd_cdir);
4319 if (newfdp->fd_rdir)
4320 vnode_rele(newfdp->fd_rdir);
4321
4322 FREE_ZONE(newfdp, sizeof(*newfdp), M_FILEDESC);
4323 return(NULL);
4324 }
4325 (void) memset(newfdp->fd_ofiles, 0, i * OFILESIZE);
4326 proc_fdlock(p);
4327
4328 newfdp->fd_ofileflags = (char *) &newfdp->fd_ofiles[i];
4329 newfdp->fd_nfiles = i;
4330
4331 if (fdp->fd_nfiles > 0) {
4332 struct fileproc **fpp;
4333 char *flags;
4334
4335 (void) memcpy(newfdp->fd_ofiles, fdp->fd_ofiles,
4336 (newfdp->fd_lastfile + 1) * sizeof(*fdp->fd_ofiles));
4337 (void) memcpy(newfdp->fd_ofileflags, fdp->fd_ofileflags,
4338 (newfdp->fd_lastfile + 1) * sizeof(*fdp->fd_ofileflags));
4339
4340 /*
4341 * kq descriptors cannot be copied.
4342 */
4343 if (newfdp->fd_knlistsize != -1) {
4344 fpp = &newfdp->fd_ofiles[newfdp->fd_lastfile];
4345 for (i = newfdp->fd_lastfile; i >= 0; i--, fpp--) {
4346 if (*fpp != NULL && (*fpp)->f_type == DTYPE_KQUEUE) {
4347 *fpp = NULL;
4348 newfdp->fd_ofileflags[i] = 0;
4349 if (i < newfdp->fd_freefile)
4350 newfdp->fd_freefile = i;
4351 }
4352 if (*fpp == NULL && i == newfdp->fd_lastfile && i > 0)
4353 newfdp->fd_lastfile--;
4354 }
4355 newfdp->fd_knlist = NULL;
4356 newfdp->fd_knlistsize = -1;
4357 newfdp->fd_knhash = NULL;
4358 newfdp->fd_knhashmask = 0;
4359 }
4360 fpp = newfdp->fd_ofiles;
4361 flags = newfdp->fd_ofileflags;
4362
4363 for (i = newfdp->fd_lastfile + 1; --i >= 0; fpp++, flags++)
4364 if ((ofp = *fpp) != NULL && !(*flags & UF_RESERVED)) {
4365 MALLOC_ZONE(fp, struct fileproc *, sizeof(struct fileproc), M_FILEPROC, M_WAITOK);
4366 if (fp == NULL) {
4367 /*
4368 * XXX no room to copy, unable to
4369 * XXX safely unwind state at present
4370 */
4371 *fpp = NULL;
4372 } else {
4373 bzero(fp, sizeof(struct fileproc));
4374 fp->f_flags = ofp->f_flags;
4375 //fp->f_iocount = ofp->f_iocount;
4376 fp->f_iocount = 0;
4377 fp->f_fglob = ofp->f_fglob;
4378 (void)fg_ref(fp);
4379 *fpp = fp;
4380 }
4381 } else {
4382 if (i < newfdp->fd_freefile)
4383 newfdp->fd_freefile = i;
4384 *fpp = NULL;
4385 *flags = 0;
4386 }
4387 }
4388
4389 proc_fdunlock(p);
4390 return (newfdp);
4391 }
4392
4393
4394 /*
4395 * fdfree
4396 *
4397 * Description: Release a filedesc (per process open file table) structure;
4398 * this is done on process exit(), or from forkproc_free() if
4399 * the fork fails for some reason subsequent to a successful
4400 * call to fdcopy()
4401 *
4402 * Parameters: p Pointer to process going away
4403 *
4404 * Returns: void
4405 *
4406 * Locks: This function internally takes and drops proc_fdlock()
4407 */
4408 void
4409 fdfree(proc_t p)
4410 {
4411 struct filedesc *fdp;
4412 struct fileproc *fp;
4413 int i;
4414
4415 proc_fdlock(p);
4416
4417 /* Certain daemons might not have file descriptors */
4418 fdp = p->p_fd;
4419
4420 if ((fdp == NULL) || (--fdp->fd_refcnt > 0)) {
4421 proc_fdunlock(p);
4422 return;
4423 }
4424 if (fdp->fd_refcnt == 0xffff)
4425 panic("fdfree: bad fd_refcnt");
4426
4427 /* Last reference: the structure can't change out from under us */
4428
4429 if (fdp->fd_nfiles > 0 && fdp->fd_ofiles) {
4430 for (i = fdp->fd_lastfile; i >= 0; i--) {
4431 if ((fp = fdp->fd_ofiles[i]) != NULL) {
4432
4433 if (fdp->fd_ofileflags[i] & UF_RESERVED)
4434 panic("fdfree: found fp with UF_RESERVED");
4435
4436 /* closef drops the iocount ... */
4437 if ((fp->f_flags & FP_INCHRREAD) != 0)
4438 fp->f_iocount++;
4439 procfdtbl_reservefd(p, i);
4440
4441 if (i < fdp->fd_knlistsize)
