]> 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-3789.31.2.tar.gz
[apple/xnu.git] / bsd / kern / kern_descrip.c
1 /*
2 * Copyright (c) 2000-2016 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/guarded.h>
84 #include <sys/priv.h>
85 #include <sys/socket.h>
86 #include <sys/socketvar.h>
87 #include <sys/stat.h>
88 #include <sys/ioctl.h>
89 #include <sys/fcntl.h>
90 #include <sys/malloc.h>
91 #include <sys/mman.h>
92 #include <sys/syslog.h>
93 #include <sys/unistd.h>
94 #include <sys/resourcevar.h>
95 #include <sys/aio_kern.h>
96 #include <sys/ev.h>
97 #include <kern/locks.h>
98 #include <sys/uio_internal.h>
99 #include <sys/codesign.h>
100 #include <sys/codedir_internal.h>
101
102 #include <security/audit/audit.h>
103
104 #include <sys/mount_internal.h>
105 #include <sys/kdebug.h>
106 #include <sys/sysproto.h>
107 #include <sys/pipe.h>
108 #include <sys/spawn.h>
109 #include <sys/cprotect.h>
110 #include <kern/kern_types.h>
111 #include <kern/kalloc.h>
112 #include <kern/waitq.h>
113 #include <libkern/OSAtomic.h>
114
115 #include <sys/ubc_internal.h>
116
117 #include <kern/ipc_misc.h>
118 #include <vm/vm_protos.h>
119
120 #include <mach/mach_port.h>
121 #include <stdbool.h>
122
123 kern_return_t ipc_object_copyin(ipc_space_t, mach_port_name_t,
124 mach_msg_type_name_t, ipc_port_t *);
125 void ipc_port_release_send(ipc_port_t);
126
127 struct psemnode;
128 struct pshmnode;
129
130 static int finishdup(proc_t p,
131 struct filedesc *fdp, int old, int new, int flags, int32_t *retval);
132
133 int falloc_locked(proc_t p, struct fileproc **resultfp, int *resultfd, vfs_context_t ctx, int locked);
134 void fg_drop(struct fileproc * fp);
135 void fg_free(struct fileglob *fg);
136 void fg_ref(struct fileproc * fp);
137 void fileport_releasefg(struct fileglob *fg);
138
139 /* flags for close_internal_locked */
140 #define FD_DUP2RESV 1
141
142 /* We don't want these exported */
143
144 __private_extern__
145 int unlink1(vfs_context_t, vnode_t, user_addr_t, enum uio_seg, int);
146
147 static void _fdrelse(struct proc * p, int fd);
148
149
150 extern void file_lock_init(void);
151
152 extern kauth_scope_t kauth_scope_fileop;
153
154 /* Conflict wait queue for when selects collide (opaque type) */
155 extern struct waitq select_conflict_queue;
156
157 #define f_flag f_fglob->fg_flag
158 #define f_type f_fglob->fg_ops->fo_type
159 #define f_msgcount f_fglob->fg_msgcount
160 #define f_cred f_fglob->fg_cred
161 #define f_ops f_fglob->fg_ops
162 #define f_offset f_fglob->fg_offset
163 #define f_data f_fglob->fg_data
164 #define CHECK_ADD_OVERFLOW_INT64L(x, y) \
165 (((((x) > 0) && ((y) > 0) && ((x) > LLONG_MAX - (y))) || \
166 (((x) < 0) && ((y) < 0) && ((x) < LLONG_MIN - (y)))) \
167 ? 1 : 0)
168 /*
169 * Descriptor management.
170 */
171 struct fmsglist fmsghead; /* head of list of open files */
172 struct fmsglist fmsg_ithead; /* head of list of open files */
173 int nfiles; /* actual number of open files */
174
175
176 lck_grp_attr_t * file_lck_grp_attr;
177 lck_grp_t * file_lck_grp;
178 lck_attr_t * file_lck_attr;
179
180 lck_mtx_t * uipc_lock;
181
182
183 /*
184 * check_file_seek_range
185 *
186 * Description: Checks if seek offsets are in the range of 0 to LLONG_MAX.
187 *
188 * Parameters: fl Flock structure.
189 * cur_file_offset Current offset in the file.
190 *
191 * Returns: 0 on Success.
192 * EOVERFLOW on overflow.
193 * EINVAL on offset less than zero.
194 */
195
196 static int
197 check_file_seek_range(struct flock *fl, off_t cur_file_offset)
198 {
199 if (fl->l_whence == SEEK_CUR) {
200 /* Check if the start marker is beyond LLONG_MAX. */
201 if (CHECK_ADD_OVERFLOW_INT64L(fl->l_start, cur_file_offset)) {
202 /* Check if start marker is negative */
203 if (fl->l_start < 0) {
204 return EINVAL;
205 }
206 return EOVERFLOW;
207 }
208 /* Check if the start marker is negative. */
209 if (fl->l_start + cur_file_offset < 0) {
210 return EINVAL;
211 }
212 /* Check if end marker is beyond LLONG_MAX. */
213 if ((fl->l_len > 0) && (CHECK_ADD_OVERFLOW_INT64L(fl->l_start +
214 cur_file_offset, fl->l_len - 1))) {
215 return EOVERFLOW;
216 }
217 /* Check if the end marker is negative. */
218 if ((fl->l_len <= 0) && (fl->l_start + cur_file_offset +
219 fl->l_len < 0)) {
220 return EINVAL;
221 }
222 } else if (fl->l_whence == SEEK_SET) {
223 /* Check if the start marker is negative. */
224 if (fl->l_start < 0) {
225 return EINVAL;
226 }
227 /* Check if the end marker is beyond LLONG_MAX. */
228 if ((fl->l_len > 0) &&
229 CHECK_ADD_OVERFLOW_INT64L(fl->l_start, fl->l_len - 1)) {
230 return EOVERFLOW;
231 }
232 /* Check if the end marker is negative. */
233 if ((fl->l_len < 0) && fl->l_start + fl->l_len < 0) {
234 return EINVAL;
235 }
236 }
237 return 0;
238 }
239
240
241 /*
242 * file_lock_init
243 *
244 * Description: Initialize the file lock group and the uipc and flist locks
245 *
246 * Parameters: (void)
247 *
248 * Returns: void
249 *
250 * Notes: Called at system startup from bsd_init().
251 */
252 void
253 file_lock_init(void)
254 {
255 /* allocate file lock group attribute and group */
256 file_lck_grp_attr= lck_grp_attr_alloc_init();
257
258 file_lck_grp = lck_grp_alloc_init("file", file_lck_grp_attr);
259
260 /* Allocate file lock attribute */
261 file_lck_attr = lck_attr_alloc_init();
262
263 uipc_lock = lck_mtx_alloc_init(file_lck_grp, file_lck_attr);
264 }
265
266
267 /*
268 * proc_fdlock, proc_fdlock_spin
269 *
270 * Description: Lock to control access to the per process struct fileproc
271 * and struct filedesc
272 *
273 * Parameters: p Process to take the lock on
274 *
275 * Returns: void
276 *
277 * Notes: The lock is initialized in forkproc() and destroyed in
278 * reap_child_process().
279 */
280 void
281 proc_fdlock(proc_t p)
282 {
283 lck_mtx_lock(&p->p_fdmlock);
284 }
285
286 void
287 proc_fdlock_spin(proc_t p)
288 {
289 lck_mtx_lock_spin(&p->p_fdmlock);
290 }
291
292 void
293 proc_fdlock_assert(proc_t p, int assertflags)
294 {
295 lck_mtx_assert(&p->p_fdmlock, assertflags);
296 }
297
298
299 /*
300 * proc_fdunlock
301 *
302 * Description: Unlock the lock previously locked by a call to proc_fdlock()
303 *
304 * Parameters: p Process to drop the lock on
305 *
306 * Returns: void
307 */
308 void
309 proc_fdunlock(proc_t p)
310 {
311 lck_mtx_unlock(&p->p_fdmlock);
312 }
313
314
315 /*
316 * System calls on descriptors.
317 */
318
319
320 /*
321 * getdtablesize
322 *
323 * Description: Returns the per process maximum size of the descriptor table
324 *
325 * Parameters: p Process being queried
326 * retval Pointer to the call return area
327 *
328 * Returns: 0 Success
329 *
330 * Implicit returns:
331 * *retval (modified) Size of dtable
332 */
333 int
334 getdtablesize(proc_t p, __unused struct getdtablesize_args *uap, int32_t *retval)
335 {
336 proc_fdlock_spin(p);
337 *retval = min((int)p->p_rlimit[RLIMIT_NOFILE].rlim_cur, maxfiles);
338 proc_fdunlock(p);
339
340 return (0);
341 }
342
343
344 void
345 procfdtbl_reservefd(struct proc * p, int fd)
346 {
347 p->p_fd->fd_ofiles[fd] = NULL;
348 p->p_fd->fd_ofileflags[fd] |= UF_RESERVED;
349 }
350
351 void
352 procfdtbl_markclosefd(struct proc * p, int fd)
353 {
354 p->p_fd->fd_ofileflags[fd] |= (UF_RESERVED | UF_CLOSING);
355 }
356
357 void
358 procfdtbl_releasefd(struct proc * p, int fd, struct fileproc * fp)
359 {
360 if (fp != NULL)
361 p->p_fd->fd_ofiles[fd] = fp;
362 p->p_fd->fd_ofileflags[fd] &= ~UF_RESERVED;
363 if ((p->p_fd->fd_ofileflags[fd] & UF_RESVWAIT) == UF_RESVWAIT) {
364 p->p_fd->fd_ofileflags[fd] &= ~UF_RESVWAIT;
365 wakeup(&p->p_fd);
366 }
367 }
368
369 void
370 procfdtbl_waitfd(struct proc * p, int fd)
371 {
372 p->p_fd->fd_ofileflags[fd] |= UF_RESVWAIT;
373 msleep(&p->p_fd, &p->p_fdmlock, PRIBIO, "ftbl_waitfd", NULL);
374 }
375
376
377 void
378 procfdtbl_clearfd(struct proc * p, int fd)
379 {
380 int waiting;
381
382 waiting = (p->p_fd->fd_ofileflags[fd] & UF_RESVWAIT);
383 p->p_fd->fd_ofiles[fd] = NULL;
384 p->p_fd->fd_ofileflags[fd] = 0;
385 if ( waiting == UF_RESVWAIT) {
386 wakeup(&p->p_fd);
387 }
388 }
389
390 /*
391 * _fdrelse
392 *
393 * Description: Inline utility function to free an fd in a filedesc
394 *
395 * Parameters: fdp Pointer to filedesc fd lies in
396 * fd fd to free
397 * reserv fd should be reserved
398 *
399 * Returns: void
400 *
401 * Locks: Assumes proc_fdlock for process pointing to fdp is held by
402 * the caller
403 */
404 static void
405 _fdrelse(struct proc * p, int fd)
406 {
407 struct filedesc *fdp = p->p_fd;
408 int nfd = 0;
409
410 if (fd < fdp->fd_freefile)
411 fdp->fd_freefile = fd;
412 #if DIAGNOSTIC
413 if (fd > fdp->fd_lastfile)
414 panic("fdrelse: fd_lastfile inconsistent");
415 #endif
416 procfdtbl_clearfd(p, fd);
417
418 while ((nfd = fdp->fd_lastfile) > 0 &&
419 fdp->fd_ofiles[nfd] == NULL &&
420 !(fdp->fd_ofileflags[nfd] & UF_RESERVED))
421 fdp->fd_lastfile--;
422 }
423
424
425 int
426 fd_rdwr(
427 int fd,
428 enum uio_rw rw,
429 uint64_t base,
430 int64_t len,
431 enum uio_seg segflg,
432 off_t offset,
433 int io_flg,
434 int64_t *aresid)
435 {
436 struct fileproc *fp;
437 proc_t p;
438 int error = 0;
439 int flags = 0;
440 int spacetype;
441 uio_t auio = NULL;
442 char uio_buf[ UIO_SIZEOF(1) ];
443 struct vfs_context context = *(vfs_context_current());
444 bool wrote_some = false;
445
446 p = current_proc();
447
448 error = fp_lookup(p, fd, &fp, 0);
449 if (error)
450 return(error);
451
452 if (fp->f_type != DTYPE_VNODE && fp->f_type != DTYPE_PIPE && fp->f_type != DTYPE_SOCKET) {
453 error = EINVAL;
454 goto out;
455 }
456 if (rw == UIO_WRITE && !(fp->f_flag & FWRITE)) {
457 error = EBADF;
458 goto out;
459 }
460
461 if (rw == UIO_READ && !(fp->f_flag & FREAD)) {
462 error = EBADF;
463 goto out;
464 }
465
466 context.vc_ucred = fp->f_fglob->fg_cred;
467
468 if (UIO_SEG_IS_USER_SPACE(segflg))
469 spacetype = proc_is64bit(p) ? UIO_USERSPACE64 : UIO_USERSPACE32;
470 else
471 spacetype = UIO_SYSSPACE;
472
473 auio = uio_createwithbuffer(1, offset, spacetype, rw, &uio_buf[0], sizeof(uio_buf));
474
475 uio_addiov(auio, base, len);
476
477 if ( !(io_flg & IO_APPEND))
478 flags = FOF_OFFSET;
479
480 if (rw == UIO_WRITE) {
481 user_ssize_t orig_resid = uio_resid(auio);
482 error = fo_write(fp, auio, flags, &context);
483 wrote_some = uio_resid(auio) < orig_resid;
484 } else
485 error = fo_read(fp, auio, flags, &context);
486
487 if (aresid)
488 *aresid = uio_resid(auio);
489 else {
490 if (uio_resid(auio) && error == 0)
491 error = EIO;
492 }
493 out:
494 if (wrote_some)
495 fp_drop_written(p, fd, fp);
496 else
497 fp_drop(p, fd, fp, 0);
498
499 return error;
500 }
501
502
503
504 /*
505 * dup
506 *
507 * Description: Duplicate a file descriptor.
508 *
509 * Parameters: p Process performing the dup
510 * uap->fd The fd to dup
511 * retval Pointer to the call return area
512 *
513 * Returns: 0 Success
514 * !0 Errno
515 *
516 * Implicit returns:
517 * *retval (modified) The new descriptor
518 */
519 int
520 dup(proc_t p, struct dup_args *uap, int32_t *retval)
521 {
522 struct filedesc *fdp = p->p_fd;
523 int old = uap->fd;
524 int new, error;
525 struct fileproc *fp;
526
527 proc_fdlock(p);
528 if ( (error = fp_lookup(p, old, &fp, 1)) ) {
529 proc_fdunlock(p);
530 return(error);
531 }
532 if (FP_ISGUARDED(fp, GUARD_DUP)) {
533 error = fp_guard_exception(p, old, fp, kGUARD_EXC_DUP);
534 (void) fp_drop(p, old, fp, 1);
535 proc_fdunlock(p);
536 return (error);
537 }
538 if ( (error = fdalloc(p, 0, &new)) ) {
539 fp_drop(p, old, fp, 1);
540 proc_fdunlock(p);
541 return (error);
542 }
543 error = finishdup(p, fdp, old, new, 0, retval);
544 fp_drop(p, old, fp, 1);
545 proc_fdunlock(p);
546
547 if (ENTR_SHOULDTRACE && fp->f_type == DTYPE_SOCKET) {
548 KERNEL_ENERGYTRACE(kEnTrActKernSocket, DBG_FUNC_START,
549 new, 0, (int64_t)VM_KERNEL_ADDRPERM(fp->f_data));
550 }
551
552 return (error);
553 }
554
555 /*
556 * dup2
557 *
558 * Description: Duplicate a file descriptor to a particular value.
559 *
560 * Parameters: p Process performing the dup
561 * uap->from The fd to dup
562 * uap->to The fd to dup it to
563 * retval Pointer to the call return area
564 *
565 * Returns: 0 Success
566 * !0 Errno
567 *
568 * Implicit returns:
569 * *retval (modified) The new descriptor
570 */
571 int
572 dup2(proc_t p, struct dup2_args *uap, int32_t *retval)
573 {
574 struct filedesc *fdp = p->p_fd;
575 int old = uap->from, new = uap->to;
576 int i, error;
577 struct fileproc *fp, *nfp;
578
579 proc_fdlock(p);
580
581 startover:
582 if ( (error = fp_lookup(p, old, &fp, 1)) ) {
583 proc_fdunlock(p);
584 return(error);
585 }
586 if (FP_ISGUARDED(fp, GUARD_DUP)) {
587 error = fp_guard_exception(p, old, fp, kGUARD_EXC_DUP);
588 (void) fp_drop(p, old, fp, 1);
589 proc_fdunlock(p);
590 return (error);
591 }
592 if (new < 0 ||
593 (rlim_t)new >= p->p_rlimit[RLIMIT_NOFILE].rlim_cur ||
594 new >= maxfiles) {
595 fp_drop(p, old, fp, 1);
596 proc_fdunlock(p);
597 return (EBADF);
598 }
599 if (old == new) {
600 fp_drop(p, old, fp, 1);
601 *retval = new;
602 proc_fdunlock(p);
603 return (0);
604 }
605 if (new < 0 || new >= fdp->fd_nfiles) {
606 if ( (error = fdalloc(p, new, &i)) ) {
607 fp_drop(p, old, fp, 1);
608 proc_fdunlock(p);
609 return (error);
610 }
611 if (new != i) {
612 fdrelse(p, i);
613 goto closeit;
614 }
615 } else {
616 closeit:
617 while ((fdp->fd_ofileflags[new] & UF_RESERVED) == UF_RESERVED) {
618 fp_drop(p, old, fp, 1);
619 procfdtbl_waitfd(p, new);
620 #if DIAGNOSTIC
621 proc_fdlock_assert(p, LCK_MTX_ASSERT_OWNED);
622 #endif
623 goto startover;
624 }
625
626 if ((fdp->fd_ofiles[new] != NULL) &&
627 ((error = fp_lookup(p, new, &nfp, 1)) == 0)) {
628 fp_drop(p, old, fp, 1);
629 if (FP_ISGUARDED(nfp, GUARD_CLOSE)) {
630 error = fp_guard_exception(p,
631 new, nfp, kGUARD_EXC_CLOSE);
632 (void) fp_drop(p, new, nfp, 1);
633 proc_fdunlock(p);
634 return (error);
635 }
636 (void)close_internal_locked(p, new, nfp, FD_DUP2RESV);
637 #if DIAGNOSTIC
638 proc_fdlock_assert(p, LCK_MTX_ASSERT_OWNED);
639 #endif
640 procfdtbl_clearfd(p, new);
641 goto startover;
642 } else {
643 #if DIAGNOSTIC
644 if (fdp->fd_ofiles[new] != NULL)
645 panic("dup2: no ref on fileproc %d", new);
646 #endif
647 procfdtbl_reservefd(p, new);
648 }
649
650 #if DIAGNOSTIC
651 proc_fdlock_assert(p, LCK_MTX_ASSERT_OWNED);
652 #endif
653
654 }
655 #if DIAGNOSTIC
656 if (fdp->fd_ofiles[new] != 0)
657 panic("dup2: overwriting fd_ofiles with new %d", new);
658 if ((fdp->fd_ofileflags[new] & UF_RESERVED) == 0)
659 panic("dup2: unreserved fileflags with new %d", new);
660 #endif
661 error = finishdup(p, fdp, old, new, 0, retval);
662 fp_drop(p, old, fp, 1);
663 proc_fdunlock(p);
664
665 return(error);
666 }
667
668
669 /*
670 * fcntl
671 *
672 * Description: The file control system call.
673 *
674 * Parameters: p Process performing the fcntl
675 * uap->fd The fd to operate against
676 * uap->cmd The command to perform
677 * uap->arg Pointer to the command argument
678 * retval Pointer to the call return area
679 *
680 * Returns: 0 Success
681 * !0 Errno (see fcntl_nocancel)
682 *
683 * Implicit returns:
684 * *retval (modified) fcntl return value (if any)
685 *
686 * Notes: This system call differs from fcntl_nocancel() in that it
687 * tests for cancellation prior to performing a potentially
688 * blocking operation.
689 */
690 int
691 fcntl(proc_t p, struct fcntl_args *uap, int32_t *retval)
692 {
693 __pthread_testcancel(1);
694 return(fcntl_nocancel(p, (struct fcntl_nocancel_args *)uap, retval));
695 }
696
697
698 /*
699 * fcntl_nocancel
700 *
701 * Description: A non-cancel-testing file control system call.
702 *
703 * Parameters: p Process performing the fcntl
704 * uap->fd The fd to operate against
705 * uap->cmd The command to perform
706 * uap->arg Pointer to the command argument
707 * retval Pointer to the call return area
708 *
709 * Returns: 0 Success
710 * EINVAL
711 * fp_lookup:EBADF Bad file descriptor
712 * [F_DUPFD]
713 * fdalloc:EMFILE
714 * fdalloc:ENOMEM
715 * finishdup:EBADF
716 * finishdup:ENOMEM
717 * [F_SETOWN]
718 * ESRCH
719 * [F_SETLK]
720 * EBADF
721 * EOVERFLOW
722 * copyin:EFAULT
723 * vnode_getwithref:???
724 * VNOP_ADVLOCK:???
725 * msleep:ETIMEDOUT
726 * [F_GETLK]
727 * EBADF
728 * EOVERFLOW
729 * copyin:EFAULT
730 * copyout:EFAULT
731 * vnode_getwithref:???
732 * VNOP_ADVLOCK:???
733 * [F_PREALLOCATE]
734 * EBADF
735 * EINVAL
736 * copyin:EFAULT
737 * copyout:EFAULT
738 * vnode_getwithref:???
739 * VNOP_ALLOCATE:???
740 * [F_SETSIZE,F_RDADVISE]
741 * EBADF
742 * copyin:EFAULT
743 * vnode_getwithref:???
744 * [F_RDAHEAD,F_NOCACHE]
745 * EBADF
746 * vnode_getwithref:???
747 * [???]