4442 knote_fdclose(p, i);
4443 if (fp->f_flags & FP_WAITEVENT)
4444 (void)waitevent_close(p, fp);
4445 (void) closef_locked(fp, fp->f_fglob, p);
4446 FREE_ZONE(fp, sizeof(*fp), M_FILEPROC);
4447 }
4448 }
4449 FREE_ZONE(fdp->fd_ofiles, fdp->fd_nfiles * OFILESIZE, M_OFILETABL);
4450 fdp->fd_ofiles = NULL;
4451 fdp->fd_nfiles = 0;
4452 }
4453
4454 proc_fdunlock(p);
4455
4456 if (fdp->fd_cdir)
4457 vnode_rele(fdp->fd_cdir);
4458 if (fdp->fd_rdir)
4459 vnode_rele(fdp->fd_rdir);
4460
4461 proc_fdlock_spin(p);
4462 p->p_fd = NULL;
4463 proc_fdunlock(p);
4464
4465 if (fdp->fd_knlist)
4466 FREE(fdp->fd_knlist, M_KQUEUE);
4467 if (fdp->fd_knhash)
4468 FREE(fdp->fd_knhash, M_KQUEUE);
4469
4470 FREE_ZONE(fdp, sizeof(*fdp), M_FILEDESC);
4471 }
4472
4473
4474 /*
4475 * closef_finish
4476 *
4477 * Description: Called on last open instance for a fileglob for a file being
4478 * closed.
4479 *
4480 * Parameters: fp Pointer to fileproc for fd
4481 * fg Pointer to fileglob for fd
4482 * p Pointer to proc structure
4483 *
4484 * Returns: 0 Success
4485 * <fo_close>:??? Anything returnable by a per-fileops
4486 * close function
4487 *
4488 * Note: fp can only be non-NULL if p is also non-NULL. If p is NULL,
4489 * then fg must eith be locked (FHASLOCK) or must not have a
4490 * type of DTYPE_VNODE.
4491 *
4492 * On return, the fg is freed.
4493 *
4494 * This function may block draining output to a character
4495 * device on last close of that device.
4496 */
4497 static int
4498 closef_finish(struct fileproc *fp, struct fileglob *fg, proc_t p, vfs_context_t ctx)
4499 {
4500 int error;
4501
4502
4503 /* fg_ops completed initialization? */
4504 if (fg->fg_ops)
4505 error = fo_close(fg, ctx);
4506 else
4507 error = 0;
4508
4509 /* if fp is non-NULL, drain it out */
4510 if (((fp != (struct fileproc *)0) && ((fp->f_flags & FP_INCHRREAD) != 0))) {
4511 proc_fdlock_spin(p);
4512 if ( ((fp->f_flags & FP_INCHRREAD) != 0) ) {
4513 fileproc_drain(p, fp);
4514 }
4515 proc_fdunlock(p);
4516 }
4517 fg_free(fg);
4518
4519 return (error);
4520 }
4521
4522 /*
4523 * closef_locked
4524 *
4525 * Description: Internal form of closef; called with proc_fdlock held
4526 *
4527 * Parameters: fp Pointer to fileproc for fd
4528 * fg Pointer to fileglob for fd
4529 * p Pointer to proc structure
4530 *
4531 * Returns: 0 Success
4532 * closef_finish:??? Anything returnable by a per-fileops
4533 * close function
4534 *
4535 * Note: Decrements reference count on file structure; if this was the
4536 * last reference, then closef_finish() is called
4537 *
4538 * p and fp are allowed to be NULL when closing a file that was
4539 * being passed in a message (but only if we are called when this
4540 * is NOT the last reference).
4541 */
4542 int
4543 closef_locked(struct fileproc *fp, struct fileglob *fg, proc_t p)
4544 {
4545 struct vnode *vp;
4546 struct flock lf;
4547 struct vfs_context context;
4548 int error;
4549
4550 if (fg == NULL) {
4551 return (0);
4552 }
4553
4554 /* Set up context with cred stashed in fg */
4555 if (p == current_proc())
4556 context.vc_thread = current_thread();
4557 else
4558 context.vc_thread = NULL;
4559 context.vc_ucred = fg->fg_cred;
4560
4561 /*
4562 * POSIX record locking dictates that any close releases ALL
4563 * locks owned by this process. This is handled by setting
4564 * a flag in the unlock to free ONLY locks obeying POSIX
4565 * semantics, and not to free BSD-style file locks.