748 *
749 * Implicit returns:
750 * *retval (modified) fcntl return value (if any)
751 */
752 int
753 fcntl_nocancel(proc_t p, struct fcntl_nocancel_args *uap, int32_t *retval)
754 {
755 int fd = uap->fd;
756 struct filedesc *fdp = p->p_fd;
757 struct fileproc *fp;
758 char *pop;
759 struct vnode *vp = NULLVP; /* for AUDIT_ARG() at end */
760 int i, tmp, error, error2, flg = 0;
761 struct flock fl;
762 struct flocktimeout fltimeout;
763 struct timespec *timeout = NULL;
764 struct vfs_context context;
765 off_t offset;
766 int newmin;
767 daddr64_t lbn, bn;
768 unsigned int fflag;
769 user_addr_t argp;
770 boolean_t is64bit;
771
772 AUDIT_ARG(fd, uap->fd);
773 AUDIT_ARG(cmd, uap->cmd);
774
775 proc_fdlock(p);
776 if ( (error = fp_lookup(p, fd, &fp, 1)) ) {
777 proc_fdunlock(p);
778 return(error);
779 }
780 context.vc_thread = current_thread();
781 context.vc_ucred = fp->f_cred;
782
783 is64bit = proc_is64bit(p);
784 if (is64bit) {
785 argp = uap->arg;
786 }
787 else {
788 /*
789 * Since the arg parameter is defined as a long but may be
790 * either a long or a pointer we must take care to handle
791 * sign extension issues. Our sys call munger will sign
792 * extend a long when we are called from a 32-bit process.
793 * Since we can never have an address greater than 32-bits
794 * from a 32-bit process we lop off the top 32-bits to avoid
795 * getting the wrong address
796 */
797 argp = CAST_USER_ADDR_T((uint32_t)uap->arg);
798 }
799
800 pop = &fdp->fd_ofileflags[fd];
801
802 #if CONFIG_MACF
803 error = mac_file_check_fcntl(proc_ucred(p), fp->f_fglob, uap->cmd,
804 uap->arg);
805 if (error)
806 goto out;
807 #endif
808
809 switch (uap->cmd) {
810
811 case F_DUPFD:
812 case F_DUPFD_CLOEXEC:
813 if (FP_ISGUARDED(fp, GUARD_DUP)) {
814 error = fp_guard_exception(p, fd, fp, kGUARD_EXC_DUP);
815 goto out;
816 }
817 newmin = CAST_DOWN_EXPLICIT(int, uap->arg); /* arg is an int, so we won't lose bits */
818 AUDIT_ARG(value32, newmin);
819 if ((u_int)newmin >= p->p_rlimit[RLIMIT_NOFILE].rlim_cur ||
820 newmin >= maxfiles) {
821 error = EINVAL;
822 goto out;
823 }
824 if ( (error = fdalloc(p, newmin, &i)) )
825 goto out;
826 error = finishdup(p, fdp, fd, i,
827 uap->cmd == F_DUPFD_CLOEXEC ? UF_EXCLOSE : 0, retval);
828 goto out;
829
830 case F_GETFD:
831 *retval = (*pop & UF_EXCLOSE)? FD_CLOEXEC : 0;
832 error = 0;
833 goto out;
834
835 case F_SETFD:
836 AUDIT_ARG(value32, uap->arg);
837 if (uap->arg & FD_CLOEXEC)
838 *pop |= UF_EXCLOSE;
839 else {
840 if (FILEPROC_TYPE(fp) == FTYPE_GUARDED) {
841 error = fp_guard_exception(p,
842 fd, fp, kGUARD_EXC_NOCLOEXEC);
843 goto out;
844 }
845 *pop &= ~UF_EXCLOSE;
846 }
847 error = 0;
848 goto out;
849
850 case F_GETFL:
851 *retval = OFLAGS(fp->f_flag);
852 error = 0;
853 goto out;
854
855 case F_SETFL:
856 fp->f_flag &= ~FCNTLFLAGS;
857 tmp = CAST_DOWN_EXPLICIT(int, uap->arg); /* arg is an int, so we won't lose bits */
858 AUDIT_ARG(value32, tmp);
859 fp->f_flag |= FFLAGS(tmp) & FCNTLFLAGS;
860 tmp = fp->f_flag & FNONBLOCK;
861 error = fo_ioctl(fp, FIONBIO, (caddr_t)&tmp, &context);
862 if (error)
863 goto out;
864 tmp = fp->f_flag & FASYNC;
865 error = fo_ioctl(fp, FIOASYNC, (caddr_t)&tmp, &context);
866 if (!error)
867 goto out;
868 fp->f_flag &= ~FNONBLOCK;
869 tmp = 0;
870 (void)fo_ioctl(fp, FIONBIO, (caddr_t)&tmp, &context);
871 goto out;
872
873 case F_GETOWN:
874 if (fp->f_type == DTYPE_SOCKET) {
875 *retval = ((struct socket *)fp->f_data)->so_pgid;
876 error = 0;
877 goto out;
878 }
879 error = fo_ioctl(fp, (int)TIOCGPGRP, (caddr_t)retval, &context);
880 *retval = -*retval;
881 goto out;
882
883 case F_SETOWN:
884 tmp = CAST_DOWN_EXPLICIT(pid_t, uap->arg); /* arg is an int, so we won't lose bits */
885 AUDIT_ARG(value32, tmp);
886 if (fp->f_type == DTYPE_SOCKET) {
887 ((struct socket *)fp->f_data)->so_pgid = tmp;
888 error =0;
889 goto out;
890 }
891 if (fp->f_type == DTYPE_PIPE) {
892 error = fo_ioctl(fp, TIOCSPGRP, (caddr_t)&tmp, &context);
893 goto out;
894 }
895
896 if (tmp <= 0) {
897 tmp = -tmp;
898 } else {
899 proc_t p1 = proc_find(tmp);
900 if (p1 == 0) {
901 error = ESRCH;
902 goto out;
903 }
904 tmp = (int)p1->p_pgrpid;
905 proc_rele(p1);
906 }
907 error = fo_ioctl(fp, (int)TIOCSPGRP, (caddr_t)&tmp, &context);
908 goto out;
909
910 case F_SETNOSIGPIPE:
911 tmp = CAST_DOWN_EXPLICIT(int, uap->arg);
912 if (fp->f_type == DTYPE_SOCKET) {
913 #if SOCKETS
914 error = sock_setsockopt((struct socket *)fp->f_data,
915 SOL_SOCKET, SO_NOSIGPIPE, &tmp, sizeof (tmp));
916 #else
917 error = EINVAL;
918 #endif
919 } else {
920 struct fileglob *fg = fp->f_fglob;
921
922 lck_mtx_lock_spin(&fg->fg_lock);
923 if (tmp)
924 fg->fg_lflags |= FG_NOSIGPIPE;
925 else
926 fg->fg_lflags &= FG_NOSIGPIPE;
927 lck_mtx_unlock(&fg->fg_lock);
928 error = 0;
929 }
930 goto out;
931
932 case F_GETNOSIGPIPE:
933 if (fp->f_type == DTYPE_SOCKET) {
934 #if SOCKETS
935 int retsize = sizeof (*retval);
936 error = sock_getsockopt((struct socket *)fp->f_data,
937 SOL_SOCKET, SO_NOSIGPIPE, retval, &retsize);
938 #else
939 error = EINVAL;
940 #endif
941 } else {
942 *retval = (fp->f_fglob->fg_lflags & FG_NOSIGPIPE) ?
943 1 : 0;
944 error = 0;
945 }
946 goto out;
947
948 case F_SETCONFINED:
949 /*
950 * If this is the only reference to this fglob in the process
951 * and it's already marked as close-on-fork then mark it as
952 * (immutably) "confined" i.e. any fd that points to it will
953 * forever be close-on-fork, and attempts to use an IPC
954 * mechanism to move the descriptor elsewhere will fail.
955 */
956 if (CAST_DOWN_EXPLICIT(int, uap->arg)) {
957 struct fileglob *fg = fp->f_fglob;
958
959 lck_mtx_lock_spin(&fg->fg_lock);
960 if (fg->fg_lflags & FG_CONFINED)
961 error = 0;
962 else if (1 != fg->fg_count)
963 error = EAGAIN; /* go close the dup .. */
964 else if (UF_FORKCLOSE == (*pop & UF_FORKCLOSE)) {
965 fg->fg_lflags |= FG_CONFINED;
966 error = 0;
967 } else
968 error = EBADF; /* open without O_CLOFORK? */
969 lck_mtx_unlock(&fg->fg_lock);
970 } else {
971 /*
972 * Other subsystems may have built on the immutability
973 * of FG_CONFINED; clearing it may be tricky.
974 */
975 error = EPERM; /* immutable */
976 }
977 goto out;
978
979 case F_GETCONFINED:
980 *retval = (fp->f_fglob->fg_lflags & FG_CONFINED) ? 1 : 0;
981 error = 0;
982 goto out;
983
984 case F_SETLKWTIMEOUT:
985 case F_SETLKW:
986 case F_OFD_SETLKWTIMEOUT:
987 case F_OFD_SETLKW:
988 flg |= F_WAIT;
989 /* Fall into F_SETLK */
990
991 case F_SETLK:
992 case F_OFD_SETLK:
993 if (fp->f_type != DTYPE_VNODE) {
994 error = EBADF;
995 goto out;
996 }
997 vp = (struct vnode *)fp->f_data;
998
999 fflag = fp->f_flag;
1000 offset = fp->f_offset;
1001 proc_fdunlock(p);
1002
1003 /* Copy in the lock structure */
1004 if (F_SETLKWTIMEOUT == uap->cmd ||
1005 F_OFD_SETLKWTIMEOUT == uap->cmd) {
1006 error = copyin(argp, (caddr_t) &fltimeout, sizeof(fltimeout));
1007 if (error) {
1008 goto outdrop;
1009 }
1010 fl = fltimeout.fl;
1011 timeout = &fltimeout.timeout;
1012 } else {
1013 error = copyin(argp, (caddr_t)&fl, sizeof(fl));
1014 if (error) {
1015 goto outdrop;
1016 }
1017 }
1018
1019 /* Check starting byte and ending byte for EOVERFLOW in SEEK_CUR */
1020 /* and ending byte for EOVERFLOW in SEEK_SET */
1021 error = check_file_seek_range(&fl, offset);
1022 if (error) {
1023 goto outdrop;
1024 }
1025
1026 if ( (error = vnode_getwithref(vp)) ) {
1027 goto outdrop;
1028 }
1029 if (fl.l_whence == SEEK_CUR)
1030 fl.l_start += offset;
1031
1032 #if CONFIG_MACF
1033 error = mac_file_check_lock(proc_ucred(p), fp->f_fglob,
1034 F_SETLK, &fl);
1035 if (error) {
1036 (void)vnode_put(vp);
1037 goto outdrop;
1038 }
1039 #endif
1040 switch (uap->cmd) {
1041 case F_OFD_SETLK:
1042 case F_OFD_SETLKW:
1043 case F_OFD_SETLKWTIMEOUT:
1044 flg |= F_OFD_LOCK;
1045 switch (fl.l_type) {
1046 case F_RDLCK:
1047 if ((fflag & FREAD) == 0) {
1048 error = EBADF;
1049 break;
1050 }
1051 error = VNOP_ADVLOCK(vp, (caddr_t)fp->f_fglob,
1052 F_SETLK, &fl, flg, &context, timeout);
1053 break;
1054 case F_WRLCK:
1055 if ((fflag & FWRITE) == 0) {
1056 error = EBADF;
1057 break;
1058 }
1059 error = VNOP_ADVLOCK(vp, (caddr_t)fp->f_fglob,
1060 F_SETLK, &fl, flg, &context, timeout);
1061 break;
1062 case F_UNLCK:
1063 error = VNOP_ADVLOCK(vp, (caddr_t)fp->f_fglob,
1064 F_UNLCK, &fl, F_OFD_LOCK, &context,
1065 timeout);
1066 break;
1067 default:
1068 error = EINVAL;
1069 break;
1070 }
1071 if (0 == error &&
1072 (F_RDLCK == fl.l_type || F_WRLCK == fl.l_type)) {
1073 struct fileglob *fg = fp->f_fglob;
1074
1075 /*
1076 * arrange F_UNLCK on last close (once
1077 * set, FG_HAS_OFDLOCK is immutable)
1078 */
1079 if ((fg->fg_lflags & FG_HAS_OFDLOCK) == 0) {
1080 lck_mtx_lock_spin(&fg->fg_lock);
1081 fg->fg_lflags |= FG_HAS_OFDLOCK;
1082 lck_mtx_unlock(&fg->fg_lock);
1083 }
1084 }
1085 break;
1086 default:
1087 flg |= F_POSIX;
1088 switch (fl.l_type) {
1089 case F_RDLCK:
1090 if ((fflag & FREAD) == 0) {
1091 error = EBADF;
1092 break;
1093 }
1094 // XXX UInt32 unsafe for LP64 kernel
1095 OSBitOrAtomic(P_LADVLOCK, &p->p_ladvflag);
1096 error = VNOP_ADVLOCK(vp, (caddr_t)p,
1097 F_SETLK, &fl, flg, &context, timeout);
1098 break;
1099 case F_WRLCK:
1100 if ((fflag & FWRITE) == 0) {
1101 error = EBADF;
1102 break;
1103 }
1104 // XXX UInt32 unsafe for LP64 kernel
1105 OSBitOrAtomic(P_LADVLOCK, &p->p_ladvflag);
1106 error = VNOP_ADVLOCK(vp, (caddr_t)p,
1107 F_SETLK, &fl, flg, &context, timeout);
1108 break;
1109 case F_UNLCK:
1110 error = VNOP_ADVLOCK(vp, (caddr_t)p,
1111 F_UNLCK, &fl, F_POSIX, &context, timeout);
1112 break;
1113 default:
1114 error = EINVAL;
1115 break;
1116 }
1117 break;
1118 }
1119 (void) vnode_put(vp);
1120 goto outdrop;
1121
1122 case F_GETLK:
1123 case F_OFD_GETLK:
1124 if (fp->f_type != DTYPE_VNODE) {
1125 error = EBADF;
1126 goto out;
1127 }
1128 vp = (struct vnode *)fp->f_data;
1129
1130 offset = fp->f_offset;
1131 proc_fdunlock(p);
1132
1133 /* Copy in the lock structure */
1134 error = copyin(argp, (caddr_t)&fl, sizeof(fl));
1135 if (error)
1136 goto outdrop;
1137
1138 /* Check starting byte and ending byte for EOVERFLOW in SEEK_CUR */
1139 /* and ending byte for EOVERFLOW in SEEK_SET */
1140 error = check_file_seek_range(&fl, offset);
1141 if (error) {
1142 goto outdrop;
1143 }
1144
1145 if ((fl.l_whence == SEEK_SET) && (fl.l_start < 0)) {
1146 error = EINVAL;
1147 goto outdrop;
1148 }
1149
1150 switch (fl.l_type) {
1151 case F_RDLCK:
1152 case F_UNLCK:
1153 case F_WRLCK:
1154 break;
1155 default:
1156 error = EINVAL;
1157 goto outdrop;
1158 }
1159
1160 switch (fl.l_whence) {
1161 case SEEK_CUR:
1162 case SEEK_SET:
1163 case SEEK_END:
1164 break;
1165 default:
1166 error = EINVAL;
1167 goto outdrop;
1168 }
1169
1170 if ( (error = vnode_getwithref(vp)) == 0 ) {
1171 if (fl.l_whence == SEEK_CUR)
1172 fl.l_start += offset;
1173
1174 #if CONFIG_MACF
1175 error = mac_file_check_lock(proc_ucred(p), fp->f_fglob,
1176 uap->cmd, &fl);
1177 if (error == 0)
1178 #endif
1179 switch (uap->cmd) {
1180 case F_OFD_GETLK:
1181 error = VNOP_ADVLOCK(vp, (caddr_t)fp->f_fglob,
1182 F_GETLK, &fl, F_OFD_LOCK, &context, NULL);
1183 break;
1184 case F_OFD_GETLKPID:
1185 error = VNOP_ADVLOCK(vp, (caddr_t)fp->f_fglob,
1186 F_GETLKPID, &fl, F_OFD_LOCK, &context, NULL);
1187 break;
1188 default:
1189 error = VNOP_ADVLOCK(vp, (caddr_t)p,
1190 uap->cmd, &fl, F_POSIX, &context, NULL);
1191 break;
1192 }
1193
1194 (void)vnode_put(vp);
1195
1196 if (error == 0)
1197 error = copyout((caddr_t)&fl, argp, sizeof(fl));
1198 }
1199 goto outdrop;
1200
1201 case F_PREALLOCATE: {
1202 fstore_t alloc_struct; /* structure for allocate command */
1203 u_int32_t alloc_flags = 0;
1204
1205 if (fp->f_type != DTYPE_VNODE) {
1206 error = EBADF;
1207 goto out;
1208 }
1209
1210 vp = (struct vnode *)fp->f_data;
1211 proc_fdunlock(p);
1212
1213 /* make sure that we have write permission */
1214 if ((fp->f_flag & FWRITE) == 0) {
1215 error = EBADF;
1216 goto outdrop;
1217 }
1218
1219 error = copyin(argp, (caddr_t)&alloc_struct, sizeof(alloc_struct));
1220 if (error)
1221 goto outdrop;
1222
1223 /* now set the space allocated to 0 */
1224 alloc_struct.fst_bytesalloc = 0;
1225
1226 /*
1227 * Do some simple parameter checking
1228 */
1229
1230 /* set up the flags */
1231
1232 alloc_flags |= PREALLOCATE;
1233
1234 if (alloc_struct.fst_flags & F_ALLOCATECONTIG)
1235 alloc_flags |= ALLOCATECONTIG;
1236
1237 if (alloc_struct.fst_flags & F_ALLOCATEALL)
1238 alloc_flags |= ALLOCATEALL;
1239
1240 /*
1241 * Do any position mode specific stuff. The only
1242 * position mode supported now is PEOFPOSMODE
1243 */
1244
1245 switch (alloc_struct.fst_posmode) {
1246
1247 case F_PEOFPOSMODE:
1248 if (alloc_struct.fst_offset != 0) {
1249 error = EINVAL;
1250 goto outdrop;
1251 }
1252
1253 alloc_flags |= ALLOCATEFROMPEOF;
1254 break;
1255
1256 case F_VOLPOSMODE:
1257 if (alloc_struct.fst_offset <= 0) {
1258 error = EINVAL;
1259 goto outdrop;
1260 }
1261
1262 alloc_flags |= ALLOCATEFROMVOL;
1263 break;
1264
1265 default: {
1266 error = EINVAL;
1267 goto outdrop;
1268 }
1269 }
1270 if ( (error = vnode_getwithref(vp)) == 0 ) {
1271 /*
1272 * call allocate to get the space
1273 */
1274 error = VNOP_ALLOCATE(vp,alloc_struct.fst_length,alloc_flags,
1275 &alloc_struct.fst_bytesalloc, alloc_struct.fst_offset,
1276 &context);
1277 (void)vnode_put(vp);
1278
1279 error2 = copyout((caddr_t)&alloc_struct, argp, sizeof(alloc_struct));
1280
1281 if (error == 0)
1282 error = error2;
1283 }
1284 goto outdrop;
1285
1286 }
1287 case F_SETSIZE:
1288 if (fp->f_type != DTYPE_VNODE) {
1289 error = EBADF;
1290 goto out;
1291 }
1292 vp = (struct vnode *)fp->f_data;
1293 proc_fdunlock(p);
1294
1295 error = copyin(argp, (caddr_t)&offset, sizeof (off_t));
1296 if (error)
1297 goto outdrop;
1298 AUDIT_ARG(value64, offset);
1299
1300 error = vnode_getwithref(vp);
1301 if (error)
1302 goto outdrop;
1303
1304 #if CONFIG_MACF
1305 error = mac_vnode_check_truncate(&context,
1306 fp->f_fglob->fg_cred, vp);
1307 if (error) {
1308 (void)vnode_put(vp);
1309 goto outdrop;
1310 }
1311 #endif
1312 /*
1313 * Make sure that we are root. Growing a file
1314 * without zero filling the data is a security hole
1315 * root would have access anyway so we'll allow it
1316 */
1317 if (!kauth_cred_issuser(kauth_cred_get())) {
1318 error = EACCES;
1319 } else {
1320 /*
1321 * set the file size
1322 */
1323 error = vnode_setsize(vp, offset, IO_NOZEROFILL,
1324 &context);
1325
1326 #if CONFIG_MACF
1327 if (error == 0)
1328 mac_vnode_notify_truncate(&context, fp->f_fglob->fg_cred, vp);
1329 #endif
1330 }
1331
1332 (void)vnode_put(vp);
1333 goto outdrop;
1334
1335 case F_RDAHEAD:
1336 if (fp->f_type != DTYPE_VNODE) {
1337 error = EBADF;
1338 goto out;
1339 }
1340 if (uap->arg)
1341 fp->f_fglob->fg_flag &= ~FNORDAHEAD;
1342 else
1343 fp->f_fglob->fg_flag |= FNORDAHEAD;
1344
1345 goto out;
1346
1347 case F_NOCACHE:
1348 if (fp->f_type != DTYPE_VNODE) {
1349 error = EBADF;
1350 goto out;
1351 }
1352 if (uap->arg)
1353 fp->f_fglob->fg_flag |= FNOCACHE;
1354 else
1355 fp->f_fglob->fg_flag &= ~FNOCACHE;
1356
1357 goto out;
1358
1359 case F_NODIRECT:
1360 if (fp->f_type != DTYPE_VNODE) {
1361 error = EBADF;
1362 goto out;
1363 }
1364 if (uap->arg)
1365 fp->f_fglob->fg_flag |= FNODIRECT;
1366 else
1367 fp->f_fglob->fg_flag &= ~FNODIRECT;
1368
1369 goto out;
1370
1371 case F_SINGLE_WRITER:
1372 if (fp->f_type != DTYPE_VNODE) {
1373 error = EBADF;
1374 goto out;
1375 }
1376 if (uap->arg)
1377 fp->f_fglob->fg_flag |= FSINGLE_WRITER;
1378 else
1379 fp->f_fglob->fg_flag &= ~FSINGLE_WRITER;
1380
1381 goto out;
1382
1383 case F_GLOBAL_NOCACHE:
1384 if (fp->f_type != DTYPE_VNODE) {
1385 error = EBADF;
1386 goto out;
1387 }
1388 vp = (struct vnode *)fp->f_data;
1389 proc_fdunlock(p);
1390
1391 if ( (error = vnode_getwithref(vp)) == 0 ) {
1392
1393 *retval = vnode_isnocache(vp);
1394
1395 if (uap->arg)
1396 vnode_setnocache(vp);
1397 else
1398 vnode_clearnocache(vp);
1399
1400 (void)vnode_put(vp);
1401 }
1402 goto outdrop;
1403
1404 case F_CHECK_OPENEVT:
1405 if (fp->f_type != DTYPE_VNODE) {
1406 error = EBADF;
1407 goto out;
1408 }
1409 vp = (struct vnode *)fp->f_data;
1410 proc_fdunlock(p);
1411
1412 if ( (error = vnode_getwithref(vp)) == 0 ) {
1413
1414 *retval = vnode_is_openevt(vp);
1415
1416 if (uap->arg)
1417 vnode_set_openevt(vp);
1418 else
1419 vnode_clear_openevt(vp);
1420
1421 (void)vnode_put(vp);
1422 }
1423 goto outdrop;
1424
1425 case F_RDADVISE: {
1426 struct radvisory ra_struct;
1427
1428 if (fp->f_type != DTYPE_VNODE) {
1429 error = EBADF;
1430 goto out;
1431 }
1432 vp = (struct vnode *)fp->f_data;
1433 proc_fdunlock(p);
1434
1435 if ( (error = copyin(argp, (caddr_t)&ra_struct, sizeof(ra_struct))) )
1436 goto outdrop;
1437 if ( (error = vnode_getwithref(vp)) == 0 ) {
1438 error = VNOP_IOCTL(vp, F_RDADVISE, (caddr_t)&ra_struct, 0, &context);
1439
1440 (void)vnode_put(vp);
1441 }
1442 goto outdrop;
1443 }
1444
1445 case F_FLUSH_DATA:
1446
1447 if (fp->f_type != DTYPE_VNODE) {
1448 error = EBADF;
1449 goto out;
1450 }
1451 vp = (struct vnode *)fp->f_data;
1452 proc_fdunlock(p);
1453
1454 if ( (error = vnode_getwithref(vp)) == 0 ) {
1455 error = cluster_push(vp, 0);
1456
1457 (void)vnode_put(vp);
1458 }
1459 goto outdrop;
1460
1461 case F_LOG2PHYS:
1462 case F_LOG2PHYS_EXT: {
1463 struct log2phys l2p_struct; /* structure for allocate command */
1464 int devBlockSize;
1465
1466 off_t file_offset = 0;
1467 size_t a_size = 0;
1468 size_t run = 0;
1469
1470 if (uap->cmd == F_LOG2PHYS_EXT) {
1471 error = copyin(argp, (caddr_t)&l2p_struct, sizeof(l2p_struct));
1472 if (error)
1473 goto out;
1474 file_offset = l2p_struct.l2p_devoffset;
1475 } else {
1476 file_offset = fp->f_offset;
1477 }
1478 if (fp->f_type != DTYPE_VNODE) {
1479 error = EBADF;
1480 goto out;
1481 }
1482 vp = (struct vnode *)fp->f_data;
1483 proc_fdunlock(p);
1484 if ( (error = vnode_getwithref(vp)) ) {
1485 goto outdrop;
1486 }
1487 error = VNOP_OFFTOBLK(vp, file_offset, &lbn);
1488 if (error) {
1489 (void)vnode_put(vp);
1490 goto outdrop;
1491 }
1492 error = VNOP_BLKTOOFF(vp, lbn, &offset);
1493 if (error) {
1494 (void)vnode_put(vp);
1495 goto outdrop;
1496 }
1497 devBlockSize = vfs_devblocksize(vnode_mount(vp));
1498 if (uap->cmd == F_LOG2PHYS_EXT) {
1499 if (l2p_struct.l2p_contigbytes < 0) {
1500 vnode_put(vp);
1501 error = EINVAL;
1502 goto outdrop;
1503 }
1504
1505 a_size = MIN((uint64_t)l2p_struct.l2p_contigbytes, SIZE_MAX);
1506 } else {
1507 a_size = devBlockSize;
1508 }
1509
1510 error = VNOP_BLOCKMAP(vp, offset, a_size, &bn, &run, NULL, 0, &context);
1511
1512 (void)vnode_put(vp);
1513
1514 if (!error) {
1515 l2p_struct.l2p_flags = 0; /* for now */
1516 if (uap->cmd == F_LOG2PHYS_EXT) {
1517 l2p_struct.l2p_contigbytes = run - (file_offset - offset);
1518 } else {
1519 l2p_struct.l2p_contigbytes = 0; /* for now */
1520 }
1521
1522 /*
1523 * The block number being -1 suggests that the file offset is not backed
1524 * by any real blocks on-disk. As a result, just let it be passed back up wholesale.