4566 * If the descriptor was in a message, POSIX-style locks
4567 * aren't passed with the descriptor.
4568 */
4569 if (p && (p->p_ladvflag & P_LADVLOCK) && fg->fg_type == DTYPE_VNODE) {
4570 proc_fdunlock(p);
4571
4572 lf.l_whence = SEEK_SET;
4573 lf.l_start = 0;
4574 lf.l_len = 0;
4575 lf.l_type = F_UNLCK;
4576 vp = (struct vnode *)fg->fg_data;
4577
4578 if ( (error = vnode_getwithref(vp)) == 0 ) {
4579 (void) VNOP_ADVLOCK(vp, (caddr_t)p, F_UNLCK, &lf, F_POSIX, &context);
4580 (void)vnode_put(vp);
4581 }
4582 proc_fdlock(p);
4583 }
4584 lck_mtx_lock_spin(&fg->fg_lock);
4585 fg->fg_count--;
4586
4587 if (fg->fg_count > 0) {
4588 lck_mtx_unlock(&fg->fg_lock);
4589 return (0);
4590 }
4591 #if DIAGNOSTIC
4592 if (fg->fg_count != 0)
4593 panic("fg %p: being freed with bad fg_count (%d)", fg, fg->fg_count);
4594 #endif
4595
4596 if (fp && (fp->f_flags & FP_WRITTEN))
4597 fg->fg_flag |= FWASWRITTEN;
4598
4599 fg->fg_lflags |= FG_TERM;
4600 lck_mtx_unlock(&fg->fg_lock);
4601
4602 if (p)
4603 proc_fdunlock(p);
4604 error = closef_finish(fp, fg, p, &context);
4605 if (p)
4606 proc_fdlock(p);
4607
4608 return(error);
4609 }
4610
4611
4612 /*
4613 * fileproc_drain
4614 *
4615 * Description: Drain out pending I/O operations
4616 *
4617 * Parameters: p Process closing this file
4618 * fp fileproc struct for the open
4619 * instance on the file
4620 *
4621 * Returns: void
4622 *
4623 * Locks: Assumes the caller holds the proc_fdlock
4624 *
4625 * Notes: For character devices, this occurs on the last close of the
4626 * device; for all other file descriptors, this occurs on each
4627 * close to prevent fd's from being closed out from under
4628 * operations currently in progress and blocked
4629 *
4630 * See Also: file_vnode(), file_socket(), file_drop(), and the cautions
4631 * regarding their use and interaction with this function.
4632 */
4633 void
4634 fileproc_drain(proc_t p, struct fileproc * fp)
4635 {
4636 struct vfs_context context;
4637
4638 context.vc_thread = proc_thread(p); /* XXX */
4639 context.vc_ucred = fp->f_fglob->fg_cred;
4640
4641 fp->f_iocount-- ; /* (the one the close holds) */
4642
4643 while (fp->f_iocount) {
4644
4645 lck_mtx_convert_spin(&p->p_fdmlock);
4646
4647 if (fp->f_fglob->fg_ops->fo_drain) {
4648 (*fp->f_fglob->fg_ops->fo_drain)(fp, &context);
4649 }
4650 if ((fp->f_flags & FP_INSELECT) == FP_INSELECT) {
4651 if (wait_queue_wakeup_all((wait_queue_t)fp->f_waddr, NULL, THREAD_INTERRUPTED) == KERN_INVALID_ARGUMENT)
4652 panic("bad wait queue for wait_queue_wakeup_all %p", fp->f_waddr);
4653 }
4654 if ((fp->f_flags & FP_SELCONFLICT) == FP_SELCONFLICT) {
4655 if (wait_queue_wakeup_all(&select_conflict_queue, NULL, THREAD_INTERRUPTED) == KERN_INVALID_ARGUMENT)
4656 panic("bad select_conflict_queue");
4657 }
4658 p->p_fpdrainwait = 1;
4659
4660 msleep(&p->p_fpdrainwait, &p->p_fdmlock, PRIBIO, "fpdrain", NULL);
4661
4662 }
4663 #if DIAGNOSTIC
4664 if ((fp->f_flags & FP_INSELECT) != 0)
4665 panic("FP_INSELECT set on drained fp");
4666 #endif
4667 if ((fp->f_flags & FP_SELCONFLICT) == FP_SELCONFLICT)
4668 fp->f_flags &= ~FP_SELCONFLICT;
4669 }
4670
4671
4672 /*
4673 * fp_free
4674 *
4675 * Description: Release the fd and free the fileproc associated with the fd
4676 * in the per process open file table of the specified process;
4677 * these values must correspond.
4678 *
4679 * Parameters: p Process containing fd
4680 * fd fd to be released
4681 * fp fileproc to be freed
4682 *
4683 * Returns: 0 Success
4684 *
4685 * Notes: XXX function should be void - no one interprets the returns
4686 * XXX code
4687 */
4688 int
4689 fp_free(proc_t p, int fd, struct fileproc * fp)
4690 {
4691 proc_fdlock_spin(p);
4692 fdrelse(p, fd);
4693 proc_fdunlock(p);
4694
4695 fg_free(fp->f_fglob);
4696 FREE_ZONE(fp, sizeof(*fp), M_FILEPROC);
4697 return(0);
4698 }
4699
4700
4701 /*
4702 * flock
4703 *
4704 * Description: Apply an advisory lock on a file descriptor.
4705 *
4706 * Parameters: p Process making request
4707 * uap->fd fd on which the lock is to be
4708 * attempted
4709 * uap->how (Un)Lock bits, including type
4710 * retval Pointer to the call return area
4711 *
4712 * Returns: 0 Success
4713 * fp_getfvp:EBADF Bad file descriptor
4714 * fp_getfvp:ENOTSUP fd does not refer to a vnode
4715 * vnode_getwithref:???
4716 * VNOP_ADVLOCK:???
4717 *
4718 * Implicit returns:
4719 * *retval (modified) Size of dtable
4720 *
4721 * Notes: Just attempt to get a record lock of the requested type on
4722 * the entire file (l_whence = SEEK_SET, l_start = 0, l_len = 0).
4723 */
4724 int
4725 flock(proc_t p, struct flock_args *uap, __unused int32_t *retval)
4726 {
4727 int fd = uap->fd;
4728 int how = uap->how;
4729 struct fileproc *fp;
4730 struct vnode *vp;
4731 struct flock lf;
4732 vfs_context_t ctx = vfs_context_current();
4733 int error=0;
4734
4735 AUDIT_ARG(fd, uap->fd);
4736 if ( (error = fp_getfvp(p, fd, &fp, &vp)) ) {
4737 return(error);
4738 }
4739 if ( (error = vnode_getwithref(vp)) ) {
4740 goto out1;
4741 }
4742 AUDIT_ARG(vnpath, vp, ARG_VNODE1);
4743
4744 lf.l_whence = SEEK_SET;
4745 lf.l_start = 0;
4746 lf.l_len = 0;
4747 if (how & LOCK_UN) {
4748 lf.l_type = F_UNLCK;
4749 fp->f_flag &= ~FHASLOCK;
4750 error = VNOP_ADVLOCK(vp, (caddr_t)fp->f_fglob, F_UNLCK, &lf, F_FLOCK, ctx);
4751 goto out;
4752 }
4753 if (how & LOCK_EX)
4754 lf.l_type = F_WRLCK;
4755 else if (how & LOCK_SH)
4756 lf.l_type = F_RDLCK;
4757 else {
4758 error = EBADF;
4759 goto out;
4760 }
4761 #if CONFIG_MACF
4762 error = mac_file_check_lock(proc_ucred(p), fp->f_fglob, F_SETLK, &lf);
4763 if (error)
4764 goto out;
4765 #endif
4766 fp->f_flag |= FHASLOCK;
4767 if (how & LOCK_NB) {
4768 error = VNOP_ADVLOCK(vp, (caddr_t)fp->f_fglob, F_SETLK, &lf, F_FLOCK, ctx);
4769 goto out;
4770 }
4771 error = VNOP_ADVLOCK(vp, (caddr_t)fp->f_fglob, F_SETLK, &lf, F_FLOCK|F_WAIT, ctx);
4772 out:
4773 (void)vnode_put(vp);
4774 out1:
4775 fp_drop(p, fd, fp, 0);
4776 return(error);
4777
4778 }
4779
4780 /*
4781 * fileport_makeport
4782 *
4783 * Description: Obtain a Mach send right for a given file descriptor.
4784 *
4785 * Parameters: p Process calling fileport
4786 * uap->fd The fd to reference
4787 * uap->portnamep User address at which to place port name.
4788 *
4789 * Returns: 0 Success.
4790 * EBADF Bad file descriptor.
4791 * EINVAL File descriptor had type that cannot be sent, misc. other errors.
4792 * EFAULT Address at which to store port name is not valid.
4793 * EAGAIN Resource shortage.
4794 *
4795 * Implicit returns:
4796 * On success, name of send right is stored at user-specified address.