1525 */
1526 if (bn == -1) {
1527 /* Don't multiply it by the block size */
1528 l2p_struct.l2p_devoffset = bn;
1529 }
1530 else {
1531 l2p_struct.l2p_devoffset = bn * devBlockSize;
1532 l2p_struct.l2p_devoffset += file_offset - offset;
1533 }
1534 error = copyout((caddr_t)&l2p_struct, argp, sizeof(l2p_struct));
1535 }
1536 goto outdrop;
1537 }
1538 case F_GETPATH: {
1539 char *pathbufp;
1540 int pathlen;
1541
1542 if (fp->f_type != DTYPE_VNODE) {
1543 error = EBADF;
1544 goto out;
1545 }
1546 vp = (struct vnode *)fp->f_data;
1547 proc_fdunlock(p);
1548
1549 pathlen = MAXPATHLEN;
1550 MALLOC(pathbufp, char *, pathlen, M_TEMP, M_WAITOK);
1551 if (pathbufp == NULL) {
1552 error = ENOMEM;
1553 goto outdrop;
1554 }
1555 if ( (error = vnode_getwithref(vp)) == 0 ) {
1556 error = vn_getpath(vp, pathbufp, &pathlen);
1557 (void)vnode_put(vp);
1558
1559 if (error == 0)
1560 error = copyout((caddr_t)pathbufp, argp, pathlen);
1561 }
1562 FREE(pathbufp, M_TEMP);
1563 goto outdrop;
1564 }
1565
1566 case F_PATHPKG_CHECK: {
1567 char *pathbufp;
1568 size_t pathlen;
1569
1570 if (fp->f_type != DTYPE_VNODE) {
1571 error = EBADF;
1572 goto out;
1573 }
1574 vp = (struct vnode *)fp->f_data;
1575 proc_fdunlock(p);
1576
1577 pathlen = MAXPATHLEN;
1578 pathbufp = kalloc(MAXPATHLEN);
1579
1580 if ( (error = copyinstr(argp, pathbufp, MAXPATHLEN, &pathlen)) == 0 ) {
1581 if ( (error = vnode_getwithref(vp)) == 0 ) {
1582 AUDIT_ARG(text, pathbufp);
1583 error = vn_path_package_check(vp, pathbufp, pathlen, retval);
1584
1585 (void)vnode_put(vp);
1586 }
1587 }
1588 kfree(pathbufp, MAXPATHLEN);
1589 goto outdrop;
1590 }
1591
1592 case F_CHKCLEAN: // used by regression tests to see if all dirty pages got cleaned by fsync()
1593 case F_FULLFSYNC: // fsync + flush the journal + DKIOCSYNCHRONIZE
1594 case F_BARRIERFSYNC: // fsync + barrier
1595 case F_FREEZE_FS: // freeze all other fs operations for the fs of this fd
1596 case F_THAW_FS: { // thaw all frozen fs operations for the fs of this fd
1597 if (fp->f_type != DTYPE_VNODE) {
1598 error = EBADF;
1599 goto out;
1600 }
1601 vp = (struct vnode *)fp->f_data;
1602 proc_fdunlock(p);
1603
1604 if ( (error = vnode_getwithref(vp)) == 0 ) {
1605 error = VNOP_IOCTL(vp, uap->cmd, (caddr_t)NULL, 0, &context);
1606
1607 (void)vnode_put(vp);
1608 }
1609 break;
1610 }
1611
1612 /*
1613 * SPI (private) for opening a file starting from a dir fd
1614 */
1615 case F_OPENFROM: {
1616 struct user_fopenfrom fopen;
1617 struct vnode_attr va;
1618 struct nameidata nd;
1619 int cmode;
1620
1621 /* Check if this isn't a valid file descriptor */
1622 if ((fp->f_type != DTYPE_VNODE) ||
1623 (fp->f_flag & FREAD) == 0) {
1624 error = EBADF;
1625 goto out;
1626 }
1627 vp = (struct vnode *)fp->f_data;
1628 proc_fdunlock(p);
1629
1630 if (vnode_getwithref(vp)) {
1631 error = ENOENT;
1632 goto outdrop;
1633 }
1634
1635 /* Only valid for directories */
1636 if (vp->v_type != VDIR) {
1637 vnode_put(vp);
1638 error = ENOTDIR;
1639 goto outdrop;
1640 }
1641
1642 /* Get flags, mode and pathname arguments. */
1643 if (IS_64BIT_PROCESS(p)) {
1644 error = copyin(argp, &fopen, sizeof(fopen));
1645 } else {
1646 struct user32_fopenfrom fopen32;
1647
1648 error = copyin(argp, &fopen32, sizeof(fopen32));
1649 fopen.o_flags = fopen32.o_flags;
1650 fopen.o_mode = fopen32.o_mode;
1651 fopen.o_pathname = CAST_USER_ADDR_T(fopen32.o_pathname);
1652 }
1653 if (error) {
1654 vnode_put(vp);
1655 goto outdrop;
1656 }
1657 AUDIT_ARG(fflags, fopen.o_flags);
1658 AUDIT_ARG(mode, fopen.o_mode);
1659 VATTR_INIT(&va);
1660 /* Mask off all but regular access permissions */
1661 cmode = ((fopen.o_mode &~ fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT;
1662 VATTR_SET(&va, va_mode, cmode & ACCESSPERMS);
1663
1664 /* Start the lookup relative to the file descriptor's vnode. */
1665 NDINIT(&nd, LOOKUP, OP_OPEN, USEDVP | FOLLOW | AUDITVNPATH1, UIO_USERSPACE,
1666 fopen.o_pathname, &context);
1667 nd.ni_dvp = vp;
1668
1669 error = open1(&context, &nd, fopen.o_flags, &va,
1670 fileproc_alloc_init, NULL, retval);
1671
1672 vnode_put(vp);
1673 break;
1674 }
1675 /*
1676 * SPI (private) for unlinking a file starting from a dir fd
1677 */
1678 case F_UNLINKFROM: {
1679 user_addr_t pathname;
1680
1681 /* Check if this isn't a valid file descriptor */
1682 if ((fp->f_type != DTYPE_VNODE) ||
1683 (fp->f_flag & FREAD) == 0) {
1684 error = EBADF;
1685 goto out;
1686 }
1687 vp = (struct vnode *)fp->f_data;
1688 proc_fdunlock(p);
1689
1690 if (vnode_getwithref(vp)) {
1691 error = ENOENT;
1692 goto outdrop;
1693 }
1694
1695 /* Only valid for directories */
1696 if (vp->v_type != VDIR) {
1697 vnode_put(vp);
1698 error = ENOTDIR;
1699 goto outdrop;
1700 }
1701
1702 /* Get flags, mode and pathname arguments. */
1703 if (IS_64BIT_PROCESS(p)) {
1704 pathname = (user_addr_t)argp;
1705 } else {
1706 pathname = CAST_USER_ADDR_T(argp);
1707 }
1708
1709 /* Start the lookup relative to the file descriptor's vnode. */
1710 error = unlink1(&context, vp, pathname, UIO_USERSPACE, 0);
1711
1712 vnode_put(vp);
1713 break;
1714
1715 }
1716
1717 case F_ADDSIGS:
1718 case F_ADDFILESIGS:
1719 case F_ADDFILESIGS_FOR_DYLD_SIM:
1720 case F_ADDFILESIGS_RETURN:
1721 {
1722 struct cs_blob *blob = NULL;
1723 struct user_fsignatures fs;
1724 kern_return_t kr;
1725 vm_offset_t kernel_blob_addr;
1726 vm_size_t kernel_blob_size;
1727 int blob_add_flags = 0;
1728
1729 if (fp->f_type != DTYPE_VNODE) {
1730 error = EBADF;
1731 goto out;
1732 }
1733 vp = (struct vnode *)fp->f_data;
1734 proc_fdunlock(p);
1735
1736 if (uap->cmd == F_ADDFILESIGS_FOR_DYLD_SIM) {
1737 blob_add_flags |= MAC_VNODE_CHECK_DYLD_SIM;
1738 if ((p->p_csflags & CS_KILL) == 0) {
1739 proc_lock(p);
1740 p->p_csflags |= CS_KILL;
1741 proc_unlock(p);
1742 }
1743 }
1744
1745 error = vnode_getwithref(vp);
1746 if (error)
1747 goto outdrop;
1748
1749 if (IS_64BIT_PROCESS(p)) {
1750 error = copyin(argp, &fs, sizeof (fs));
1751 } else {
1752 struct user32_fsignatures fs32;
1753
1754 error = copyin(argp, &fs32, sizeof (fs32));
1755 fs.fs_file_start = fs32.fs_file_start;
1756 fs.fs_blob_start = CAST_USER_ADDR_T(fs32.fs_blob_start);
1757 fs.fs_blob_size = fs32.fs_blob_size;
1758 }
1759
1760 if (error) {
1761 vnode_put(vp);
1762 goto outdrop;
1763 }
1764
1765 /*
1766 * First check if we have something loaded a this offset
1767 */
1768 blob = ubc_cs_blob_get(vp, CPU_TYPE_ANY, fs.fs_file_start);
1769 if (blob != NULL)
1770 {
1771 /* If this is for dyld_sim revalidate the blob */
1772 if (uap->cmd == F_ADDFILESIGS_FOR_DYLD_SIM) {
1773 error = ubc_cs_blob_revalidate(vp, blob, NULL, blob_add_flags);
1774 if (error) {
1775 vnode_put(vp);
1776 goto outdrop;
1777 }
1778 }
1779
1780 } else {
1781 /*
1782 * An arbitrary limit, to prevent someone from mapping in a 20GB blob. This should cover
1783 * our use cases for the immediate future, but note that at the time of this commit, some
1784 * platforms are nearing 2MB blob sizes (with a prior soft limit of 2.5MB).
1785 *
1786 * We should consider how we can manage this more effectively; the above means that some
1787 * platforms are using megabytes of memory for signing data; it merely hasn't crossed the
1788 * threshold considered ridiculous at the time of this change.
1789 */
1790 #define CS_MAX_BLOB_SIZE (40ULL * 1024ULL * 1024ULL)
1791 if (fs.fs_blob_size > CS_MAX_BLOB_SIZE) {
1792 error = E2BIG;
1793 vnode_put(vp);
1794 goto outdrop;
1795 }
1796
1797 kernel_blob_size = CAST_DOWN(vm_size_t, fs.fs_blob_size);
1798 kr = ubc_cs_blob_allocate(&kernel_blob_addr, &kernel_blob_size);
1799 if (kr != KERN_SUCCESS) {
1800 error = ENOMEM;
1801 vnode_put(vp);
1802 goto outdrop;
1803 }
1804
1805 if(uap->cmd == F_ADDSIGS) {
1806 error = copyin(fs.fs_blob_start,
1807 (void *) kernel_blob_addr,
1808 kernel_blob_size);
1809 } else /* F_ADDFILESIGS || F_ADDFILESIGS_RETURN || F_ADDFILESIGS_FOR_DYLD_SIM */ {
1810 int resid;
1811
1812 error = vn_rdwr(UIO_READ,
1813 vp,
1814 (caddr_t) kernel_blob_addr,
1815 kernel_blob_size,
1816 fs.fs_file_start + fs.fs_blob_start,
1817 UIO_SYSSPACE,
1818 0,
1819 kauth_cred_get(),
1820 &resid,
1821 p);
1822 if ((error == 0) && resid) {
1823 /* kernel_blob_size rounded to a page size, but signature may be at end of file */
1824 memset((void *)(kernel_blob_addr + (kernel_blob_size - resid)), 0x0, resid);
1825 }
1826 }
1827
1828 if (error) {
1829 ubc_cs_blob_deallocate(kernel_blob_addr,
1830 kernel_blob_size);
1831 vnode_put(vp);
1832 goto outdrop;
1833 }
1834
1835 blob = NULL;
1836 error = ubc_cs_blob_add(vp,
1837 CPU_TYPE_ANY, /* not for a specific architecture */
1838 fs.fs_file_start,
1839 &kernel_blob_addr,
1840 kernel_blob_size,
1841 NULL,
1842 blob_add_flags,
1843 &blob);
1844
1845 /* ubc_blob_add() has consumed "kernel_blob_addr" if it is zeroed */
1846 if (error) {
1847 if (kernel_blob_addr) {
1848 ubc_cs_blob_deallocate(kernel_blob_addr,
1849 kernel_blob_size);
1850 }
1851 vnode_put(vp);
1852 goto outdrop;
1853 } else {
1854 #if CHECK_CS_VALIDATION_BITMAP
1855 ubc_cs_validation_bitmap_allocate( vp );
1856 #endif
1857 }
1858 }
1859
1860 if (uap->cmd == F_ADDFILESIGS_RETURN || uap->cmd == F_ADDFILESIGS_FOR_DYLD_SIM) {
1861 /*
1862 * The first element of the structure is a
1863 * off_t that happen to have the same size for
1864 * all archs. Lets overwrite that.
1865 */
1866 off_t end_offset = 0;
1867 if (blob)
1868 end_offset = blob->csb_end_offset;
1869 error = copyout(&end_offset, argp, sizeof (end_offset));
1870 }
1871
1872 (void) vnode_put(vp);
1873 break;
1874 }
1875 case F_GETCODEDIR:
1876 case F_FINDSIGS: {
1877 error = ENOTSUP;
1878 goto out;
1879 }
1880 case F_CHECK_LV: {
1881 struct fileglob *fg;
1882 fchecklv_t lv;
1883
1884 if (fp->f_type != DTYPE_VNODE) {
1885 error = EBADF;
1886 goto out;
1887 }
1888 fg = fp->f_fglob;
1889 proc_fdunlock(p);
1890
1891 if (IS_64BIT_PROCESS(p)) {
1892 error = copyin(argp, &lv, sizeof (lv));
1893 } else {
1894 struct user32_fchecklv lv32;
1895
1896 error = copyin(argp, &lv32, sizeof (lv32));
1897 lv.lv_file_start = lv32.lv_file_start;
1898 lv.lv_error_message = CAST_USER_ADDR_T(lv32.lv_error_message);
1899 lv.lv_error_message_size = lv32.lv_error_message;
1900 }
1901 if (error)
1902 goto outdrop;
1903
1904 #if CONFIG_MACF
1905 error = mac_file_check_library_validation(p, fg, lv.lv_file_start,
1906 lv.lv_error_message, lv.lv_error_message_size);
1907 #endif
1908
1909 break;
1910 }
1911 #if CONFIG_PROTECT
1912 case F_GETPROTECTIONCLASS: {
1913 if (fp->f_type != DTYPE_VNODE) {
1914 error = EBADF;
1915 goto out;
1916 }
1917 vp = (struct vnode *)fp->f_data;
1918
1919 proc_fdunlock(p);
1920
1921 if (vnode_getwithref(vp)) {
1922 error = ENOENT;
1923 goto outdrop;
1924 }
1925
1926 struct vnode_attr va;
1927
1928 VATTR_INIT(&va);
1929 VATTR_WANTED(&va, va_dataprotect_class);
1930 error = VNOP_GETATTR(vp, &va, &context);
1931 if (!error) {
1932 if (VATTR_IS_SUPPORTED(&va, va_dataprotect_class))
1933 *retval = va.va_dataprotect_class;
1934 else
1935 error = ENOTSUP;
1936 }
1937
1938 vnode_put(vp);
1939 break;
1940 }
1941
1942 case F_SETPROTECTIONCLASS: {
1943 /* tmp must be a valid PROTECTION_CLASS_* */
1944 tmp = CAST_DOWN_EXPLICIT(uint32_t, uap->arg);
1945
1946 if (fp->f_type != DTYPE_VNODE) {
1947 error = EBADF;
1948 goto out;
1949 }
1950 vp = (struct vnode *)fp->f_data;
1951
1952 proc_fdunlock(p);
1953
1954 if (vnode_getwithref(vp)) {
1955 error = ENOENT;
1956 goto outdrop;
1957 }
1958
1959 /* Only go forward if you have write access */
1960 vfs_context_t ctx = vfs_context_current();
1961 if(vnode_authorize(vp, NULLVP, (KAUTH_VNODE_ACCESS | KAUTH_VNODE_WRITE_DATA), ctx) != 0) {
1962 vnode_put(vp);
1963 error = EBADF;
1964 goto outdrop;
1965 }
1966
1967 struct vnode_attr va;
1968
1969 VATTR_INIT(&va);
1970 VATTR_SET(&va, va_dataprotect_class, tmp);
1971
1972 error = VNOP_SETATTR(vp, &va, ctx);
1973
1974 vnode_put(vp);
1975 break;
1976 }
1977
1978 case F_TRANSCODEKEY: {
1979 if (fp->f_type != DTYPE_VNODE) {
1980 error = EBADF;
1981 goto out;
1982 }
1983
1984 vp = (struct vnode *)fp->f_data;
1985 proc_fdunlock(p);
1986
1987 if (vnode_getwithref(vp)) {
1988 error = ENOENT;
1989 goto outdrop;
1990 }
1991
1992 cp_key_t k = {
1993 .len = CP_MAX_WRAPPEDKEYSIZE,
1994 };
1995
1996 MALLOC(k.key, char *, k.len, M_TEMP, M_WAITOK);
1997
1998 error = VNOP_IOCTL(vp, F_TRANSCODEKEY, (caddr_t)&k, 1, &context);
1999
2000 vnode_put(vp);
2001
2002 if (error == 0) {
2003 error = copyout(k.key, argp, k.len);
2004 *retval = k.len;
2005 }
2006
2007 FREE(k.key, M_TEMP);
2008
2009 break;
2010 }
2011
2012 case F_GETPROTECTIONLEVEL: {
2013 if (fp->f_type != DTYPE_VNODE) {
2014 error = EBADF;
2015 goto out;
2016 }
2017
2018 vp = (struct vnode*) fp->f_data;
2019 proc_fdunlock (p);
2020
2021 if (vnode_getwithref(vp)) {
2022 error = ENOENT;
2023 goto outdrop;
2024 }
2025
2026 error = VNOP_IOCTL(vp, F_GETPROTECTIONLEVEL, (caddr_t)retval, 0, &context);
2027
2028 vnode_put (vp);
2029 break;
2030 }
2031
2032 case F_GETDEFAULTPROTLEVEL: {
2033 if (fp->f_type != DTYPE_VNODE) {
2034 error = EBADF;
2035 goto out;
2036 }
2037
2038 vp = (struct vnode*) fp->f_data;
2039 proc_fdunlock (p);
2040
2041 if (vnode_getwithref(vp)) {
2042 error = ENOENT;
2043 goto outdrop;
2044 }
2045
2046 /*
2047 * if cp_get_major_vers fails, error will be set to proper errno
2048 * and cp_version will still be 0.