4797 */
4798 int
4799 fileport_makeport(proc_t p, struct fileport_makeport_args *uap,
4800 __unused int *retval)
4801 {
4802 int err;
4803 int fd = uap->fd;
4804 user_addr_t user_portaddr = uap->portnamep;
4805 struct fileproc *fp = FILEPROC_NULL;
4806 struct fileglob *fg = NULL;
4807 ipc_port_t fileport;
4808 mach_port_name_t name = MACH_PORT_NULL;
4809
4810 err = fp_lookup(p, fd, &fp, 0);
4811 if (err != 0) {
4812 goto out;
4813 }
4814
4815 if (!filetype_issendable(fp->f_type)) {
4816 err = EINVAL;
4817 goto out;
4818 }
4819
4820 /* Dropped when port is deallocated */
4821 fg = fp->f_fglob;
4822 fg_ref(fp);
4823
4824 /* Allocate and initialize a port */
4825 fileport = fileport_alloc(fg);
4826 if (fileport == IPC_PORT_NULL) {
4827 err = EAGAIN;
4828 fg_drop(fp);
4829 goto out;
4830 }
4831
4832 /* Add an entry. Deallocates port on failure. */
4833 name = ipc_port_copyout_send(fileport, get_task_ipcspace(p->task));
4834 if (!MACH_PORT_VALID(name)) {
4835 err = EINVAL;
4836 goto out;
4837 }
4838
4839 err = copyout(&name, user_portaddr, sizeof(mach_port_name_t));
4840 if (err != 0) {
4841 goto out;
4842 }
4843
4844 /* Tag the fileglob for debugging purposes */
4845 lck_mtx_lock_spin(&fg->fg_lock);
4846 fg->fg_lflags |= FG_PORTMADE;
4847 lck_mtx_unlock(&fg->fg_lock);
4848
4849 fp_drop(p, fd, fp, 0);
4850
4851 return 0;
4852
4853 out:
4854 if (MACH_PORT_VALID(name)) {
4855 /* Don't care if another thread races us to deallocate the entry */
4856 (void) mach_port_deallocate(get_task_ipcspace(p->task), name);
4857 }
4858
4859 if (fp != FILEPROC_NULL) {
4860 fp_drop(p, fd, fp, 0);
4861 }
4862
4863 return err;
4864 }
4865
4866 void
4867 fileport_releasefg(struct fileglob *fg)
4868 {
4869 (void)closef_locked(NULL, fg, PROC_NULL);
4870
4871 return;
4872 }
4873
4874
4875 /*
4876 * fileport_makefd
4877 *
4878 * Description: Obtain the file descriptor for a given Mach send right.
4879 *
4880 * Parameters: p Process calling fileport
4881 * uap->port Name of send right to file port.
4882 *
4883 * Returns: 0 Success
4884 * EINVAL Invalid Mach port name, or port is not for a file.
4885 * fdalloc:EMFILE
4886 * fdalloc:ENOMEM Unable to allocate fileproc or extend file table.
4887 *
4888 * Implicit returns:
4889 * *retval (modified) The new descriptor
4890 */
4891 int
4892 fileport_makefd(proc_t p, struct fileport_makefd_args *uap, int32_t *retval)
4893 {
4894 struct fileglob *fg;
4895 struct fileproc *fp = FILEPROC_NULL;
4896 ipc_port_t port = IPC_PORT_NULL;
4897 mach_port_name_t send = uap->port;
4898 kern_return_t res;
4899 int fd;
4900 int err;
4901
4902 res = ipc_object_copyin(get_task_ipcspace(p->task),
4903 send, MACH_MSG_TYPE_COPY_SEND, &port);
4904
4905 if (res != KERN_SUCCESS) {
4906 err = EINVAL;
4907 goto out;
4908 }
4909
4910 fg = fileport_port_to_fileglob(port);
4911 if (fg == NULL) {
4912 err = EINVAL;
4913 goto out;
4914 }
4915
4916 MALLOC_ZONE(fp, struct fileproc *, sizeof(*fp), M_FILEPROC, M_WAITOK);
4917 if (fp == FILEPROC_NULL) {
4918 err = ENOMEM;
4919 goto out;
4920 }
4921
4922 bzero(fp, sizeof(*fp));
4923
4924 fp->f_fglob = fg;
4925 fg_ref(fp);
4926
4927 proc_fdlock(p);
4928 err = fdalloc(p, 0, &fd);
4929 if (err != 0) {
4930 proc_fdunlock(p);
4931 goto out;
4932 }
4933 *fdflags(p, fd) |= UF_EXCLOSE;
4934
4935 procfdtbl_releasefd(p, fd, fp);
4936 proc_fdunlock(p);
4937
4938 *retval = fd;
4939 err = 0;
4940 out:
4941 if ((fp != NULL) && (0 != err)) {
4942 FREE_ZONE(fp, sizeof(*fp), M_FILEPROC);
4943 }
4944
4945 if (IPC_PORT_NULL != port) {
4946 ipc_port_release_send(port);
4947 }
4948
4949 return err;
4950 }
4951
4952
4953 /*
4954 * dupfdopen
4955 *
4956 * Description: Duplicate the specified descriptor to a free descriptor;
4957 * this is the second half of fdopen(), above.