2049 */
2050
2051 error = VNOP_IOCTL(vp, F_GETDEFAULTPROTLEVEL, (caddr_t)retval, 0, &context);
2052
2053 vnode_put (vp);
2054 break;
2055 }
2056
2057 #endif /* CONFIG_PROTECT */
2058
2059 case F_MOVEDATAEXTENTS: {
2060 struct fileproc *fp2 = NULL;
2061 struct vnode *src_vp = NULLVP;
2062 struct vnode *dst_vp = NULLVP;
2063 /* We need to grab the 2nd FD out of the argments before moving on. */
2064 int fd2 = CAST_DOWN_EXPLICIT(int32_t, uap->arg);
2065
2066 error = priv_check_cred(kauth_cred_get(), PRIV_VFS_MOVE_DATA_EXTENTS, 0);
2067 if (error)
2068 goto out;
2069
2070 if (fp->f_type != DTYPE_VNODE) {
2071 error = EBADF;
2072 goto out;
2073 }
2074
2075 /* For now, special case HFS+ only, since this is SPI. */
2076 src_vp = (struct vnode *)fp->f_data;
2077 if (src_vp->v_tag != VT_HFS) {
2078 error = EINVAL;
2079 goto out;
2080 }
2081
2082 /*
2083 * Get the references before we start acquiring iocounts on the vnodes,
2084 * while we still hold the proc fd lock
2085 */
2086 if ( (error = fp_lookup(p, fd2, &fp2, 1)) ) {
2087 error = EBADF;
2088 goto out;
2089 }
2090 if (fp2->f_type != DTYPE_VNODE) {
2091 fp_drop(p, fd2, fp2, 1);
2092 error = EBADF;
2093 goto out;
2094 }
2095 dst_vp = (struct vnode *)fp2->f_data;
2096 if (dst_vp->v_tag != VT_HFS) {
2097 fp_drop(p, fd2, fp2, 1);
2098 error = EINVAL;
2099 goto out;
2100 }
2101
2102 #if CONFIG_MACF
2103 /* Re-do MAC checks against the new FD, pass in a fake argument */
2104 error = mac_file_check_fcntl(proc_ucred(p), fp2->f_fglob, uap->cmd, 0);
2105 if (error) {
2106 fp_drop(p, fd2, fp2, 1);
2107 goto out;
2108 }
2109 #endif
2110 /* Audit the 2nd FD */
2111 AUDIT_ARG(fd, fd2);
2112
2113 proc_fdunlock(p);
2114
2115 if (vnode_getwithref(src_vp)) {
2116 fp_drop(p, fd2, fp2, 0);
2117 error = ENOENT;
2118 goto outdrop;
2119 }
2120 if (vnode_getwithref(dst_vp)) {
2121 vnode_put (src_vp);
2122 fp_drop(p, fd2, fp2, 0);
2123 error = ENOENT;
2124 goto outdrop;
2125 }
2126
2127 /*
2128 * Basic asserts; validate they are not the same and that
2129 * both live on the same filesystem.
2130 */
2131 if (dst_vp == src_vp) {
2132 vnode_put (src_vp);
2133 vnode_put (dst_vp);
2134 fp_drop (p, fd2, fp2, 0);
2135 error = EINVAL;
2136 goto outdrop;
2137 }
2138
2139 if (dst_vp->v_mount != src_vp->v_mount) {
2140 vnode_put (src_vp);
2141 vnode_put (dst_vp);
2142 fp_drop (p, fd2, fp2, 0);
2143 error = EXDEV;
2144 goto outdrop;
2145 }
2146
2147 /* Now we have a legit pair of FDs. Go to work */
2148
2149 /* Now check for write access to the target files */
2150 if(vnode_authorize(src_vp, NULLVP,
2151 (KAUTH_VNODE_ACCESS | KAUTH_VNODE_WRITE_DATA), &context) != 0) {
2152 vnode_put(src_vp);
2153 vnode_put(dst_vp);
2154 fp_drop(p, fd2, fp2, 0);
2155 error = EBADF;
2156 goto outdrop;
2157 }
2158
2159 if(vnode_authorize(dst_vp, NULLVP,
2160 (KAUTH_VNODE_ACCESS | KAUTH_VNODE_WRITE_DATA), &context) != 0) {
2161 vnode_put(src_vp);
2162 vnode_put(dst_vp);
2163 fp_drop(p, fd2, fp2, 0);
2164 error = EBADF;
2165 goto outdrop;
2166 }
2167
2168 /* Verify that both vps point to files and not directories */
2169 if ( !vnode_isreg(src_vp) || !vnode_isreg(dst_vp)) {
2170 error = EINVAL;
2171 vnode_put (src_vp);
2172 vnode_put (dst_vp);
2173 fp_drop (p, fd2, fp2, 0);
2174 goto outdrop;
2175 }
2176
2177 /*
2178 * The exchangedata syscall handler passes in 0 for the flags to VNOP_EXCHANGE.
2179 * We'll pass in our special bit indicating that the new behavior is expected
2180 */
2181
2182 error = VNOP_EXCHANGE(src_vp, dst_vp, FSOPT_EXCHANGE_DATA_ONLY, &context);
2183
2184 vnode_put (src_vp);
2185 vnode_put (dst_vp);
2186 fp_drop(p, fd2, fp2, 0);
2187 break;
2188 }
2189
2190 /*
2191 * SPI for making a file compressed.
2192 */
2193 case F_MAKECOMPRESSED: {
2194 uint32_t gcounter = CAST_DOWN_EXPLICIT(uint32_t, uap->arg);
2195
2196 if (fp->f_type != DTYPE_VNODE) {
2197 error = EBADF;
2198 goto out;
2199 }
2200
2201 vp = (struct vnode*) fp->f_data;
2202 proc_fdunlock (p);
2203
2204 /* get the vnode */
2205 if (vnode_getwithref(vp)) {
2206 error = ENOENT;
2207 goto outdrop;
2208 }
2209
2210 /* Is it a file? */
2211 if ((vnode_isreg(vp) == 0) && (vnode_islnk(vp) == 0)) {
2212 vnode_put(vp);
2213 error = EBADF;
2214 goto outdrop;
2215 }
2216
2217 /* invoke ioctl to pass off to FS */
2218 /* Only go forward if you have write access */
2219 vfs_context_t ctx = vfs_context_current();
2220 if(vnode_authorize(vp, NULLVP, (KAUTH_VNODE_ACCESS | KAUTH_VNODE_WRITE_DATA), ctx) != 0) {
2221 vnode_put(vp);
2222 error = EBADF;
2223 goto outdrop;
2224 }
2225
2226 error = VNOP_IOCTL(vp, uap->cmd, (caddr_t)&gcounter, 0, &context);
2227
2228 vnode_put (vp);
2229 break;
2230 }
2231
2232 /*
2233 * SPI (private) for indicating to a filesystem that subsequent writes to
2234 * the open FD will written to the Fastflow.
2235 */
2236 case F_SET_GREEDY_MODE:
2237 /* intentionally drop through to the same handler as F_SETSTATIC.
2238 * both fcntls should pass the argument and their selector into VNOP_IOCTL.
2239 */
2240
2241 /*
2242 * SPI (private) for indicating to a filesystem that subsequent writes to
2243 * the open FD will represent static content.
2244 */
2245 case F_SETSTATICCONTENT: {
2246 caddr_t ioctl_arg = NULL;
2247
2248 if (uap->arg) {
2249 ioctl_arg = (caddr_t) 1;
2250 }
2251
2252 if (fp->f_type != DTYPE_VNODE) {
2253 error = EBADF;
2254 goto out;
2255 }
2256 vp = (struct vnode *)fp->f_data;
2257 proc_fdunlock(p);
2258
2259 error = vnode_getwithref(vp);
2260 if (error) {
2261 error = ENOENT;
2262 goto outdrop;
2263 }
2264
2265 /* Only go forward if you have write access */
2266 vfs_context_t ctx = vfs_context_current();
2267 if(vnode_authorize(vp, NULLVP, (KAUTH_VNODE_ACCESS | KAUTH_VNODE_WRITE_DATA), ctx) != 0) {
2268 vnode_put(vp);
2269 error = EBADF;
2270 goto outdrop;
2271 }
2272
2273 error = VNOP_IOCTL(vp, uap->cmd, ioctl_arg, 0, &context);
2274 (void)vnode_put(vp);
2275
2276 break;
2277 }
2278
2279 /*
2280 * SPI (private) for indicating to the lower level storage driver that the
2281 * subsequent writes should be of a particular IO type (burst, greedy, static),
2282 * or other flavors that may be necessary.
2283 */
2284 case F_SETIOTYPE: {
2285 caddr_t param_ptr;
2286 uint32_t param;
2287
2288 if (uap->arg) {
2289 /* extract 32 bits of flags from userland */
2290 param_ptr = (caddr_t) uap->arg;
2291 param = (uint32_t) param_ptr;
2292 }
2293 else {
2294 /* If no argument is specified, error out */
2295 error = EINVAL;
2296 goto out;
2297 }
2298
2299 /*
2300 * Validate the different types of flags that can be specified:
2301 * all of them are mutually exclusive for now.
2302 */
2303 switch (param) {
2304 case F_IOTYPE_ISOCHRONOUS:
2305 break;
2306
2307 default:
2308 error = EINVAL;
2309 goto out;
2310 }
2311
2312
2313 if (fp->f_type != DTYPE_VNODE) {
2314 error = EBADF;
2315 goto out;
2316 }
2317 vp = (struct vnode *)fp->f_data;
2318 proc_fdunlock(p);
2319
2320 error = vnode_getwithref(vp);
2321 if (error) {
2322 error = ENOENT;
2323 goto outdrop;
2324 }
2325
2326 /* Only go forward if you have write access */
2327 vfs_context_t ctx = vfs_context_current();
2328 if(vnode_authorize(vp, NULLVP, (KAUTH_VNODE_ACCESS | KAUTH_VNODE_WRITE_DATA), ctx) != 0) {
2329 vnode_put(vp);
2330 error = EBADF;
2331 goto outdrop;
2332 }
2333
2334 error = VNOP_IOCTL(vp, uap->cmd, param_ptr, 0, &context);
2335 (void)vnode_put(vp);
2336
2337 break;
2338 }
2339
2340 /*
2341 * Set the vnode pointed to by 'fd'
2342 * and tag it as the (potentially future) backing store
2343 * for another filesystem
2344 */
2345 case F_SETBACKINGSTORE: {
2346 if (fp->f_type != DTYPE_VNODE) {
2347 error = EBADF;
2348 goto out;
2349 }
2350
2351 vp = (struct vnode *)fp->f_data;
2352
2353 if (vp->v_tag != VT_HFS) {
2354 error = EINVAL;
2355 goto out;
2356 }
2357 proc_fdunlock(p);
2358
2359 if (vnode_getwithref(vp)) {
2360 error = ENOENT;
2361 goto outdrop;
2362 }
2363
2364 /* only proceed if you have write access */
2365 vfs_context_t ctx = vfs_context_current();
2366 if(vnode_authorize(vp, NULLVP, (KAUTH_VNODE_ACCESS | KAUTH_VNODE_WRITE_DATA), ctx) != 0) {
2367 vnode_put(vp);
2368 error = EBADF;
2369 goto outdrop;
2370 }
2371
2372
2373 /* If arg != 0, set, otherwise unset */
2374 if (uap->arg) {
2375 error = VNOP_IOCTL (vp, uap->cmd, (caddr_t)1, 0, &context);
2376 }
2377 else {
2378 error = VNOP_IOCTL (vp, uap->cmd, (caddr_t)NULL, 0, &context);
2379 }
2380
2381 vnode_put(vp);
2382 break;
2383 }
2384
2385 /*
2386 * like F_GETPATH, but special semantics for
2387 * the mobile time machine handler.
2388 */
2389 case F_GETPATH_MTMINFO: {
2390 char *pathbufp;
2391 int pathlen;
2392
2393 if (fp->f_type != DTYPE_VNODE) {
2394 error = EBADF;
2395 goto out;
2396 }
2397 vp = (struct vnode *)fp->f_data;
2398 proc_fdunlock(p);
2399
2400 pathlen = MAXPATHLEN;
2401 MALLOC(pathbufp, char *, pathlen, M_TEMP, M_WAITOK);
2402 if (pathbufp == NULL) {
2403 error = ENOMEM;
2404 goto outdrop;
2405 }
2406 if ( (error = vnode_getwithref(vp)) == 0 ) {
2407 int backingstore = 0;
2408
2409 /* Check for error from vn_getpath before moving on */
2410 if ((error = vn_getpath(vp, pathbufp, &pathlen)) == 0) {
2411 if (vp->v_tag == VT_HFS) {
2412 error = VNOP_IOCTL (vp, uap->cmd, (caddr_t) &backingstore, 0, &context);
2413 }
2414 (void)vnode_put(vp);
2415
2416 if (error == 0) {
2417 error = copyout((caddr_t)pathbufp, argp, pathlen);
2418 }
2419 if (error == 0) {
2420 /*
2421 * If the copyout was successful, now check to ensure
2422 * that this vnode is not a BACKINGSTORE vnode. mtmd
2423 * wants the path regardless.
2424 */
2425 if (backingstore) {
2426 error = EBUSY;
2427 }
2428 }
2429 } else
2430 (void)vnode_put(vp);
2431 }
2432 FREE(pathbufp, M_TEMP);
2433 goto outdrop;
2434 }
2435
2436 #if DEBUG || DEVELOPMENT
2437 case F_RECYCLE:
2438 if (fp->f_type != DTYPE_VNODE) {
2439 error = EBADF;
2440 goto out;
2441 }
2442 vp = (struct vnode *)fp->f_data;
2443 proc_fdunlock(p);
2444
2445 vnode_recycle(vp);
2446 break;
2447 #endif
2448
2449 default:
2450 /*
2451 * This is an fcntl() that we d not recognize at this level;
2452 * if this is a vnode, we send it down into the VNOP_IOCTL
2453 * for this vnode; this can include special devices, and will
2454 * effectively overload fcntl() to send ioctl()'s.
2455 */
2456 if((uap->cmd & IOC_VOID) && (uap->cmd & IOC_INOUT)){
2457 error = EINVAL;
2458 goto out;
2459 }
2460
2461 /* Catch any now-invalid fcntl() selectors */
2462 switch (uap->cmd) {
2463 case F_MARKDEPENDENCY:
2464 error = EINVAL;
2465 goto out;
2466 default:
2467 break;
2468 }
2469
2470 if (fp->f_type != DTYPE_VNODE) {
2471 error = EBADF;
2472 goto out;
2473 }
2474 vp = (struct vnode *)fp->f_data;
2475 proc_fdunlock(p);
2476
2477 if ( (error = vnode_getwithref(vp)) == 0 ) {
2478 #define STK_PARAMS 128
2479 char stkbuf[STK_PARAMS];
2480 unsigned int size;
2481 caddr_t data, memp;
2482 /*
2483 * For this to work properly, we have to copy in the
2484 * ioctl() cmd argument if there is one; we must also
2485 * check that a command parameter, if present, does
2486 * not exceed the maximum command length dictated by
2487 * the number of bits we have available in the command
2488 * to represent a structure length. Finally, we have
2489 * to copy the results back out, if it is that type of
2490 * ioctl().
2491 */
2492 size = IOCPARM_LEN(uap->cmd);
2493 if (size > IOCPARM_MAX) {
2494 (void)vnode_put(vp);
2495 error = EINVAL;
2496 break;
2497 }
2498
2499 memp = NULL;
2500 if (size > sizeof (stkbuf)) {
2501 if ((memp = (caddr_t)kalloc(size)) == 0) {
2502 (void)vnode_put(vp);
2503 error = ENOMEM;
2504 goto outdrop;
2505 }
2506 data = memp;
2507 } else {
2508 data = &stkbuf[0];
2509 }
2510
2511 if (uap->cmd & IOC_IN) {
2512 if (size) {
2513 /* structure */
2514 error = copyin(argp, data, size);
2515 if (error) {
2516 (void)vnode_put(vp);
2517 if (memp)
2518 kfree(memp, size);
2519 goto outdrop;
2520 }
2521
2522 /* Bzero the section beyond that which was needed */
2523 if (size <= sizeof(stkbuf)) {
2524 bzero ( (((uint8_t*)data) + size), (sizeof(stkbuf) - size));
2525 }
2526 } else {
2527 /* int */
2528 if (is64bit) {
2529 *(user_addr_t *)data = argp;
2530 } else {
2531 *(uint32_t *)data = (uint32_t)argp;
2532 }
2533 };
2534 } else if ((uap->cmd & IOC_OUT) && size) {
2535 /*
2536 * Zero the buffer so the user always
2537 * gets back something deterministic.
2538 */
2539 bzero(data, size);
2540 } else if (uap->cmd & IOC_VOID) {
2541 if (is64bit) {
2542 *(user_addr_t *)data = argp;
2543 } else {
2544 *(uint32_t *)data = (uint32_t)argp;
2545 }
2546 }
2547
2548 error = VNOP_IOCTL(vp, uap->cmd, CAST_DOWN(caddr_t, data), 0, &context);
2549
2550 (void)vnode_put(vp);
2551
2552 /* Copy any output data to user */
2553 if (error == 0 && (uap->cmd & IOC_OUT) && size)
2554 error = copyout(data, argp, size);
2555 if (memp)
2556 kfree(memp, size);
2557 }
2558 break;
2559 }
2560
2561 outdrop:
2562 AUDIT_ARG(vnpath_withref, vp, ARG_VNODE1);
2563 fp_drop(p, fd, fp, 0);
2564 return(error);
2565 out:
2566 fp_drop(p, fd, fp, 1);
2567 proc_fdunlock(p);
2568 return(error);
2569 }
2570
2571
2572 /*
2573 * finishdup
2574 *
2575 * Description: Common code for dup, dup2, and fcntl(F_DUPFD).
2576 *
2577 * Parameters: p Process performing the dup
2578 * old The fd to dup
2579 * new The fd to dup it to
2580 * fd_flags Flags to augment the new fd
2581 * retval Pointer to the call return area
2582 *
2583 * Returns: 0 Success
2584 * EBADF
2585 * ENOMEM
2586 *
2587 * Implicit returns:
2588 * *retval (modified) The new descriptor
2589 *
2590 * Locks: Assumes proc_fdlock for process pointing to fdp is held by
2591 * the caller
2592 *
2593 * Notes: This function may drop and reacquire this lock; it is unsafe
2594 * for a caller to assume that other state protected by the lock
2595 * has not been subsequently changed out from under it.
2596 */
2597 int
2598 finishdup(proc_t p,
2599 struct filedesc *fdp, int old, int new, int fd_flags, int32_t *retval)
2600 {
2601 struct fileproc *nfp;
2602 struct fileproc *ofp;
2603 #if CONFIG_MACF
2604 int error;
2605 #endif
2606
2607 #if DIAGNOSTIC
2608 proc_fdlock_assert(p, LCK_MTX_ASSERT_OWNED);
2609 #endif
2610 if ((ofp = fdp->fd_ofiles[old]) == NULL ||
2611 (fdp->fd_ofileflags[old] & UF_RESERVED)) {
2612 fdrelse(p, new);
2613 return (EBADF);
2614 }
2615 fg_ref(ofp);
2616
2617 #if CONFIG_MACF
2618 error = mac_file_check_dup(proc_ucred(p), ofp->f_fglob, new);
2619 if (error) {
2620 fg_drop(ofp);
2621 fdrelse(p, new);
2622 return (error);
2623 }
2624 #endif
2625
2626 proc_fdunlock(p);
2627
2628 nfp = fileproc_alloc_init(NULL);
2629
2630 proc_fdlock(p);
2631
2632 if (nfp == NULL) {
2633 fg_drop(ofp);
2634 fdrelse(p, new);
2635 return (ENOMEM);
2636 }
2637
2638 nfp->f_fglob = ofp->f_fglob;
2639
2640 #if DIAGNOSTIC
2641 if (fdp->fd_ofiles[new] != 0)
2642 panic("finishdup: overwriting fd_ofiles with new %d", new);
2643 if ((fdp->fd_ofileflags[new] & UF_RESERVED) == 0)
2644 panic("finishdup: unreserved fileflags with new %d", new);
2645 #endif
2646
2647 if (new > fdp->fd_lastfile)
2648 fdp->fd_lastfile = new;
2649 *fdflags(p, new) |= fd_flags;
2650 procfdtbl_releasefd(p, new, nfp);
2651 *retval = new;
2652 return (0);
2653 }
2654
2655
2656 /*
2657 * close
2658 *
2659 * Description: The implementation of the close(2) system call
2660 *
2661 * Parameters: p Process in whose per process file table
2662 * the close is to occur
2663 * uap->fd fd to be closed
2664 * retval <unused>
2665 *
2666 * Returns: 0 Success
2667 * fp_lookup:EBADF Bad file descriptor
2668 * fp_guard_exception:??? Guarded file descriptor
2669 * close_internal:EBADF
2670 * close_internal:??? Anything returnable by a per-fileops
2671 * close function
2672 */
2673 int
2674 close(proc_t p, struct close_args *uap, int32_t *retval)
2675 {
2676 __pthread_testcancel(1);
2677 return(close_nocancel(p, (struct close_nocancel_args *)uap, retval));
2678 }
2679
2680
2681 int
2682 close_nocancel(proc_t p, struct close_nocancel_args *uap, __unused int32_t *retval)
2683 {
2684 struct fileproc *fp;
2685 int fd = uap->fd;
2686 int error;
2687
2688 AUDIT_SYSCLOSE(p, fd);
2689
2690 proc_fdlock(p);
2691
2692 if ( (error = fp_lookup(p,fd,&fp, 1)) ) {
2693 proc_fdunlock(p);
2694 return(error);
2695 }
2696
2697 if (FP_ISGUARDED(fp, GUARD_CLOSE)) {
2698 error = fp_guard_exception(p, fd, fp, kGUARD_EXC_CLOSE);
2699 (void) fp_drop(p, fd, fp, 1);
2700 proc_fdunlock(p);
2701 return (error);
2702 }
2703
2704 error = close_internal_locked(p, fd, fp, 0);
2705
2706 proc_fdunlock(p);
2707
2708 return (error);
2709 }
2710
2711
2712 /*
2713 * close_internal_locked
2714 *
2715 * Close a file descriptor.