4958 *
4959 * Parameters: fdp filedesc pointer to fill in
4960 * indx fd to dup to
4961 * dfd fd to dup from
4962 * mode mode to set on new fd
4963 * error command code
4964 *
4965 * Returns: 0 Success
4966 * EBADF Source fd is bad
4967 * EACCES Requested mode not allowed
4968 * !0 'error', if not ENODEV or
4969 * ENXIO
4970 *
4971 * Notes: XXX This is not thread safe; see fdopen() above
4972 */
4973 int
4974 dupfdopen(struct filedesc *fdp, int indx, int dfd, int flags, int error)
4975 {
4976 struct fileproc *wfp;
4977 struct fileproc *fp;
4978 #if CONFIG_MACF
4979 int myerror;
4980 #endif
4981 proc_t p = current_proc();
4982
4983 /*
4984 * If the to-be-dup'd fd number is greater than the allowed number
4985 * of file descriptors, or the fd to be dup'd has already been
4986 * closed, reject. Note, check for new == old is necessary as
4987 * falloc could allocate an already closed to-be-dup'd descriptor
4988 * as the new descriptor.
4989 */
4990 proc_fdlock(p);
4991
4992 fp = fdp->fd_ofiles[indx];
4993 if (dfd < 0 || dfd >= fdp->fd_nfiles ||
4994 (wfp = fdp->fd_ofiles[dfd]) == NULL || wfp == fp ||
4995 (fdp->fd_ofileflags[dfd] & UF_RESERVED)) {
4996
4997 proc_fdunlock(p);
4998 return (EBADF);
4999 }
5000 #if CONFIG_MACF
5001 myerror = mac_file_check_dup(proc_ucred(p), wfp->f_fglob, dfd);
5002 if (myerror) {
5003 proc_fdunlock(p);
5004 return (myerror);
5005 }
5006 #endif
5007 /*
5008 * There are two cases of interest here.
5009 *
5010 * For ENODEV simply dup (dfd) to file descriptor
5011 * (indx) and return.
5012 *
5013 * For ENXIO steal away the file structure from (dfd) and
5014 * store it in (indx). (dfd) is effectively closed by
5015 * this operation.
5016 *
5017 * Any other error code is just returned.
5018 */
5019 switch (error) {
5020 case ENODEV:
5021 /*
5022 * Check that the mode the file is being opened for is a
5023 * subset of the mode of the existing descriptor.
5024 */
5025 if (((flags & (FREAD|FWRITE)) | wfp->f_flag) != wfp->f_flag) {
5026 proc_fdunlock(p);
5027 return (EACCES);
5028 }
5029 if (indx > fdp->fd_lastfile)
5030 fdp->fd_lastfile = indx;
5031 (void)fg_ref(wfp);
5032
5033 if (fp->f_fglob)
5034 fg_free(fp->f_fglob);
5035 fp->f_fglob = wfp->f_fglob;
5036
5037 fdp->fd_ofileflags[indx] = fdp->fd_ofileflags[dfd] |
5038 (flags & O_CLOEXEC) ? UF_EXCLOSE : 0;
5039
5040 proc_fdunlock(p);
5041 return (0);
5042
5043 default:
5044 proc_fdunlock(p);
5045 return (error);
5046 }
5047 /* NOTREACHED */
5048 }
5049
5050
5051 /*
5052 * fg_ref
5053 *
5054 * Description: Add a reference to a fileglob by fileproc
5055 *
5056 * Parameters: fp fileproc containing fileglob
5057 * pointer
5058 *
5059 * Returns: void
5060 *
5061 * Notes: XXX Should use OSAddAtomic?
5062 */
5063 void
5064 fg_ref(struct fileproc * fp)
5065 {
5066 struct fileglob *fg;
5067
5068 fg = fp->f_fglob;
5069
5070 lck_mtx_lock_spin(&fg->fg_lock);
5071
5072 #if DIAGNOSTIC
5073 if ((fp->f_flags & ~((unsigned int)FP_VALID_FLAGS)) != 0)
5074 panic("fg_ref: invalid bits on fp %p", fp);
5075
5076 if (fg->fg_count == 0)
5077 panic("fg_ref: adding fgcount to zeroed fg: fp %p fg %p",
5078 fp, fg);
5079 #endif
5080 fg->fg_count++;
5081 lck_mtx_unlock(&fg->fg_lock);
5082 }
5083
5084
5085 /*
5086 * fg_drop
5087 *
5088 * Description: Remove a reference to a fileglob by fileproc
5089 *
5090 * Parameters: fp fileproc containing fileglob
5091 * pointer
5092 *
5093 * Returns: void
5094 *
5095 * Notes: XXX Should use OSAddAtomic?