2716 *
2717 * Parameters: p Process in whose per process file table
2718 * the close is to occur
2719 * fd fd to be closed
2720 * fp fileproc associated with the fd
2721 *
2722 * Returns: 0 Success
2723 * EBADF fd already in close wait state
2724 * closef_locked:??? Anything returnable by a per-fileops
2725 * close function
2726 *
2727 * Locks: Assumes proc_fdlock for process is held by the caller and returns
2728 * with lock held
2729 *
2730 * Notes: This function may drop and reacquire this lock; it is unsafe
2731 * for a caller to assume that other state protected by the lock
2732 * has not been subsequently changed out from under it.
2733 */
2734 int
2735 close_internal_locked(proc_t p, int fd, struct fileproc *fp, int flags)
2736 {
2737 struct filedesc *fdp = p->p_fd;
2738 int error =0;
2739 int resvfd = flags & FD_DUP2RESV;
2740
2741
2742 #if DIAGNOSTIC
2743 proc_fdlock_assert(p, LCK_MTX_ASSERT_OWNED);
2744 #endif
2745
2746 /* Keep people from using the filedesc while we are closing it */
2747 procfdtbl_markclosefd(p, fd);
2748
2749
2750 if ((fp->f_flags & FP_CLOSING) == FP_CLOSING) {
2751 panic("close_internal_locked: being called on already closing fd");
2752 }
2753
2754
2755 #if DIAGNOSTIC
2756 if ((fdp->fd_ofileflags[fd] & UF_RESERVED) == 0)
2757 panic("close_internal: unreserved fileflags with fd %d", fd);
2758 #endif
2759
2760 fp->f_flags |= FP_CLOSING;
2761
2762 if ( (fp->f_flags & FP_AIOISSUED) || kauth_authorize_fileop_has_listeners() ) {
2763
2764 proc_fdunlock(p);
2765
2766 if ( (fp->f_type == DTYPE_VNODE) && kauth_authorize_fileop_has_listeners() ) {
2767 /*
2768 * call out to allow 3rd party notification of close.
2769 * Ignore result of kauth_authorize_fileop call.
2770 */
2771 if (vnode_getwithref((vnode_t)fp->f_data) == 0) {
2772 u_int fileop_flags = 0;
2773 if ((fp->f_flags & FP_WRITTEN) != 0)
2774 fileop_flags |= KAUTH_FILEOP_CLOSE_MODIFIED;
2775 kauth_authorize_fileop(fp->f_fglob->fg_cred, KAUTH_FILEOP_CLOSE,
2776 (uintptr_t)fp->f_data, (uintptr_t)fileop_flags);
2777 vnode_put((vnode_t)fp->f_data);
2778 }
2779 }
2780 if (fp->f_flags & FP_AIOISSUED)
2781 /*
2782 * cancel all async IO requests that can be cancelled.
2783 */
2784 _aio_close( p, fd );
2785
2786 proc_fdlock(p);
2787 }
2788
2789 if (fd < fdp->fd_knlistsize)
2790 knote_fdclose(p, fd, FALSE);
2791
2792 if (fp->f_flags & FP_WAITEVENT)
2793 (void)waitevent_close(p, fp);
2794
2795 fileproc_drain(p, fp);
2796
2797 if (resvfd == 0) {
2798 _fdrelse(p, fd);
2799 } else {
2800 procfdtbl_reservefd(p, fd);
2801 }
2802
2803 if (ENTR_SHOULDTRACE && fp->f_type == DTYPE_SOCKET)
2804 KERNEL_ENERGYTRACE(kEnTrActKernSocket, DBG_FUNC_END,
2805 fd, 0, (int64_t)VM_KERNEL_ADDRPERM(fp->f_data));
2806
2807 error = closef_locked(fp, fp->f_fglob, p);
2808 if ((fp->f_flags & FP_WAITCLOSE) == FP_WAITCLOSE)
2809 wakeup(&fp->f_flags);
2810 fp->f_flags &= ~(FP_WAITCLOSE | FP_CLOSING);
2811
2812 proc_fdunlock(p);
2813
2814 fileproc_free(fp);
2815
2816 proc_fdlock(p);
2817
2818 #if DIAGNOSTIC
2819 if (resvfd != 0) {
2820 if ((fdp->fd_ofileflags[fd] & UF_RESERVED) == 0)
2821 panic("close with reserved fd returns with freed fd:%d: proc: %p", fd, p);
2822 }
2823 #endif
2824
2825 return(error);
2826 }
2827
2828
2829 /*
2830 * fstat1
2831 *
2832 * Description: Return status information about a file descriptor.
2833 *
2834 * Parameters: p The process doing the fstat
2835 * fd The fd to stat
2836 * ub The user stat buffer
2837 * xsecurity The user extended security
2838 * buffer, or 0 if none
2839 * xsecurity_size The size of xsecurity, or 0
2840 * if no xsecurity
2841 * isstat64 Flag to indicate 64 bit version
2842 * for inode size, etc.
2843 *
2844 * Returns: 0 Success
2845 * EBADF
2846 * EFAULT
2847 * fp_lookup:EBADF Bad file descriptor
2848 * vnode_getwithref:???
2849 * copyout:EFAULT
2850 * vnode_getwithref:???
2851 * vn_stat:???
2852 * soo_stat:???
2853 * pipe_stat:???
2854 * pshm_stat:???
2855 * kqueue_stat:???
2856 *
2857 * Notes: Internal implementation for all other fstat() related
2858 * functions
2859 *
2860 * XXX switch on node type is bogus; need a stat in struct
2861 * XXX fileops instead.
2862 */
2863 static int
2864 fstat1(proc_t p, int fd, user_addr_t ub, user_addr_t xsecurity, user_addr_t xsecurity_size, int isstat64)
2865 {
2866 struct fileproc *fp;
2867 union {
2868 struct stat sb;
2869 struct stat64 sb64;
2870 } source;
2871 union {
2872 struct user64_stat user64_sb;
2873 struct user32_stat user32_sb;
2874 struct user64_stat64 user64_sb64;
2875 struct user32_stat64 user32_sb64;
2876 } dest;
2877 int error, my_size;
2878 file_type_t type;
2879 caddr_t data;
2880 kauth_filesec_t fsec;
2881 user_size_t xsecurity_bufsize;
2882 vfs_context_t ctx = vfs_context_current();
2883 void * sbptr;
2884
2885
2886 AUDIT_ARG(fd, fd);
2887
2888 if ((error = fp_lookup(p, fd, &fp, 0)) != 0) {
2889 return(error);
2890 }
2891 type = fp->f_type;
2892 data = fp->f_data;
2893 fsec = KAUTH_FILESEC_NONE;
2894
2895 sbptr = (void *)&source;
2896
2897 switch (type) {
2898
2899 case DTYPE_VNODE:
2900 if ((error = vnode_getwithref((vnode_t)data)) == 0) {
2901 /*
2902 * If the caller has the file open, and is not
2903 * requesting extended security information, we are
2904 * going to let them get the basic stat information.
2905 */
2906 if (xsecurity == USER_ADDR_NULL) {
2907 error = vn_stat_noauth((vnode_t)data, sbptr, NULL, isstat64, ctx,
2908 fp->f_fglob->fg_cred);
2909 } else {
2910 error = vn_stat((vnode_t)data, sbptr, &fsec, isstat64, ctx);
2911 }
2912
2913 AUDIT_ARG(vnpath, (struct vnode *)data, ARG_VNODE1);
2914 (void)vnode_put((vnode_t)data);
2915 }
2916 break;
2917
2918 #if SOCKETS
2919 case DTYPE_SOCKET:
2920 error = soo_stat((struct socket *)data, sbptr, isstat64);
2921 break;
2922 #endif /* SOCKETS */
2923
2924 case DTYPE_PIPE:
2925 error = pipe_stat((void *)data, sbptr, isstat64);
2926 break;
2927
2928 case DTYPE_PSXSHM:
2929 error = pshm_stat((void *)data, sbptr, isstat64);
2930 break;
2931
2932 case DTYPE_KQUEUE:
2933 error = kqueue_stat((void *)data, sbptr, isstat64, p);
2934 break;
2935
2936 default:
2937 error = EBADF;
2938 goto out;
2939 }
2940 if (error == 0) {
2941 caddr_t sbp;
2942
2943 if (isstat64 != 0) {
2944 source.sb64.st_lspare = 0;
2945 source.sb64.st_qspare[0] = 0LL;
2946 source.sb64.st_qspare[1] = 0LL;
2947
2948 if (IS_64BIT_PROCESS(current_proc())) {
2949 munge_user64_stat64(&source.sb64, &dest.user64_sb64);
2950 my_size = sizeof(dest.user64_sb64);
2951 sbp = (caddr_t)&dest.user64_sb64;
2952 } else {
2953 munge_user32_stat64(&source.sb64, &dest.user32_sb64);
2954 my_size = sizeof(dest.user32_sb64);
2955 sbp = (caddr_t)&dest.user32_sb64;
2956 }
2957 } else {
2958 source.sb.st_lspare = 0;
2959 source.sb.st_qspare[0] = 0LL;
2960 source.sb.st_qspare[1] = 0LL;
2961 if (IS_64BIT_PROCESS(current_proc())) {
2962 munge_user64_stat(&source.sb, &dest.user64_sb);
2963 my_size = sizeof(dest.user64_sb);
2964 sbp = (caddr_t)&dest.user64_sb;
2965 } else {
2966 munge_user32_stat(&source.sb, &dest.user32_sb);
2967 my_size = sizeof(dest.user32_sb);
2968 sbp = (caddr_t)&dest.user32_sb;
2969 }
2970 }
2971
2972 error = copyout(sbp, ub, my_size);
2973 }
2974
2975 /* caller wants extended security information? */
2976 if (xsecurity != USER_ADDR_NULL) {
2977
2978 /* did we get any? */
2979 if (fsec == KAUTH_FILESEC_NONE) {
2980 if (susize(xsecurity_size, 0) != 0) {
2981 error = EFAULT;
2982 goto out;
2983 }
2984 } else {
2985 /* find the user buffer size */
2986 xsecurity_bufsize = fusize(xsecurity_size);
2987
2988 /* copy out the actual data size */
2989 if (susize(xsecurity_size, KAUTH_FILESEC_COPYSIZE(fsec)) != 0) {
2990 error = EFAULT;
2991 goto out;
2992 }
2993
2994 /* if the caller supplied enough room, copy out to it */
2995 if (xsecurity_bufsize >= KAUTH_FILESEC_COPYSIZE(fsec))
2996 error = copyout(fsec, xsecurity, KAUTH_FILESEC_COPYSIZE(fsec));
2997 }
2998 }
2999 out:
3000 fp_drop(p, fd, fp, 0);
3001 if (fsec != NULL)
3002 kauth_filesec_free(fsec);
3003 return (error);
3004 }
3005
3006
3007 /*
3008 * fstat_extended
3009 *
3010 * Description: Extended version of fstat supporting returning extended
3011 * security information
3012 *
3013 * Parameters: p The process doing the fstat
3014 * uap->fd The fd to stat
3015 * uap->ub The user stat buffer
3016 * uap->xsecurity The user extended security
3017 * buffer, or 0 if none
3018 * uap->xsecurity_size The size of xsecurity, or 0
3019 *
3020 * Returns: 0 Success
3021 * !0 Errno (see fstat1)
3022 */
3023 int
3024 fstat_extended(proc_t p, struct fstat_extended_args *uap, __unused int32_t *retval)
3025 {
3026 return(fstat1(p, uap->fd, uap->ub, uap->xsecurity, uap->xsecurity_size, 0));
3027 }
3028
3029
3030 /*
3031 * fstat
3032 *
3033 * Description: Get file status for the file associated with fd
3034 *
3035 * Parameters: p The process doing the fstat
3036 * uap->fd The fd to stat
3037 * uap->ub The user stat buffer
3038 *
3039 * Returns: 0 Success
3040 * !0 Errno (see fstat1)
3041 */
3042 int
3043 fstat(proc_t p, struct fstat_args *uap, __unused int32_t *retval)
3044 {
3045 return(fstat1(p, uap->fd, uap->ub, 0, 0, 0));
3046 }
3047
3048
3049 /*
3050 * fstat64_extended
3051 *
3052 * Description: Extended version of fstat64 supporting returning extended
3053 * security information
3054 *
3055 * Parameters: p The process doing the fstat
3056 * uap->fd The fd to stat
3057 * uap->ub The user stat buffer
3058 * uap->xsecurity The user extended security
3059 * buffer, or 0 if none
3060 * uap->xsecurity_size The size of xsecurity, or 0
3061 *
3062 * Returns: 0 Success
3063 * !0 Errno (see fstat1)
3064 */
3065 int
3066 fstat64_extended(proc_t p, struct fstat64_extended_args *uap, __unused int32_t *retval)
3067 {
3068 return(fstat1(p, uap->fd, uap->ub, uap->xsecurity, uap->xsecurity_size, 1));
3069 }
3070
3071
3072 /*
3073 * fstat64
3074 *
3075 * Description: Get 64 bit version of the file status for the file associated
3076 * with fd
3077 *
3078 * Parameters: p The process doing the fstat
3079 * uap->fd The fd to stat
3080 * uap->ub The user stat buffer
3081 *
3082 * Returns: 0 Success
3083 * !0 Errno (see fstat1)
3084 */
3085 int
3086 fstat64(proc_t p, struct fstat64_args *uap, __unused int32_t *retval)
3087 {
3088 return(fstat1(p, uap->fd, uap->ub, 0, 0, 1));
3089 }
3090
3091
3092 /*
3093 * fpathconf
3094 *
3095 * Description: Return pathconf information about a file descriptor.
3096 *
3097 * Parameters: p Process making the request
3098 * uap->fd fd to get information about
3099 * uap->name Name of information desired
3100 * retval Pointer to the call return area
3101 *
3102 * Returns: 0 Success
3103 * EINVAL
3104 * fp_lookup:EBADF Bad file descriptor
3105 * vnode_getwithref:???
3106 * vn_pathconf:???
3107 *
3108 * Implicit returns:
3109 * *retval (modified) Returned information (numeric)
3110 */
3111 int
3112 fpathconf(proc_t p, struct fpathconf_args *uap, int32_t *retval)
3113 {
3114 int fd = uap->fd;
3115 struct fileproc *fp;
3116 struct vnode *vp;
3117 int error = 0;
3118 file_type_t type;
3119 caddr_t data;
3120
3121
3122 AUDIT_ARG(fd, uap->fd);
3123 if ( (error = fp_lookup(p, fd, &fp, 0)) )
3124 return(error);
3125 type = fp->f_type;
3126 data = fp->f_data;
3127
3128 switch (type) {
3129
3130 case DTYPE_SOCKET:
3131 if (uap->name != _PC_PIPE_BUF) {
3132 error = EINVAL;
3133 goto out;
3134 }
3135 *retval = PIPE_BUF;
3136 error = 0;
3137 goto out;
3138
3139 case DTYPE_PIPE:
3140 if (uap->name != _PC_PIPE_BUF) {
3141 error = EINVAL;
3142 goto out;
3143 }
3144 *retval = PIPE_BUF;
3145 error = 0;
3146 goto out;
3147
3148 case DTYPE_VNODE:
3149 vp = (struct vnode *)data;
3150
3151 if ( (error = vnode_getwithref(vp)) == 0) {
3152 AUDIT_ARG(vnpath, vp, ARG_VNODE1);
3153
3154 error = vn_pathconf(vp, uap->name, retval, vfs_context_current());
3155
3156 (void)vnode_put(vp);
3157 }
3158 goto out;
3159
3160 default:
3161 error = EINVAL;
3162 goto out;
3163
3164 }
3165 /*NOTREACHED*/
3166 out:
3167 fp_drop(p, fd, fp, 0);
3168 return(error);
3169 }
3170
3171 /*
3172 * Statistics counter for the number of times a process calling fdalloc()
3173 * has resulted in an expansion of the per process open file table.
3174 *
3175 * XXX This would likely be of more use if it were per process
3176 */
3177 int fdexpand;
3178
3179
3180 /*
3181 * fdalloc
3182 *
3183 * Description: Allocate a file descriptor for the process.
3184 *
3185 * Parameters: p Process to allocate the fd in
3186 * want The fd we would prefer to get
3187 * result Pointer to fd we got
3188 *
3189 * Returns: 0 Success
3190 * EMFILE
3191 * ENOMEM
3192 *
3193 * Implicit returns:
3194 * *result (modified) The fd which was allocated
3195 */
3196 int
3197 fdalloc(proc_t p, int want, int *result)
3198 {
3199 struct filedesc *fdp = p->p_fd;
3200 int i;
3201 int lim, last, numfiles, oldnfiles;
3202 struct fileproc **newofiles, **ofiles;
3203 char *newofileflags;
3204
3205 /*
3206 * Search for a free descriptor starting at the higher
3207 * of want or fd_freefile. If that fails, consider
3208 * expanding the ofile array.
3209 */
3210 #if DIAGNOSTIC
3211 proc_fdlock_assert(p, LCK_MTX_ASSERT_OWNED);
3212 #endif
3213
3214 lim = min((int)p->p_rlimit[RLIMIT_NOFILE].rlim_cur, maxfiles);
3215 for (;;) {
3216 last = min(fdp->fd_nfiles, lim);
3217 if ((i = want) < fdp->fd_freefile)
3218 i = fdp->fd_freefile;
3219 for (; i < last; i++) {
3220 if (fdp->fd_ofiles[i] == NULL && !(fdp->fd_ofileflags[i] & UF_RESERVED)) {
3221 procfdtbl_reservefd(p, i);
3222 if (i > fdp->fd_lastfile)
3223 fdp->fd_lastfile = i;
3224 if (want <= fdp->fd_freefile)
3225 fdp->fd_freefile = i;
3226 *result = i;
3227 return (0);
3228 }
3229 }
3230
3231 /*
3232 * No space in current array. Expand?
3233 */
3234 if (fdp->fd_nfiles >= lim)
3235 return (EMFILE);
3236 if (fdp->fd_nfiles < NDEXTENT)
3237 numfiles = NDEXTENT;
3238 else
3239 numfiles = 2 * fdp->fd_nfiles;
3240 /* Enforce lim */
3241 if (numfiles > lim)
3242 numfiles = lim;
3243 proc_fdunlock(p);
3244 MALLOC_ZONE(newofiles, struct fileproc **,
3245 numfiles * OFILESIZE, M_OFILETABL, M_WAITOK);
3246 proc_fdlock(p);
3247 if (newofiles == NULL) {
3248 return (ENOMEM);
3249 }
3250 if (fdp->fd_nfiles >= numfiles) {
3251 FREE_ZONE(newofiles, numfiles * OFILESIZE, M_OFILETABL);
3252 continue;
3253 }
3254 newofileflags = (char *) &newofiles[numfiles];
3255 /*
3256 * Copy the existing ofile and ofileflags arrays
3257 * and zero the new portion of each array.
3258 */
3259 oldnfiles = fdp->fd_nfiles;
3260 (void) memcpy(newofiles, fdp->fd_ofiles,
3261 oldnfiles * sizeof(*fdp->fd_ofiles));
3262 (void) memset(&newofiles[oldnfiles], 0,
3263 (numfiles - oldnfiles) * sizeof(*fdp->fd_ofiles));
3264
3265 (void) memcpy(newofileflags, fdp->fd_ofileflags,
3266 oldnfiles * sizeof(*fdp->fd_ofileflags));
3267 (void) memset(&newofileflags[oldnfiles], 0,
3268 (numfiles - oldnfiles) *
3269 sizeof(*fdp->fd_ofileflags));
3270 ofiles = fdp->fd_ofiles;
3271 fdp->fd_ofiles = newofiles;
3272 fdp->fd_ofileflags = newofileflags;
3273 fdp->fd_nfiles = numfiles;
3274 FREE_ZONE(ofiles, oldnfiles * OFILESIZE, M_OFILETABL);
3275 fdexpand++;
3276 }
3277 }
3278
3279
3280 /*
3281 * fdavail
3282 *
3283 * Description: Check to see whether n user file descriptors are available
3284 * to the process p.
3285 *
3286 * Parameters: p Process to check in
3287 * n The number of fd's desired
3288 *
3289 * Returns: 0 No
3290 * 1 Yes
3291 *
3292 * Locks: Assumes proc_fdlock for process is held by the caller
3293 *
3294 * Notes: The answer only remains valid so long as the proc_fdlock is
3295 * held by the caller.
3296 */
3297 int
3298 fdavail(proc_t p, int n)
3299 {
3300 struct filedesc *fdp = p->p_fd;
3301 struct fileproc **fpp;
3302 char *flags;
3303 int i, lim;
3304
3305 lim = min((int)p->p_rlimit[RLIMIT_NOFILE].rlim_cur, maxfiles);
3306 if ((i = lim - fdp->fd_nfiles) > 0 && (n -= i) <= 0)
3307 return (1);
3308 fpp = &fdp->fd_ofiles[fdp->fd_freefile];
3309 flags = &fdp->fd_ofileflags[fdp->fd_freefile];
3310 for (i = fdp->fd_nfiles - fdp->fd_freefile; --i >= 0; fpp++, flags++)
3311 if (*fpp == NULL && !(*flags & UF_RESERVED) && --n <= 0)
3312 return (1);
3313 return (0);
3314 }
3315
3316
3317 /*
3318 * fdrelse
3319 *
3320 * Description: Legacy KPI wrapper function for _fdrelse
3321 *
3322 * Parameters: p Process in which fd lives
3323 * fd fd to free
3324 *
3325 * Returns: void
3326 *
3327 * Locks: Assumes proc_fdlock for process is held by the caller
3328 */
3329 void
3330 fdrelse(proc_t p, int fd)
3331 {
3332 _fdrelse(p, fd);
3333 }
3334
3335
3336 /*
3337 * fdgetf_noref
3338 *
3339 * Description: Get the fileproc pointer for the given fd from the per process
3340 * open file table without taking an explicit reference on it.
3341 *
3342 * Parameters: p Process containing fd
3343 * fd fd to obtain fileproc for
3344 * resultfp Pointer to pointer return area
3345 *
3346 * Returns: 0 Success
3347 * EBADF
3348 *
3349 * Implicit returns:
3350 * *resultfp (modified) Pointer to fileproc pointer
3351 *
3352 * Locks: Assumes proc_fdlock for process is held by the caller
3353 *
3354 * Notes: Because there is no reference explicitly taken, the returned
3355 * fileproc pointer is only valid so long as the proc_fdlock
3356 * remains held by the caller.
3357 */
3358 int
3359 fdgetf_noref(proc_t p, int fd, struct fileproc **resultfp)
3360 {
3361 struct filedesc *fdp = p->p_fd;
3362 struct fileproc *fp;
3363
3364 if (fd < 0 || fd >= fdp->fd_nfiles ||
3365 (fp = fdp->fd_ofiles[fd]) == NULL ||
3366 (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
3367 return (EBADF);
3368 }
3369 if (resultfp)
3370 *resultfp = fp;
3371 return (0);
3372 }
3373
3374
3375 /*
3376 * fp_getfvp
3377 *
3378 * Description: Get fileproc and vnode pointer for a given fd from the per
3379 * process open file table of the specified process, and if
3380 * successful, increment the f_iocount
3381 *
3382 * Parameters: p Process in which fd lives
3383 * fd fd to get information for
3384 * resultfp Pointer to result fileproc
3385 * pointer area, or 0 if none
3386 * resultvp Pointer to result vnode pointer
3387 * area, or 0 if none
3388 *
3389 * Returns: 0 Success
3390 * EBADF Bad file descriptor
3391 * ENOTSUP fd does not refer to a vnode
3392 *
3393 * Implicit returns:
3394 * *resultfp (modified) Fileproc pointer
3395 * *resultvp (modified) vnode pointer
3396 *
3397 * Notes: The resultfp and resultvp fields are optional, and may be
3398 * independently specified as NULL to skip returning information
3399 *
3400 * Locks: Internally takes and releases proc_fdlock
3401 */
3402 int
3403 fp_getfvp(proc_t p, int fd, struct fileproc **resultfp, struct vnode **resultvp)
3404 {
3405 struct filedesc *fdp = p->p_fd;
3406 struct fileproc *fp;
3407
3408 proc_fdlock_spin(p);
3409 if (fd < 0 || fd >= fdp->fd_nfiles ||
3410 (fp = fdp->fd_ofiles[fd]) == NULL ||
3411 (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
3412 proc_fdunlock(p);
3413 return (EBADF);
3414 }
3415 if (fp->f_type != DTYPE_VNODE) {
3416 proc_fdunlock(p);
3417 return(ENOTSUP);
3418 }
3419 fp->f_iocount++;
3420
3421 if (resultfp)
3422 *resultfp = fp;
3423 if (resultvp)
3424 *resultvp = (struct vnode *)fp->f_data;
3425 proc_fdunlock(p);
3426
3427 return (0);
3428 }
3429
3430
3431 /*
3432 * fp_getfvpandvid
3433 *
3434 * Description: Get fileproc, vnode pointer, and vid for a given fd from the
3435 * per process open file table of the specified process, and if
3436 * successful, increment the f_iocount
3437 *
3438 * Parameters: p Process in which fd lives
3439 * fd fd to get information for
3440 * resultfp Pointer to result fileproc
3441 * pointer area, or 0 if none
3442 * resultvp Pointer to result vnode pointer
3443 * area, or 0 if none
3444 * vidp Pointer to resuld vid area
3445 *
3446 * Returns: 0 Success
3447 * EBADF Bad file descriptor
3448 * ENOTSUP fd does not refer to a vnode
3449 *
3450 * Implicit returns:
3451 * *resultfp (modified) Fileproc pointer
3452 * *resultvp (modified) vnode pointer
3453 * *vidp vid value
3454 *
3455 * Notes: The resultfp and resultvp fields are optional, and may be
3456 * independently specified as NULL to skip returning information
3457 *
3458 * Locks: Internally takes and releases proc_fdlock
3459 */
3460 int
3461 fp_getfvpandvid(proc_t p, int fd, struct fileproc **resultfp,
3462 struct vnode **resultvp, uint32_t *vidp)
3463 {
3464 struct filedesc *fdp = p->p_fd;
3465 struct fileproc *fp;
3466
3467 proc_fdlock_spin(p);
3468 if (fd < 0 || fd >= fdp->fd_nfiles ||
3469 (fp = fdp->fd_ofiles[fd]) == NULL ||
3470 (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
3471 proc_fdunlock(p);
3472 return (EBADF);
3473 }
3474 if (fp->f_type != DTYPE_VNODE) {
3475 proc_fdunlock(p);
3476 return(ENOTSUP);
3477 }
3478 fp->f_iocount++;
3479
3480 if (resultfp)
3481 *resultfp = fp;
3482 if (resultvp)
3483 *resultvp = (struct vnode *)fp->f_data;
3484 if (vidp)
3485 *vidp = (uint32_t)vnode_vid((struct vnode *)fp->f_data);
3486 proc_fdunlock(p);
3487
3488 return (0);
3489 }
3490
3491
3492 /*
3493 * fp_getfsock
3494 *
3495 * Description: Get fileproc and socket pointer for a given fd from the
3496 * per process open file table of the specified process, and if
3497 * successful, increment the f_iocount
3498 *
3499 * Parameters: p Process in which fd lives
3500 * fd fd to get information for
3501 * resultfp Pointer to result fileproc
3502 * pointer area, or 0 if none
3503 * results Pointer to result socket
3504 * pointer area, or 0 if none
3505 *
3506 * Returns: EBADF The file descriptor is invalid
3507 * EOPNOTSUPP The file descriptor is not a socket
3508 * 0 Success
3509 *
3510 * Implicit returns:
3511 * *resultfp (modified) Fileproc pointer
3512 * *results (modified) socket pointer
3513 *
3514 * Notes: EOPNOTSUPP should probably be ENOTSOCK; this function is only
3515 * ever called from accept1().
3516 */
3517 int
3518 fp_getfsock(proc_t p, int fd, struct fileproc **resultfp,
3519 struct socket **results)
3520 {
3521 struct filedesc *fdp = p->p_fd;
3522 struct fileproc *fp;
3523
3524 proc_fdlock_spin(p);
3525 if (fd < 0 || fd >= fdp->fd_nfiles ||
3526 (fp = fdp->fd_ofiles[fd]) == NULL ||
3527 (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
3528 proc_fdunlock(p);
3529 return (EBADF);
3530 }
3531 if (fp->f_type != DTYPE_SOCKET) {
3532 proc_fdunlock(p);
3533 return(EOPNOTSUPP);
3534 }
3535 fp->f_iocount++;
3536
3537 if (resultfp)
3538 *resultfp = fp;
3539 if (results)
3540 *results = (struct socket *)fp->f_data;
3541 proc_fdunlock(p);
3542
3543 return (0);
3544 }
3545
3546
3547 /*
3548 * fp_getfkq
3549 *
3550 * Description: Get fileproc and kqueue pointer for a given fd from the
3551 * per process open file table of the specified process, and if
3552 * successful, increment the f_iocount
3553 *
3554 * Parameters: p Process in which fd lives
3555 * fd fd to get information for
3556 * resultfp Pointer to result fileproc
3557 * pointer area, or 0 if none
3558 * resultkq Pointer to result kqueue
3559 * pointer area, or 0 if none
3560 *
3561 * Returns: EBADF The file descriptor is invalid
3562 * EBADF The file descriptor is not a socket
3563 * 0 Success
3564 *
3565 * Implicit returns:
3566 * *resultfp (modified) Fileproc pointer
3567 * *resultkq (modified) kqueue pointer
3568 *
3569 * Notes: The second EBADF should probably be something else to make
3570 * the error condition distinct.
3571 */
3572 int
3573 fp_getfkq(proc_t p, int fd, struct fileproc **resultfp,
3574 struct kqueue **resultkq)
3575 {
3576 struct filedesc *fdp = p->p_fd;
3577 struct fileproc *fp;
3578
3579 proc_fdlock_spin(p);
3580 if ( fd < 0 || fd >= fdp->fd_nfiles ||
3581 (fp = fdp->fd_ofiles[fd]) == NULL ||
3582 (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
3583 proc_fdunlock(p);
3584 return (EBADF);
3585 }
3586 if (fp->f_type != DTYPE_KQUEUE) {
3587 proc_fdunlock(p);
3588 return(EBADF);
3589 }
3590 fp->f_iocount++;
3591
3592 if (resultfp)
3593 *resultfp = fp;
3594 if (resultkq)
3595 *resultkq = (struct kqueue *)fp->f_data;
3596 proc_fdunlock(p);
3597
3598 return (0);
3599 }
3600
3601
3602 /*
3603 * fp_getfpshm
3604 *
3605 * Description: Get fileproc and POSIX shared memory pointer for a given fd
3606 * from the per process open file table of the specified process
3607 * and if successful, increment the f_iocount
3608 *
3609 * Parameters: p Process in which fd lives
3610 * fd fd to get information for
3611 * resultfp Pointer to result fileproc
3612 * pointer area, or 0 if none
3613 * resultpshm Pointer to result POSIX
3614 * shared memory pointer
3615 * pointer area, or 0 if none
3616 *
3617 * Returns: EBADF The file descriptor is invalid
3618 * EBADF The file descriptor is not a POSIX
3619 * shared memory area
3620 * 0 Success
3621 *
3622 * Implicit returns:
3623 * *resultfp (modified) Fileproc pointer
3624 * *resultpshm (modified) POSIX shared memory pointer
3625 *
3626 * Notes: The second EBADF should probably be something else to make
3627 * the error condition distinct.
3628 */
3629 int
3630 fp_getfpshm(proc_t p, int fd, struct fileproc **resultfp,
3631 struct pshmnode **resultpshm)
3632 {
3633 struct filedesc *fdp = p->p_fd;
3634 struct fileproc *fp;
3635
3636 proc_fdlock_spin(p);
3637 if (fd < 0 || fd >= fdp->fd_nfiles ||
3638 (fp = fdp->fd_ofiles[fd]) == NULL ||
3639 (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
3640 proc_fdunlock(p);
3641 return (EBADF);
3642 }
3643 if (fp->f_type != DTYPE_PSXSHM) {
3644
3645 proc_fdunlock(p);
3646 return(EBADF);
3647 }
3648 fp->f_iocount++;
3649
3650 if (resultfp)
3651 *resultfp = fp;
3652 if (resultpshm)
3653 *resultpshm = (struct pshmnode *)fp->f_data;
3654 proc_fdunlock(p);
3655
3656 return (0);
3657 }
3658
3659
3660 /*
3661 * fp_getfsem
3662 *
3663 * Description: Get fileproc and POSIX semaphore pointer for a given fd from
3664 * the per process open file table of the specified process
3665 * and if successful, increment the f_iocount
3666 *
3667 * Parameters: p Process in which fd lives
3668 * fd fd to get information for
3669 * resultfp Pointer to result fileproc
3670 * pointer area, or 0 if none
3671 * resultpsem Pointer to result POSIX
3672 * semaphore pointer area, or
3673 * 0 if none
3674 *
3675 * Returns: EBADF The file descriptor is invalid
3676 * EBADF The file descriptor is not a POSIX
3677 * semaphore
3678 * 0 Success
3679 *
3680 * Implicit returns:
3681 * *resultfp (modified) Fileproc pointer
3682 * *resultpsem (modified) POSIX semaphore pointer
3683 *
3684 * Notes: The second EBADF should probably be something else to make
3685 * the error condition distinct.
3686 *
3687 * In order to support unnamed POSIX semaphores, the named
3688 * POSIX semaphores will have to move out of the per-process
3689 * open filetable, and into a global table that is shared with
3690 * unnamed POSIX semaphores, since unnamed POSIX semaphores
3691 * are typically used by declaring instances in shared memory,
3692 * and there's no other way to do this without changing the
3693 * underlying type, which would introduce binary compatibility
3694 * issues.
3695 */
3696 int
3697 fp_getfpsem(proc_t p, int fd, struct fileproc **resultfp,
3698 struct psemnode **resultpsem)
3699 {
3700 struct filedesc *fdp = p->p_fd;
3701 struct fileproc *fp;
3702
3703 proc_fdlock_spin(p);
3704 if (fd < 0 || fd >= fdp->fd_nfiles ||
3705 (fp = fdp->fd_ofiles[fd]) == NULL ||
3706 (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
3707 proc_fdunlock(p);
3708 return (EBADF);
3709 }
3710 if (fp->f_type != DTYPE_PSXSEM) {
3711 proc_fdunlock(p);
3712 return(EBADF);
3713 }
3714 fp->f_iocount++;
3715
3716 if (resultfp)
3717 *resultfp = fp;
3718 if (resultpsem)
3719 *resultpsem = (struct psemnode *)fp->f_data;
3720 proc_fdunlock(p);
3721
3722 return (0);
3723 }
3724
3725
3726 /*
3727 * fp_getfpipe
3728 *
3729 * Description: Get fileproc and pipe pointer for a given fd from the
3730 * per process open file table of the specified process
3731 * and if successful, increment the f_iocount
3732 *
3733 * Parameters: p Process in which fd lives
3734 * fd fd to get information for
3735 * resultfp Pointer to result fileproc
3736 * pointer area, or 0 if none
3737 * resultpipe Pointer to result pipe
3738 * pointer area, or 0 if none
3739 *
3740 * Returns: EBADF The file descriptor is invalid
3741 * EBADF The file descriptor is not a socket
3742 * 0 Success
3743 *
3744 * Implicit returns:
3745 * *resultfp (modified) Fileproc pointer
3746 * *resultpipe (modified) pipe pointer
3747 *
3748 * Notes: The second EBADF should probably be something else to make
3749 * the error condition distinct.
3750 */
3751 int
3752 fp_getfpipe(proc_t p, int fd, struct fileproc **resultfp,
3753 struct pipe **resultpipe)
3754 {
3755 struct filedesc *fdp = p->p_fd;
3756 struct fileproc *fp;
3757
3758 proc_fdlock_spin(p);
3759 if (fd < 0 || fd >= fdp->fd_nfiles ||
3760 (fp = fdp->fd_ofiles[fd]) == NULL ||
3761 (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
3762 proc_fdunlock(p);
3763 return (EBADF);
3764 }
3765 if (fp->f_type != DTYPE_PIPE) {
3766 proc_fdunlock(p);
3767 return(EBADF);
3768 }
3769 fp->f_iocount++;
3770
3771 if (resultfp)
3772 *resultfp = fp;
3773 if (resultpipe)
3774 *resultpipe = (struct pipe *)fp->f_data;
3775 proc_fdunlock(p);
3776
3777 return (0);
3778 }
3779
3780
3781 /*
3782 * fp_lookup
3783 *
3784 * Description: Get fileproc pointer for a given fd from the per process
3785 * open file table of the specified process and if successful,
3786 * increment the f_iocount
3787 *
3788 * Parameters: p Process in which fd lives
3789 * fd fd to get information for
3790 * resultfp Pointer to result fileproc
3791 * pointer area, or 0 if none
3792 * locked !0 if the caller holds the
3793 * proc_fdlock, 0 otherwise
3794 *
3795 * Returns: 0 Success
3796 * EBADF Bad file descriptor
3797 *
3798 * Implicit returns:
3799 * *resultfp (modified) Fileproc pointer
3800 *
3801 * Locks: If the argument 'locked' is non-zero, then the caller is
3802 * expected to have taken and held the proc_fdlock; if it is
3803 * zero, than this routine internally takes and drops this lock.
3804 */
3805 int
3806 fp_lookup(proc_t p, int fd, struct fileproc **resultfp, int locked)
3807 {
3808 struct filedesc *fdp = p->p_fd;
3809 struct fileproc *fp;
3810
3811 if (!locked)
3812 proc_fdlock_spin(p);
3813 if (fd < 0 || fdp == NULL || fd >= fdp->fd_nfiles ||
3814 (fp = fdp->fd_ofiles[fd]) == NULL ||
3815 (fdp->fd_ofileflags[fd] & UF_RESERVED)) {
3816 if (!locked)
3817 proc_fdunlock(p);
3818 return (EBADF);
3819 }
3820 fp->f_iocount++;
3821
3822 if (resultfp)
3823 *resultfp = fp;
3824 if (!locked)
3825 proc_fdunlock(p);
3826
3827 return (0);
3828 }
3829
3830
3831 /*
3832 * fp_tryswap
3833 *
3834 * Description: Swap the fileproc pointer for a given fd with a new
3835 * fileproc pointer in the per-process open file table of
3836 * the specified process. The fdlock must be held at entry.
3837 *
3838 * Parameters: p Process containing the fd
3839 * fd The fd of interest
3840 * nfp Pointer to the newfp
3841 *
3842 * Returns: 0 Success
3843 * EBADF Bad file descriptor
3844 * EINTR Interrupted
3845 * EKEEPLOOKING f_iocount changed while lock was dropped.
3846 */
3847 int
3848 fp_tryswap(proc_t p, int fd, struct fileproc *nfp)
3849 {
3850 struct fileproc *fp;
3851 int error;
3852
3853 proc_fdlock_assert(p, LCK_MTX_ASSERT_OWNED);
3854
3855 if (0 != (error = fp_lookup(p, fd, &fp, 1)))
3856 return (error);
3857 /*
3858 * At this point, our caller (change_guardedfd_np) has
3859 * one f_iocount reference, and we just took another
3860 * one to begin the replacement.
3861 */
3862 if (fp->f_iocount < 2) {
3863 panic("f_iocount too small %d", fp->f_iocount);
3864 } else if (2 == fp->f_iocount) {
3865
3866 /* Copy the contents of *fp, preserving the "type" of *nfp */
3867
3868 nfp->f_flags = (nfp->f_flags & FP_TYPEMASK) |
3869 (fp->f_flags & ~FP_TYPEMASK);
3870 nfp->f_iocount = fp->f_iocount;
3871 nfp->f_fglob = fp->f_fglob;
3872 nfp->f_wset = fp->f_wset;
3873
3874 p->p_fd->fd_ofiles[fd] = nfp;
3875 (void) fp_drop(p, fd, nfp, 1);
3876 } else {
3877 /*
3878 * Wait for all other active references to evaporate.
3879 */
3880 p->p_fpdrainwait = 1;
3881 error = msleep(&p->p_fpdrainwait, &p->p_fdmlock,
3882 PRIBIO | PCATCH, "tryswap fpdrain", NULL);
3883 if (0 == error) {
3884 /*
3885 * Return an "internal" errno to trigger a full
3886 * reevaluation of the change-guard attempt.
3887 */
3888 error = EKEEPLOOKING;
3889 printf("%s: lookup collision fd %d\n", __func__, fd);
3890 }
3891 (void) fp_drop(p, fd, fp, 1);
3892 }
3893 return (error);
3894 }
3895
3896
3897 /*
3898 * fp_drop_written
3899 *
3900 * Description: Set the FP_WRITTEN flag on the fileproc and drop the I/O
3901 * reference previously taken by calling fp_lookup et. al.
3902 *
3903 * Parameters: p Process in which the fd lives
3904 * fd fd associated with the fileproc
3905 * fp fileproc on which to set the
3906 * flag and drop the reference
3907 *
3908 * Returns: 0 Success
3909 * fp_drop:EBADF Bad file descriptor
3910 *
3911 * Locks: This function internally takes and drops the proc_fdlock for
3912 * the supplied process
3913 *
3914 * Notes: The fileproc must correspond to the fd in the supplied proc
3915 */
3916 int
3917 fp_drop_written(proc_t p, int fd, struct fileproc *fp)
3918 {
3919 int error;
3920
3921 proc_fdlock_spin(p);
3922
3923 fp->f_flags |= FP_WRITTEN;
3924
3925 error = fp_drop(p, fd, fp, 1);
3926
3927 proc_fdunlock(p);
3928
3929 return (error);
3930 }
3931
3932
3933 /*
3934 * fp_drop_event
3935 *
3936 * Description: Set the FP_WAITEVENT flag on the fileproc and drop the I/O
3937 * reference previously taken by calling fp_lookup et. al.
3938 *
3939 * Parameters: p Process in which the fd lives
3940 * fd fd associated with the fileproc
3941 * fp fileproc on which to set the
3942 * flag and drop the reference
3943 *
3944 * Returns: 0 Success
3945 * fp_drop:EBADF Bad file descriptor
3946 *
3947 * Locks: This function internally takes and drops the proc_fdlock for
3948 * the supplied process
3949 *
3950 * Notes: The fileproc must correspond to the fd in the supplied proc
3951 */
3952 int
3953 fp_drop_event(proc_t p, int fd, struct fileproc *fp)
3954 {
3955 int error;
3956
3957 proc_fdlock_spin(p);
3958
3959 fp->f_flags |= FP_WAITEVENT;
3960
3961 error = fp_drop(p, fd, fp, 1);
3962
3963 proc_fdunlock(p);
3964
3965 return (error);
3966 }
3967
3968
3969 /*
3970 * fp_drop
3971 *
3972 * Description: Drop the I/O reference previously taken by calling fp_lookup
3973 * et. al.
3974 *
3975 * Parameters: p Process in which the fd lives
3976 * fd fd associated with the fileproc
3977 * fp fileproc on which to set the
3978 * flag and drop the reference
3979 * locked flag to internally take and
3980 * drop proc_fdlock if it is not
3981 * already held by the caller
3982 *
3983 * Returns: 0 Success
3984 * EBADF Bad file descriptor
3985 *
3986 * Locks: This function internally takes and drops the proc_fdlock for
3987 * the supplied process if 'locked' is non-zero, and assumes that
3988 * the caller already holds this lock if 'locked' is non-zero.
3989 *
3990 * Notes: The fileproc must correspond to the fd in the supplied proc
3991 */
3992 int
3993 fp_drop(proc_t p, int fd, struct fileproc *fp, int locked)
3994 {
3995 struct filedesc *fdp = p->p_fd;
3996 int needwakeup = 0;
3997
3998 if (!locked)
3999 proc_fdlock_spin(p);
4000 if ((fp == FILEPROC_NULL) && (fd < 0 || fd >= fdp->fd_nfiles ||
4001 (fp = fdp->fd_ofiles[fd]) == NULL ||
4002 ((fdp->fd_ofileflags[fd] & UF_RESERVED) &&
4003 !(fdp->fd_ofileflags[fd] & UF_CLOSING)))) {
4004 if (!locked)
4005 proc_fdunlock(p);
4006 return (EBADF);
4007 }
4008 fp->f_iocount--;
4009
4010 if (fp->f_iocount == 0) {
4011 if (fp->f_flags & FP_SELCONFLICT)
4012 fp->f_flags &= ~FP_SELCONFLICT;
4013
4014 if (p->p_fpdrainwait) {
4015 p->p_fpdrainwait = 0;
4016 needwakeup = 1;
4017 }
4018 }
4019 if (!locked)
4020 proc_fdunlock(p);
4021 if (needwakeup)
4022 wakeup(&p->p_fpdrainwait);
4023
4024 return (0);
4025 }
4026
4027
4028 /*
4029 * file_vnode
4030 *
4031 * Description: Given an fd, look it up in the current process's per process
4032 * open file table, and return its internal vnode pointer.