5096 */
5097 void
5098 fg_drop(struct fileproc * fp)
5099 {
5100 struct fileglob *fg;
5101
5102 fg = fp->f_fglob;
5103 lck_mtx_lock_spin(&fg->fg_lock);
5104 fg->fg_count--;
5105 lck_mtx_unlock(&fg->fg_lock);
5106 }
5107
5108
5109 /*
5110 * fg_insertuipc
5111 *
5112 * Description: Insert fileglob onto message queue
5113 *
5114 * Parameters: fg Fileglob pointer to insert
5115 *
5116 * Returns: void
5117 *
5118 * Locks: Takes and drops fg_lock, potentially many times
5119 */
5120 void
5121 fg_insertuipc(struct fileglob * fg)
5122 {
5123 int insertque = 0;
5124
5125 lck_mtx_lock_spin(&fg->fg_lock);
5126
5127 while (fg->fg_lflags & FG_RMMSGQ) {
5128 lck_mtx_convert_spin(&fg->fg_lock);
5129
5130 fg->fg_lflags |= FG_WRMMSGQ;
5131 msleep(&fg->fg_lflags, &fg->fg_lock, 0, "fg_insertuipc", NULL);
5132 }
5133
5134 fg->fg_count++;
5135 fg->fg_msgcount++;
5136 if (fg->fg_msgcount == 1) {
5137 fg->fg_lflags |= FG_INSMSGQ;
5138 insertque=1;
5139 }
5140 lck_mtx_unlock(&fg->fg_lock);
5141
5142 if (insertque) {
5143 lck_mtx_lock_spin(uipc_lock);
5144 unp_gc_wait();
5145 LIST_INSERT_HEAD(&fmsghead, fg, f_msglist);
5146 lck_mtx_unlock(uipc_lock);
5147 lck_mtx_lock(&fg->fg_lock);
5148 fg->fg_lflags &= ~FG_INSMSGQ;
5149 if (fg->fg_lflags & FG_WINSMSGQ) {
5150 fg->fg_lflags &= ~FG_WINSMSGQ;
5151 wakeup(&fg->fg_lflags);
5152 }
5153 lck_mtx_unlock(&fg->fg_lock);
5154 }
5155
5156 }
5157
5158
5159 /*
5160 * fg_removeuipc
5161 *
5162 * Description: Remove fileglob from message queue
5163 *
5164 * Parameters: fg Fileglob pointer to remove
5165 *
5166 * Returns: void
5167 *
5168 * Locks: Takes and drops fg_lock, potentially many times
5169 */
5170 void
5171 fg_removeuipc(struct fileglob * fg)
5172 {
5173 int removeque = 0;
5174
5175 lck_mtx_lock_spin(&fg->fg_lock);
5176 while (fg->fg_lflags & FG_INSMSGQ) {
5177 lck_mtx_convert_spin(&fg->fg_lock);
5178
5179 fg->fg_lflags |= FG_WINSMSGQ;
5180 msleep(&fg->fg_lflags, &fg->fg_lock, 0, "fg_removeuipc", NULL);
5181 }
5182 fg->fg_msgcount--;
5183 if (fg->fg_msgcount == 0) {
5184 fg->fg_lflags |= FG_RMMSGQ;
5185 removeque=1;
5186 }
5187 lck_mtx_unlock(&fg->fg_lock);
5188
5189 if (removeque) {
5190 lck_mtx_lock_spin(uipc_lock);
5191 unp_gc_wait();
5192 LIST_REMOVE(fg, f_msglist);
5193 lck_mtx_unlock(uipc_lock);
5194 lck_mtx_lock(&fg->fg_lock);
5195 fg->fg_lflags &= ~FG_RMMSGQ;
5196 if (fg->fg_lflags & FG_WRMMSGQ) {
5197 fg->fg_lflags &= ~FG_WRMMSGQ;
5198 wakeup(&fg->fg_lflags);
5199 }
5200 lck_mtx_unlock(&fg->fg_lock);
5201 }
5202 }
5203
5204
5205 /*
5206 * fo_read
5207 *
5208 * Description: Generic fileops read indirected through the fileops pointer
5209 * in the fileproc structure
5210 *
5211 * Parameters: fp fileproc structure pointer
5212 * uio user I/O structure pointer
5213 * flags FOF_ flags
5214 * ctx VFS context for operation
5215 *
5216 * Returns: 0 Success
5217 * !0 Errno from read
5218 */
5219 int
5220 fo_read(struct fileproc *fp, struct uio *uio, int flags, vfs_context_t ctx)
5221 {
5222 return ((*fp->f_ops->fo_read)(fp, uio, flags, ctx));
5223 }
5224
5225
5226 /*
5227 * fo_write
5228 *
5229 * Description: Generic fileops write indirected through the fileops pointer
5230 * in the fileproc structure
5231 *
5232 * Parameters: fp fileproc structure pointer
5233 * uio user I/O structure pointer
5234 * flags FOF_ flags