4033 *
4034 * Parameters: fd fd to obtain vnode from
4035 * vpp pointer to vnode return area
4036 *
4037 * Returns: 0 Success
4038 * EINVAL The fd does not refer to a
4039 * vnode fileproc entry
4040 * fp_lookup:EBADF Bad file descriptor
4041 *
4042 * Implicit returns:
4043 * *vpp (modified) Returned vnode pointer
4044 *
4045 * Locks: This function internally takes and drops the proc_fdlock for
4046 * the current process
4047 *
4048 * Notes: If successful, this function increments the f_iocount on the
4049 * fd's corresponding fileproc.
4050 *
4051 * The fileproc referenced is not returned; because of this, care
4052 * must be taken to not drop the last reference (e.g. by closing
4053 * the file). This is inherently unsafe, since the reference may
4054 * not be recoverable from the vnode, if there is a subsequent
4055 * close that destroys the associate fileproc. The caller should
4056 * therefore retain their own reference on the fileproc so that
4057 * the f_iocount can be dropped subsequently. Failure to do this
4058 * can result in the returned pointer immediately becoming invalid
4059 * following the call.
4060 *
4061 * Use of this function is discouraged.
4062 */
4063 int
4064 file_vnode(int fd, struct vnode **vpp)
4065 {
4066 proc_t p = current_proc();
4067 struct fileproc *fp;
4068 int error;
4069
4070 proc_fdlock_spin(p);
4071 if ( (error = fp_lookup(p, fd, &fp, 1)) ) {
4072 proc_fdunlock(p);
4073 return(error);
4074 }
4075 if (fp->f_type != DTYPE_VNODE) {
4076 fp_drop(p, fd, fp,1);
4077 proc_fdunlock(p);
4078 return(EINVAL);
4079 }
4080 if (vpp != NULL)
4081 *vpp = (struct vnode *)fp->f_data;
4082 proc_fdunlock(p);
4083
4084 return(0);
4085 }
4086
4087
4088 /*
4089 * file_vnode_withvid
4090 *
4091 * Description: Given an fd, look it up in the current process's per process
4092 * open file table, and return its internal vnode pointer.
4093 *
4094 * Parameters: fd fd to obtain vnode from
4095 * vpp pointer to vnode return area
4096 * vidp pointer to vid of the returned vnode
4097 *
4098 * Returns: 0 Success
4099 * EINVAL The fd does not refer to a
4100 * vnode fileproc entry
4101 * fp_lookup:EBADF Bad file descriptor
4102 *
4103 * Implicit returns:
4104 * *vpp (modified) Returned vnode pointer
4105 *
4106 * Locks: This function internally takes and drops the proc_fdlock for
4107 * the current process
4108 *
4109 * Notes: If successful, this function increments the f_iocount on the
4110 * fd's corresponding fileproc.
4111 *
4112 * The fileproc referenced is not returned; because of this, care
4113 * must be taken to not drop the last reference (e.g. by closing
4114 * the file). This is inherently unsafe, since the reference may
4115 * not be recoverable from the vnode, if there is a subsequent
4116 * close that destroys the associate fileproc. The caller should
4117 * therefore retain their own reference on the fileproc so that
4118 * the f_iocount can be dropped subsequently. Failure to do this
4119 * can result in the returned pointer immediately becoming invalid
4120 * following the call.
4121 *
4122 * Use of this function is discouraged.
4123 */
4124 int
4125 file_vnode_withvid(int fd, struct vnode **vpp, uint32_t * vidp)
4126 {
4127 proc_t p = current_proc();
4128 struct fileproc *fp;
4129 vnode_t vp;
4130 int error;
4131
4132 proc_fdlock_spin(p);
4133 if ( (error = fp_lookup(p, fd, &fp, 1)) ) {
4134 proc_fdunlock(p);
4135 return(error);
4136 }
4137 if (fp->f_type != DTYPE_VNODE) {
4138 fp_drop(p, fd, fp,1);
4139 proc_fdunlock(p);
4140 return(EINVAL);
4141 }
4142 vp = (struct vnode *)fp->f_data;
4143 if (vpp != NULL)
4144 *vpp = vp;
4145
4146 if ((vidp != NULL) && (vp != NULLVP))
4147 *vidp = (uint32_t)vp->v_id;
4148
4149 proc_fdunlock(p);
4150
4151 return(0);
4152 }
4153
4154
4155 /*
4156 * file_socket
4157 *
4158 * Description: Given an fd, look it up in the current process's per process
4159 * open file table, and return its internal socket pointer.
4160 *
4161 * Parameters: fd fd to obtain vnode from
4162 * sp pointer to socket return area
4163 *
4164 * Returns: 0 Success
4165 * ENOTSOCK Not a socket
4166 * fp_lookup:EBADF Bad file descriptor
4167 *
4168 * Implicit returns:
4169 * *sp (modified) Returned socket pointer
4170 *
4171 * Locks: This function internally takes and drops the proc_fdlock for
4172 * the current process
4173 *
4174 * Notes: If successful, this function increments the f_iocount on the
4175 * fd's corresponding fileproc.
4176 *
4177 * The fileproc referenced is not returned; because of this, care
4178 * must be taken to not drop the last reference (e.g. by closing
4179 * the file). This is inherently unsafe, since the reference may
4180 * not be recoverable from the socket, if there is a subsequent
4181 * close that destroys the associate fileproc. The caller should
4182 * therefore retain their own reference on the fileproc so that
4183 * the f_iocount can be dropped subsequently. Failure to do this
4184 * can result in the returned pointer immediately becoming invalid
4185 * following the call.
4186 *
4187 * Use of this function is discouraged.
4188 */
4189 int
4190 file_socket(int fd, struct socket **sp)
4191 {
4192 proc_t p = current_proc();
4193 struct fileproc *fp;
4194 int error;
4195
4196 proc_fdlock_spin(p);
4197 if ( (error = fp_lookup(p, fd, &fp, 1)) ) {
4198 proc_fdunlock(p);
4199 return(error);
4200 }
4201 if (fp->f_type != DTYPE_SOCKET) {
4202 fp_drop(p, fd, fp,1);
4203 proc_fdunlock(p);
4204 return(ENOTSOCK);
4205 }
4206 *sp = (struct socket *)fp->f_data;
4207 proc_fdunlock(p);
4208
4209 return(0);
4210 }
4211
4212
4213 /*
4214 * file_flags
4215 *
4216 * Description: Given an fd, look it up in the current process's per process
4217 * open file table, and return its fileproc's flags field.
4218 *
4219 * Parameters: fd fd whose flags are to be
4220 * retrieved
4221 * flags pointer to flags data area
4222 *
4223 * Returns: 0 Success
4224 * ENOTSOCK Not a socket
4225 * fp_lookup:EBADF Bad file descriptor
4226 *
4227 * Implicit returns:
4228 * *flags (modified) Returned flags field
4229 *
4230 * Locks: This function internally takes and drops the proc_fdlock for
4231 * the current process
4232 *
4233 * Notes: This function will internally increment and decrement the
4234 * f_iocount of the fileproc as part of its operation.
4235 */
4236 int
4237 file_flags(int fd, int *flags)
4238 {
4239
4240 proc_t p = current_proc();
4241 struct fileproc *fp;
4242 int error;
4243
4244 proc_fdlock_spin(p);
4245 if ( (error = fp_lookup(p, fd, &fp, 1)) ) {
4246 proc_fdunlock(p);
4247 return(error);
4248 }
4249 *flags = (int)fp->f_flag;
4250 fp_drop(p, fd, fp,1);
4251 proc_fdunlock(p);
4252
4253 return(0);
4254 }
4255
4256
4257 /*
4258 * file_drop
4259 *
4260 * Description: Drop an iocount reference on an fd, and wake up any waiters
4261 * for draining (i.e. blocked in fileproc_drain() called during
4262 * the last attempt to close a file).
4263 *
4264 * Parameters: fd fd on which an ioreference is
4265 * to be dropped
4266 *
4267 * Returns: 0 Success
4268 * EBADF Bad file descriptor
4269 *
4270 * Description: Given an fd, look it up in the current process's per process
4271 * open file table, and drop it's fileproc's f_iocount by one
4272 *
4273 * Notes: This is intended as a corresponding operation to the functions
4274 * file_vnode() and file_socket() operations.
4275 *
4276 * Technically, the close reference is supposed to be protected
4277 * by a fileproc_drain(), however, a drain will only block if
4278 * the fd refers to a character device, and that device has had
4279 * preparefileread() called on it. If it refers to something
4280 * other than a character device, then the drain will occur and
4281 * block each close attempt, rather than merely the last close.
4282 *
4283 * Since it's possible for an fd that refers to a character
4284 * device to have an intermediate close followed by an open to
4285 * cause a different file to correspond to that descriptor,
4286 * unless there was a cautionary reference taken on the fileproc,
4287 * this is an inherently unsafe function. This happens in the
4288 * case where multiple fd's in a process refer to the same
4289 * character device (e.g. stdin/out/err pointing to a tty, etc.).
4290 *
4291 * Use of this function is discouraged.
4292 */
4293 int
4294 file_drop(int fd)
4295 {
4296 struct fileproc *fp;
4297 proc_t p = current_proc();
4298 int needwakeup = 0;
4299
4300 proc_fdlock_spin(p);
4301 if (fd < 0 || fd >= p->p_fd->fd_nfiles ||
4302 (fp = p->p_fd->fd_ofiles[fd]) == NULL ||
4303 ((p->p_fd->fd_ofileflags[fd] & UF_RESERVED) &&
4304 !(p->p_fd->fd_ofileflags[fd] & UF_CLOSING))) {
4305 proc_fdunlock(p);
4306 return (EBADF);
4307 }
4308 fp->f_iocount --;
4309
4310 if (fp->f_iocount == 0) {
4311 if (fp->f_flags & FP_SELCONFLICT)
4312 fp->f_flags &= ~FP_SELCONFLICT;
4313
4314 if (p->p_fpdrainwait) {
4315 p->p_fpdrainwait = 0;
4316 needwakeup = 1;
4317 }
4318 }
4319 proc_fdunlock(p);
4320
4321 if (needwakeup)
4322 wakeup(&p->p_fpdrainwait);
4323 return(0);
4324 }
4325
4326
4327 static int falloc_withalloc_locked(proc_t, struct fileproc **, int *,
4328 vfs_context_t, struct fileproc * (*)(void *), void *, int);
4329
4330 /*
4331 * falloc
4332 *
4333 * Description: Allocate an entry in the per process open file table and
4334 * return the corresponding fileproc and fd.
4335 *
4336 * Parameters: p The process in whose open file
4337 * table the fd is to be allocated
4338 * resultfp Pointer to fileproc pointer
4339 * return area
4340 * resultfd Pointer to fd return area
4341 * ctx VFS context
4342 *
4343 * Returns: 0 Success
4344 * falloc:ENFILE Too many open files in system
4345 * falloc:EMFILE Too many open files in process
4346 * falloc:ENOMEM M_FILEPROC or M_FILEGLOB zone
4347 * exhausted
4348 *
4349 * Implicit returns:
4350 * *resultfd (modified) Returned fileproc pointer
4351 * *resultfd (modified) Returned fd
4352 *
4353 * Locks: This function takes and drops the proc_fdlock; if this lock
4354 * is already held, use falloc_locked() instead.
4355 *
4356 * Notes: This function takes separate process and context arguments
4357 * solely to support kern_exec.c; otherwise, it would take
4358 * neither, and expect falloc_locked() to use the
4359 * vfs_context_current() routine internally.
4360 */
4361 int
4362 falloc(proc_t p, struct fileproc **resultfp, int *resultfd, vfs_context_t ctx)
4363 {
4364 return (falloc_withalloc(p, resultfp, resultfd, ctx,
4365 fileproc_alloc_init, NULL));
4366 }
4367
4368 /*
4369 * Like falloc, but including the fileproc allocator and create-args
4370 */
4371 int
4372 falloc_withalloc(proc_t p, struct fileproc **resultfp, int *resultfd,
4373 vfs_context_t ctx, fp_allocfn_t fp_zalloc, void *arg)
4374 {
4375 int error;
4376
4377 proc_fdlock(p);
4378 error = falloc_withalloc_locked(p,
4379 resultfp, resultfd, ctx, fp_zalloc, arg, 1);
4380 proc_fdunlock(p);
4381
4382 return (error);
4383 }
4384
4385 /*
4386 * "uninitialized" ops -- ensure fg->fg_ops->fo_type always exists
4387 */
4388 static const struct fileops uninitops;
4389
4390 /*
4391 * falloc_locked
4392 *
4393 * Create a new open file structure and allocate
4394 * a file descriptor for the process that refers to it.
4395 *
4396 * Returns: 0 Success
4397 *
4398 * Description: Allocate an entry in the per process open file table and
4399 * return the corresponding fileproc and fd.
4400 *
4401 * Parameters: p The process in whose open file
4402 * table the fd is to be allocated
4403 * resultfp Pointer to fileproc pointer
4404 * return area
4405 * resultfd Pointer to fd return area
4406 * ctx VFS context
4407 * locked Flag to indicate whether the
4408 * caller holds proc_fdlock
4409 *
4410 * Returns: 0 Success
4411 * ENFILE Too many open files in system
4412 * fdalloc:EMFILE Too many open files in process
4413 * ENOMEM M_FILEPROC or M_FILEGLOB zone
4414 * exhausted
4415 * fdalloc:ENOMEM
4416 *
4417 * Implicit returns:
4418 * *resultfd (modified) Returned fileproc pointer
4419 * *resultfd (modified) Returned fd
4420 *
4421 * Locks: If the parameter 'locked' is zero, this function takes and
4422 * drops the proc_fdlock; if non-zero, the caller must hold the
4423 * lock.
4424 *
4425 * Notes: If you intend to use a non-zero 'locked' parameter, use the
4426 * utility function falloc() instead.
4427 *
4428 * This function takes separate process and context arguments
4429 * solely to support kern_exec.c; otherwise, it would take
4430 * neither, and use the vfs_context_current() routine internally.
4431 */
4432 int
4433 falloc_locked(proc_t p, struct fileproc **resultfp, int *resultfd,
4434 vfs_context_t ctx, int locked)
4435 {
4436 return (falloc_withalloc_locked(p, resultfp, resultfd, ctx,
4437 fileproc_alloc_init, NULL, locked));
4438 }
4439
4440 static int
4441 falloc_withalloc_locked(proc_t p, struct fileproc **resultfp, int *resultfd,
4442 vfs_context_t ctx, fp_allocfn_t fp_zalloc, void *crarg,
4443 int locked)
4444 {
4445 struct fileproc *fp;
4446 struct fileglob *fg;
4447 int error, nfd;
4448
4449 if (!locked)
4450 proc_fdlock(p);
4451 if ( (error = fdalloc(p, 0, &nfd)) ) {
4452 if (!locked)
4453 proc_fdunlock(p);
4454 return (error);
4455 }
4456 if (nfiles >= maxfiles) {
4457 if (!locked)
4458 proc_fdunlock(p);
4459 tablefull("file");
4460 return (ENFILE);
4461 }
4462 #if CONFIG_MACF
4463 error = mac_file_check_create(proc_ucred(p));
4464 if (error) {
4465 if (!locked)
4466 proc_fdunlock(p);
4467 return (error);
4468 }
4469 #endif
4470
4471 /*
4472 * Allocate a new file descriptor.
4473 * If the process has file descriptor zero open, add to the list
4474 * of open files at that point, otherwise put it at the front of
4475 * the list of open files.
4476 */
4477 proc_fdunlock(p);
4478
4479 fp = (*fp_zalloc)(crarg);
4480 if (fp == NULL) {
4481 if (locked)
4482 proc_fdlock(p);
4483 return (ENOMEM);
4484 }
4485 MALLOC_ZONE(fg, struct fileglob *, sizeof(struct fileglob), M_FILEGLOB, M_WAITOK);
4486 if (fg == NULL) {
4487 fileproc_free(fp);
4488 if (locked)
4489 proc_fdlock(p);
4490 return (ENOMEM);
4491 }
4492 bzero(fg, sizeof(struct fileglob));
4493 lck_mtx_init(&fg->fg_lock, file_lck_grp, file_lck_attr);
4494
4495 fp->f_iocount = 1;
4496 fg->fg_count = 1;
4497 fg->fg_ops = &uninitops;
4498 fp->f_fglob = fg;
4499 #if CONFIG_MACF
4500 mac_file_label_init(fg);
4501 #endif
4502
4503 kauth_cred_ref(ctx->vc_ucred);
4504
4505 proc_fdlock(p);
4506
4507 fp->f_cred = ctx->vc_ucred;
4508
4509 #if CONFIG_MACF
4510 mac_file_label_associate(fp->f_cred, fg);
4511 #endif
4512
4513 OSAddAtomic(1, &nfiles);
4514
4515 p->p_fd->fd_ofiles[nfd] = fp;
4516
4517 if (!locked)
4518 proc_fdunlock(p);
4519
4520 if (resultfp)
4521 *resultfp = fp;
4522 if (resultfd)
4523 *resultfd = nfd;
4524
4525 return (0);
4526 }
4527
4528
4529 /*
4530 * fg_free
4531 *
4532 * Description: Free a file structure; drop the global open file count, and
4533 * drop the credential reference, if the fileglob has one, and
4534 * destroy the instance mutex before freeing
4535 *
4536 * Parameters: fg Pointer to fileglob to be
4537 * freed
4538 *
4539 * Returns: void
4540 */
4541 void
4542 fg_free(struct fileglob *fg)
4543 {
4544 OSAddAtomic(-1, &nfiles);
4545
4546 if (fg->fg_vn_data) {
4547 fg_vn_data_free(fg->fg_vn_data);
4548 fg->fg_vn_data = NULL;
4549 }
4550
4551 if (IS_VALID_CRED(fg->fg_cred)) {
4552 kauth_cred_unref(&fg->fg_cred);
4553 }
4554 lck_mtx_destroy(&fg->fg_lock, file_lck_grp);
4555
4556 #if CONFIG_MACF
4557 mac_file_label_destroy(fg);
4558 #endif
4559 FREE_ZONE(fg, sizeof *fg, M_FILEGLOB);
4560 }
4561
4562
4563 /*
4564 * fdexec
4565 *
4566 * Description: Perform close-on-exec processing for all files in a process
4567 * that are either marked as close-on-exec, or which were in the
4568 * process of being opened at the time of the execve
4569 *
4570 * Also handles the case (via posix_spawn()) where -all-
4571 * files except those marked with "inherit" as treated as
4572 * close-on-exec.
4573 *
4574 * Parameters: p Pointer to process calling
4575 * execve
4576 *
4577 * Returns: void
4578 *
4579 * Locks: This function internally takes and drops proc_fdlock()
4580 *
4581 */
4582 void
4583 fdexec(proc_t p, short flags)
4584 {
4585 struct filedesc *fdp = p->p_fd;
4586 int i;
4587 boolean_t cloexec_default = (flags & POSIX_SPAWN_CLOEXEC_DEFAULT) != 0;
4588
4589 proc_fdlock(p);
4590 for (i = fdp->fd_lastfile; i >= 0; i--) {
4591
4592 struct fileproc *fp = fdp->fd_ofiles[i];
4593 char *flagp = &fdp->fd_ofileflags[i];
4594
4595 if (fp && cloexec_default) {
4596 /*
4597 * Reverse the usual semantics of file descriptor
4598 * inheritance - all of them should be closed
4599 * except files marked explicitly as "inherit" and
4600 * not marked close-on-exec.
4601 */
4602 if ((*flagp & (UF_EXCLOSE|UF_INHERIT)) != UF_INHERIT)
4603 *flagp |= UF_EXCLOSE;
4604 *flagp &= ~UF_INHERIT;
4605 }
4606
4607 if (
4608 ((*flagp & (UF_RESERVED|UF_EXCLOSE)) == UF_EXCLOSE)
4609 #if CONFIG_MACF
4610 || (fp && mac_file_check_inherit(proc_ucred(p), fp->f_fglob))
4611 #endif
4612 ) {
4613 if (i < fdp->fd_knlistsize)
4614 knote_fdclose(p, i, TRUE);
4615 procfdtbl_clearfd(p, i);
4616 if (i == fdp->fd_lastfile && i > 0)
4617 fdp->fd_lastfile--;
4618 if (i < fdp->fd_freefile)
4619 fdp->fd_freefile = i;
4620
4621 /*
4622 * Wait for any third party viewers (e.g., lsof)
4623 * to release their references to this fileproc.
4624 */
4625 while (fp->f_iocount > 0) {
4626 p->p_fpdrainwait = 1;
4627 msleep(&p->p_fpdrainwait, &p->p_fdmlock, PRIBIO,
4628 "fpdrain", NULL);
4629 }
4630
4631 closef_locked(fp, fp->f_fglob, p);
4632
4633 fileproc_free(fp);
4634 }
4635 }
4636 proc_fdunlock(p);
4637 }
4638
4639
4640 /*
4641 * fdcopy
4642 *
4643 * Description: Copy a filedesc structure. This is normally used as part of
4644 * forkproc() when forking a new process, to copy the per process
4645 * open file table over to the new process.
4646 *
4647 * Parameters: p Process whose open file table
4648 * is to be copied (parent)
4649 * uth_cdir Per thread current working
4650 * cirectory, or NULL
4651 *
4652 * Returns: NULL Copy failed
4653 * !NULL Pointer to new struct filedesc
4654 *
4655 * Locks: This function internally takes and drops proc_fdlock()
4656 *
4657 * Notes: Files are copied directly, ignoring the new resource limits
4658 * for the process that's being copied into. Since the descriptor
4659 * references are just additional references, this does not count
4660 * against the number of open files on the system.
4661 *
4662 * The struct filedesc includes the current working directory,
4663 * and the current root directory, if the process is chroot'ed.
4664 *
4665 * If the exec was called by a thread using a per thread current
4666 * working directory, we inherit the working directory from the
4667 * thread making the call, rather than from the process.