5235 * ctx VFS context for operation
5236 *
5237 * Returns: 0 Success
5238 * !0 Errno from write
5239 */
5240 int
5241 fo_write(struct fileproc *fp, struct uio *uio, int flags, vfs_context_t ctx)
5242 {
5243 return((*fp->f_ops->fo_write)(fp, uio, flags, ctx));
5244 }
5245
5246
5247 /*
5248 * fo_ioctl
5249 *
5250 * Description: Generic fileops ioctl indirected through the fileops pointer
5251 * in the fileproc structure
5252 *
5253 * Parameters: fp fileproc structure pointer
5254 * com ioctl command
5255 * data pointer to internalized copy
5256 * of user space ioctl command
5257 * parameter data in kernel space
5258 * ctx VFS context for operation
5259 *
5260 * Returns: 0 Success
5261 * !0 Errno from ioctl
5262 *
5263 * Locks: The caller is assumed to have held the proc_fdlock; this
5264 * function releases and reacquires this lock. If the caller
5265 * accesses data protected by this lock prior to calling this
5266 * function, it will need to revalidate/reacquire any cached
5267 * protected data obtained prior to the call.
5268 */
5269 int
5270 fo_ioctl(struct fileproc *fp, u_long com, caddr_t data, vfs_context_t ctx)
5271 {
5272 int error;
5273
5274 proc_fdunlock(vfs_context_proc(ctx));
5275 error = (*fp->f_ops->fo_ioctl)(fp, com, data, ctx);
5276 proc_fdlock(vfs_context_proc(ctx));
5277 return(error);
5278 }
5279
5280
5281 /*
5282 * fo_select
5283 *
5284 * Description: Generic fileops select indirected through the fileops pointer
5285 * in the fileproc structure
5286 *
5287 * Parameters: fp fileproc structure pointer
5288 * which select which
5289 * wql pointer to wait queue list
5290 * ctx VFS context for operation
5291 *
5292 * Returns: 0 Success
5293 * !0 Errno from select
5294 */
5295 int
5296 fo_select(struct fileproc *fp, int which, void *wql, vfs_context_t ctx)
5297 {
5298 return((*fp->f_ops->fo_select)(fp, which, wql, ctx));
5299 }
5300
5301
5302 /*
5303 * fo_close
5304 *
5305 * Description: Generic fileops close indirected through the fileops pointer
5306 * in the fileproc structure
5307 *
5308 * Parameters: fp fileproc structure pointer for
5309 * file to close
5310 * ctx VFS context for operation
5311 *
5312 * Returns: 0 Success
5313 * !0 Errno from close
5314 */
5315 int
5316 fo_close(struct fileglob *fg, vfs_context_t ctx)
5317 {
5318 return((*fg->fg_ops->fo_close)(fg, ctx));
5319 }
5320
5321
5322 /*
5323 * fo_kqfilter
5324 *
5325 * Description: Generic fileops kqueue filter indirected through the fileops
5326 * pointer in the fileproc structure
5327 *
5328 * Parameters: fp fileproc structure pointer
5329 * kn pointer to knote to filter on
5330 * ctx VFS context for operation
5331 *
5332 * Returns: 0 Success
5333 * !0 Errno from kqueue filter
5334 */
5335 int
5336 fo_kqfilter(struct fileproc *fp, struct knote *kn, vfs_context_t ctx)
5337 {
5338 return ((*fp->f_ops->fo_kqfilter)(fp, kn, ctx));
5339 }
5340
5341 /*
5342 * The ability to send a file descriptor to another
5343 * process is opt-in by file type.
5344 */
5345 boolean_t
5346 filetype_issendable(file_type_t fdtype)
5347 {
5348 switch (fdtype) {
5349 case DTYPE_VNODE:
5350 case DTYPE_SOCKET:
5351 case DTYPE_PIPE:
5352 case DTYPE_PSXSHM:
5353 return TRUE;
5354 default:
5355 /* DTYPE_KQUEUE, DTYPE_FSEVENTS, DTYPE_PSXSEM */
5356 return FALSE;
5357 }
5358 }
mirror of XNU