4668 *
4669 * In the case of a failure to obtain a reference, for most cases,
4670 * the file entry will be silently dropped. There's an exception
4671 * for the case of a chroot dir, since a failure to to obtain a
4672 * reference there would constitute an "escape" from the chroot
4673 * environment, which must not be allowed. In that case, we will
4674 * deny the execve() operation, rather than allowing the escape.
4675 */
4676 struct filedesc *
4677 fdcopy(proc_t p, vnode_t uth_cdir)
4678 {
4679 struct filedesc *newfdp, *fdp = p->p_fd;
4680 int i;
4681 struct fileproc *ofp, *fp;
4682 vnode_t v_dir;
4683
4684 MALLOC_ZONE(newfdp, struct filedesc *,
4685 sizeof(*newfdp), M_FILEDESC, M_WAITOK);
4686 if (newfdp == NULL)
4687 return(NULL);
4688
4689 proc_fdlock(p);
4690
4691 /*
4692 * the FD_CHROOT flag will be inherited via this copy
4693 */
4694 (void) memcpy(newfdp, fdp, sizeof(*newfdp));
4695
4696 /*
4697 * If we are running with per-thread current working directories,
4698 * inherit the new current working directory from the current thread
4699 * instead, before we take our references.
4700 */
4701 if (uth_cdir != NULLVP)
4702 newfdp->fd_cdir = uth_cdir;
4703
4704 /*
4705 * For both fd_cdir and fd_rdir make sure we get
4706 * a valid reference... if we can't, than set
4707 * set the pointer(s) to NULL in the child... this
4708 * will keep us from using a non-referenced vp
4709 * and allows us to do the vnode_rele only on
4710 * a properly referenced vp
4711 */
4712 if ( (v_dir = newfdp->fd_cdir) ) {
4713 if (vnode_getwithref(v_dir) == 0) {
4714 if ( (vnode_ref(v_dir)) )
4715 newfdp->fd_cdir = NULL;
4716 vnode_put(v_dir);
4717 } else
4718 newfdp->fd_cdir = NULL;
4719 }
4720 if (newfdp->fd_cdir == NULL && fdp->fd_cdir) {
4721 /*
4722 * we couldn't get a new reference on
4723 * the current working directory being
4724 * inherited... we might as well drop
4725 * our reference from the parent also
4726 * since the vnode has gone DEAD making
4727 * it useless... by dropping it we'll
4728 * be that much closer to recycling it
4729 */
4730 vnode_rele(fdp->fd_cdir);
4731 fdp->fd_cdir = NULL;
4732 }
4733
4734 if ( (v_dir = newfdp->fd_rdir) ) {
4735 if (vnode_getwithref(v_dir) == 0) {
4736 if ( (vnode_ref(v_dir)) )
4737 newfdp->fd_rdir = NULL;
4738 vnode_put(v_dir);
4739 } else {
4740 newfdp->fd_rdir = NULL;
4741 }
4742 }
4743 /* Coming from a chroot environment and unable to get a reference... */
4744 if (newfdp->fd_rdir == NULL && fdp->fd_rdir) {
4745 /*
4746 * We couldn't get a new reference on
4747 * the chroot directory being
4748 * inherited... this is fatal, since
4749 * otherwise it would constitute an
4750 * escape from a chroot environment by
4751 * the new process.
4752 */
4753 if (newfdp->fd_cdir)
4754 vnode_rele(newfdp->fd_cdir);
4755 FREE_ZONE(newfdp, sizeof *newfdp, M_FILEDESC);
4756 return(NULL);
4757 }
4758
4759 /*
4760 * If the number of open files fits in the internal arrays
4761 * of the open file structure, use them, otherwise allocate
4762 * additional memory for the number of descriptors currently
4763 * in use.
4764 */
4765 if (newfdp->fd_lastfile < NDFILE)
4766 i = NDFILE;
4767 else {
4768 /*
4769 * Compute the smallest multiple of NDEXTENT needed
4770 * for the file descriptors currently in use,
4771 * allowing the table to shrink.
4772 */
4773 i = newfdp->fd_nfiles;
4774 while (i > 1 + 2 * NDEXTENT && i > 1 + newfdp->fd_lastfile * 2)
4775 i /= 2;
4776 }
4777 proc_fdunlock(p);
4778
4779 MALLOC_ZONE(newfdp->fd_ofiles, struct fileproc **,
4780 i * OFILESIZE, M_OFILETABL, M_WAITOK);
4781 if (newfdp->fd_ofiles == NULL) {
4782 if (newfdp->fd_cdir)
4783 vnode_rele(newfdp->fd_cdir);
4784 if (newfdp->fd_rdir)
4785 vnode_rele(newfdp->fd_rdir);
4786
4787 FREE_ZONE(newfdp, sizeof(*newfdp), M_FILEDESC);
4788 return(NULL);
4789 }
4790 (void) memset(newfdp->fd_ofiles, 0, i * OFILESIZE);
4791 proc_fdlock(p);
4792
4793 newfdp->fd_ofileflags = (char *) &newfdp->fd_ofiles[i];
4794 newfdp->fd_nfiles = i;
4795
4796 if (fdp->fd_nfiles > 0) {
4797 struct fileproc **fpp;
4798 char *flags;
4799
4800 (void) memcpy(newfdp->fd_ofiles, fdp->fd_ofiles,
4801 (newfdp->fd_lastfile + 1) * sizeof(*fdp->fd_ofiles));
4802 (void) memcpy(newfdp->fd_ofileflags, fdp->fd_ofileflags,
4803 (newfdp->fd_lastfile + 1) * sizeof(*fdp->fd_ofileflags));
4804
4805 /*
4806 * kq descriptors cannot be copied.
4807 */
4808 if (newfdp->fd_knlistsize != -1) {
4809 fpp = &newfdp->fd_ofiles[newfdp->fd_lastfile];
4810 flags = &newfdp->fd_ofileflags[newfdp->fd_lastfile];
4811 for (i = newfdp->fd_lastfile;
4812 i >= 0; i--, fpp--, flags--) {
4813 if (*flags & UF_RESERVED)
4814 continue; /* (removed below) */
4815 if (*fpp != NULL && (*fpp)->f_type == DTYPE_KQUEUE) {
4816 *fpp = NULL;
4817 *flags = 0;
4818 if (i < newfdp->fd_freefile)
4819 newfdp->fd_freefile = i;
4820 }
4821 if (*fpp == NULL && i == newfdp->fd_lastfile && i > 0)
4822 newfdp->fd_lastfile--;
4823 }
4824 newfdp->fd_knlist = NULL;
4825 newfdp->fd_knlistsize = -1;
4826 newfdp->fd_knhash = NULL;
4827 newfdp->fd_knhashmask = 0;
4828 }
4829 fpp = newfdp->fd_ofiles;
4830 flags = newfdp->fd_ofileflags;
4831
4832 for (i = newfdp->fd_lastfile + 1; --i >= 0; fpp++, flags++)
4833 if ((ofp = *fpp) != NULL &&
4834 0 == (ofp->f_fglob->fg_lflags & FG_CONFINED) &&
4835 0 == (*flags & (UF_FORKCLOSE|UF_RESERVED))) {
4836 #if DEBUG
4837 if (FILEPROC_TYPE(ofp) != FTYPE_SIMPLE)
4838 panic("complex fileproc");
4839 #endif
4840 fp = fileproc_alloc_init(NULL);
4841 if (fp == NULL) {
4842 /*
4843 * XXX no room to copy, unable to
4844 * XXX safely unwind state at present
4845 */
4846 *fpp = NULL;
4847 } else {
4848 fp->f_flags |=
4849 (ofp->f_flags & ~FP_TYPEMASK);
4850 fp->f_fglob = ofp->f_fglob;
4851 (void)fg_ref(fp);
4852 *fpp = fp;
4853 }
4854 } else {
4855 if (i < newfdp->fd_freefile)
4856 newfdp->fd_freefile = i;
4857 *fpp = NULL;
4858 *flags = 0;
4859 }
4860 }
4861
4862 proc_fdunlock(p);
4863 return (newfdp);
4864 }
4865
4866
4867 /*
4868 * fdfree
4869 *
4870 * Description: Release a filedesc (per process open file table) structure;
4871 * this is done on process exit(), or from forkproc_free() if
4872 * the fork fails for some reason subsequent to a successful
4873 * call to fdcopy()
4874 *
4875 * Parameters: p Pointer to process going away
4876 *
4877 * Returns: void
4878 *
4879 * Locks: This function internally takes and drops proc_fdlock()
4880 */
4881 void
4882 fdfree(proc_t p)
4883 {
4884 struct filedesc *fdp;
4885 struct fileproc *fp;
4886 int i;
4887
4888 proc_fdlock(p);
4889
4890 if (p == kernproc || NULL == (fdp = p->p_fd)) {
4891 proc_fdunlock(p);
4892 return;
4893 }
4894
4895 extern struct filedesc filedesc0;
4896
4897 if (&filedesc0 == fdp)
4898 panic("filedesc0");
4899
4900 if (fdp->fd_nfiles > 0 && fdp->fd_ofiles) {
4901 for (i = fdp->fd_lastfile; i >= 0; i--) {
4902
4903 /* May still have knotes for fd without open file */
4904 if (i < fdp->fd_knlistsize)
4905 knote_fdclose(p, i, TRUE);
4906
4907 if ((fp = fdp->fd_ofiles[i]) != NULL) {
4908
4909 if (fdp->fd_ofileflags[i] & UF_RESERVED)
4910 panic("fdfree: found fp with UF_RESERVED");
4911
4912 procfdtbl_reservefd(p, i);
4913
4914 if (fp->f_flags & FP_WAITEVENT)
4915 (void)waitevent_close(p, fp);
4916 (void) closef_locked(fp, fp->f_fglob, p);
4917 fileproc_free(fp);
4918 }
4919 }
4920 FREE_ZONE(fdp->fd_ofiles, fdp->fd_nfiles * OFILESIZE, M_OFILETABL);
4921 fdp->fd_ofiles = NULL;
4922 fdp->fd_nfiles = 0;
4923 }
4924
4925 proc_fdunlock(p);
4926
4927 if (fdp->fd_cdir)
4928 vnode_rele(fdp->fd_cdir);
4929 if (fdp->fd_rdir)
4930 vnode_rele(fdp->fd_rdir);
4931
4932 proc_fdlock_spin(p);
4933 p->p_fd = NULL;
4934 proc_fdunlock(p);
4935
4936 if (fdp->fd_knlist)
4937 FREE(fdp->fd_knlist, M_KQUEUE);
4938 if (fdp->fd_knhash)
4939 FREE(fdp->fd_knhash, M_KQUEUE);
4940
4941 FREE_ZONE(fdp, sizeof(*fdp), M_FILEDESC);
4942 }
4943
4944 /*
4945 * closef_locked
4946 *
4947 * Description: Internal form of closef; called with proc_fdlock held
4948 *
4949 * Parameters: fp Pointer to fileproc for fd
4950 * fg Pointer to fileglob for fd
4951 * p Pointer to proc structure
4952 *
4953 * Returns: 0 Success
4954 * closef_finish:??? Anything returnable by a per-fileops
4955 * close function
4956 *
4957 * Note: Decrements reference count on file structure; if this was the
4958 * last reference, then closef_finish() is called
4959 *
4960 * p and fp are allowed to be NULL when closing a file that was
4961 * being passed in a message (but only if we are called when this
4962 * is NOT the last reference).
4963 */
4964 int
4965 closef_locked(struct fileproc *fp, struct fileglob *fg, proc_t p)
4966 {
4967 struct vnode *vp;
4968 struct flock lf;
4969 struct vfs_context context;
4970 int error;
4971
4972 if (fg == NULL) {
4973 return (0);
4974 }
4975
4976 /* Set up context with cred stashed in fg */
4977 if (p == current_proc())
4978 context.vc_thread = current_thread();
4979 else
4980 context.vc_thread = NULL;
4981 context.vc_ucred = fg->fg_cred;
4982
4983 /*
4984 * POSIX record locking dictates that any close releases ALL
4985 * locks owned by this process. This is handled by setting
4986 * a flag in the unlock to free ONLY locks obeying POSIX
4987 * semantics, and not to free BSD-style file locks.
4988 * If the descriptor was in a message, POSIX-style locks
4989 * aren't passed with the descriptor.
4990 */
4991 if (p && (p->p_ladvflag & P_LADVLOCK) &&
4992 DTYPE_VNODE == FILEGLOB_DTYPE(fg)) {
4993 proc_fdunlock(p);
4994
4995 lf.l_whence = SEEK_SET;
4996 lf.l_start = 0;
4997 lf.l_len = 0;
4998 lf.l_type = F_UNLCK;
4999 vp = (struct vnode *)fg->fg_data;
5000
5001 if ( (error = vnode_getwithref(vp)) == 0 ) {
5002 (void) VNOP_ADVLOCK(vp, (caddr_t)p, F_UNLCK, &lf, F_POSIX, &context, NULL);
5003 (void)vnode_put(vp);
5004 }
5005 proc_fdlock(p);
5006 }
5007 lck_mtx_lock_spin(&fg->fg_lock);
5008 fg->fg_count--;
5009
5010 if (fg->fg_count > 0) {
5011 lck_mtx_unlock(&fg->fg_lock);
5012 return (0);
5013 }
5014 #if DIAGNOSTIC
5015 if (fg->fg_count != 0)
5016 panic("fg %p: being freed with bad fg_count (%d)", fg, fg->fg_count);
5017 #endif
5018
5019 if (fp && (fp->f_flags & FP_WRITTEN))
5020 fg->fg_flag |= FWASWRITTEN;
5021
5022 fg->fg_lflags |= FG_TERM;
5023 lck_mtx_unlock(&fg->fg_lock);
5024
5025 if (p)
5026 proc_fdunlock(p);
5027
5028 /* Since we ensure that fg->fg_ops is always initialized,
5029 * it is safe to invoke fo_close on the fg */
5030 error = fo_close(fg, &context);
5031
5032 fg_free(fg);
5033
5034 if (p)
5035 proc_fdlock(p);
5036
5037 return(error);
5038 }
5039
5040
5041 /*
5042 * fileproc_drain
5043 *
5044 * Description: Drain out pending I/O operations
5045 *
5046 * Parameters: p Process closing this file
5047 * fp fileproc struct for the open
5048 * instance on the file
5049 *
5050 * Returns: void
5051 *
5052 * Locks: Assumes the caller holds the proc_fdlock
5053 *
5054 * Notes: For character devices, this occurs on the last close of the
5055 * device; for all other file descriptors, this occurs on each
5056 * close to prevent fd's from being closed out from under
5057 * operations currently in progress and blocked
5058 *
5059 * See Also: file_vnode(), file_socket(), file_drop(), and the cautions
5060 * regarding their use and interaction with this function.
5061 */
5062 void
5063 fileproc_drain(proc_t p, struct fileproc * fp)
5064 {
5065 struct vfs_context context;
5066
5067 context.vc_thread = proc_thread(p); /* XXX */
5068 context.vc_ucred = fp->f_fglob->fg_cred;
5069
5070 fp->f_iocount-- ; /* (the one the close holds) */
5071
5072 while (fp->f_iocount) {
5073
5074 lck_mtx_convert_spin(&p->p_fdmlock);
5075
5076 if (fp->f_fglob->fg_ops->fo_drain) {
5077 (*fp->f_fglob->fg_ops->fo_drain)(fp, &context);
5078 }
5079 if ((fp->f_flags & FP_INSELECT) == FP_INSELECT) {
5080 if (waitq_wakeup64_all((struct waitq *)fp->f_wset, NO_EVENT64,
5081 THREAD_INTERRUPTED, WAITQ_ALL_PRIORITIES) == KERN_INVALID_ARGUMENT)
5082 panic("bad wait queue for waitq_wakeup64_all %p (fp:%p)", fp->f_wset, fp);
5083 }
5084 if ((fp->f_flags & FP_SELCONFLICT) == FP_SELCONFLICT) {
5085 if (waitq_wakeup64_all(&select_conflict_queue, NO_EVENT64,
5086 THREAD_INTERRUPTED, WAITQ_ALL_PRIORITIES) == KERN_INVALID_ARGUMENT)
5087 panic("bad select_conflict_queue");
5088 }
5089 p->p_fpdrainwait = 1;
5090
5091 msleep(&p->p_fpdrainwait, &p->p_fdmlock, PRIBIO, "fpdrain", NULL);
5092
5093 }
5094 #if DIAGNOSTIC
5095 if ((fp->f_flags & FP_INSELECT) != 0)
5096 panic("FP_INSELECT set on drained fp");
5097 #endif
5098 if ((fp->f_flags & FP_SELCONFLICT) == FP_SELCONFLICT)
5099 fp->f_flags &= ~FP_SELCONFLICT;
5100 }
5101
5102
5103 /*
5104 * fp_free
5105 *
5106 * Description: Release the fd and free the fileproc associated with the fd
5107 * in the per process open file table of the specified process;
5108 * these values must correspond.
5109 *
5110 * Parameters: p Process containing fd
5111 * fd fd to be released
5112 * fp fileproc to be freed
5113 *
5114 * Returns: 0 Success
5115 *
5116 * Notes: XXX function should be void - no one interprets the returns
5117 * XXX code
5118 */
5119 int
5120 fp_free(proc_t p, int fd, struct fileproc * fp)
5121 {
5122 proc_fdlock_spin(p);
5123 fdrelse(p, fd);
5124 proc_fdunlock(p);
5125
5126 fg_free(fp->f_fglob);
5127 fileproc_free(fp);
5128 return(0);
5129 }
5130
5131
5132 /*
5133 * flock
5134 *
5135 * Description: Apply an advisory lock on a file descriptor.
5136 *
5137 * Parameters: p Process making request
5138 * uap->fd fd on which the lock is to be
5139 * attempted
5140 * uap->how (Un)Lock bits, including type
5141 * retval Pointer to the call return area
5142 *
5143 * Returns: 0 Success
5144 * fp_getfvp:EBADF Bad file descriptor
5145 * fp_getfvp:ENOTSUP fd does not refer to a vnode
5146 * vnode_getwithref:???
5147 * VNOP_ADVLOCK:???
5148 *
5149 * Implicit returns:
5150 * *retval (modified) Size of dtable
5151 *
5152 * Notes: Just attempt to get a record lock of the requested type on
5153 * the entire file (l_whence = SEEK_SET, l_start = 0, l_len = 0).
5154 */
5155 int
5156 flock(proc_t p, struct flock_args *uap, __unused int32_t *retval)
5157 {
5158 int fd = uap->fd;
5159 int how = uap->how;
5160 struct fileproc *fp;
5161 struct vnode *vp;
5162 struct flock lf;
5163 vfs_context_t ctx = vfs_context_current();
5164 int error=0;
5165
5166 AUDIT_ARG(fd, uap->fd);
5167 if ( (error = fp_getfvp(p, fd, &fp, &vp)) ) {
5168 return(error);
5169 }
5170 if ( (error = vnode_getwithref(vp)) ) {
5171 goto out1;
5172 }
5173 AUDIT_ARG(vnpath, vp, ARG_VNODE1);
5174
5175 lf.l_whence = SEEK_SET;
5176 lf.l_start = 0;
5177 lf.l_len = 0;
5178 if (how & LOCK_UN) {
5179 lf.l_type = F_UNLCK;
5180 fp->f_flag &= ~FHASLOCK;
5181 error = VNOP_ADVLOCK(vp, (caddr_t)fp->f_fglob, F_UNLCK, &lf, F_FLOCK, ctx, NULL);
5182 goto out;
5183 }
5184 if (how & LOCK_EX)
5185 lf.l_type = F_WRLCK;
5186 else if (how & LOCK_SH)
5187 lf.l_type = F_RDLCK;
5188 else {
5189 error = EBADF;
5190 goto out;
5191 }
5192 #if CONFIG_MACF
5193 error = mac_file_check_lock(proc_ucred(p), fp->f_fglob, F_SETLK, &lf);
5194 if (error)
5195 goto out;
5196 #endif
5197 error = VNOP_ADVLOCK(vp, (caddr_t)fp->f_fglob, F_SETLK, &lf,
5198 (how & LOCK_NB ? F_FLOCK : F_FLOCK | F_WAIT),
5199 ctx, NULL);
5200 if (!error)
5201 fp->f_flag |= FHASLOCK;
5202 out:
5203 (void)vnode_put(vp);
5204 out1:
5205 fp_drop(p, fd, fp, 0);
5206 return(error);
5207
5208 }
5209
5210 /*
5211 * fileport_makeport
5212 *
5213 * Description: Obtain a Mach send right for a given file descriptor.
5214 *
5215 * Parameters: p Process calling fileport
5216 * uap->fd The fd to reference
5217 * uap->portnamep User address at which to place port name.
5218 *
5219 * Returns: 0 Success.
5220 * EBADF Bad file descriptor.
5221 * EINVAL File descriptor had type that cannot be sent, misc. other errors.
5222 * EFAULT Address at which to store port name is not valid.
5223 * EAGAIN Resource shortage.
5224 *
5225 * Implicit returns:
5226 * On success, name of send right is stored at user-specified address.
5227 */
5228 int
5229 fileport_makeport(proc_t p, struct fileport_makeport_args *uap,
5230 __unused int *retval)
5231 {
5232 int err;
5233 int fd = uap->fd;
5234 user_addr_t user_portaddr = uap->portnamep;
5235 struct fileproc *fp = FILEPROC_NULL;
5236 struct fileglob *fg = NULL;
5237 ipc_port_t fileport;
5238 mach_port_name_t name = MACH_PORT_NULL;
5239
5240 proc_fdlock(p);
5241 err = fp_lookup(p, fd, &fp, 1);
5242 if (err != 0) {
5243 goto out_unlock;
5244 }
5245
5246 if (!file_issendable(p, fp)) {
5247 err = EINVAL;
5248 goto out_unlock;
5249 }
5250
5251 if (FP_ISGUARDED(fp, GUARD_FILEPORT)) {
5252 err = fp_guard_exception(p, fd, fp, kGUARD_EXC_FILEPORT);
5253 goto out_unlock;
5254 }
5255
5256 /* Dropped when port is deallocated */
5257 fg = fp->f_fglob;
5258 fg_ref(fp);
5259
5260 proc_fdunlock(p);
5261
5262 /* Allocate and initialize a port */
5263 fileport = fileport_alloc(fg);
5264 if (fileport == IPC_PORT_NULL) {
5265 err = EAGAIN;
5266 fg_drop(fp);
5267 goto out;
5268 }
5269
5270 /* Add an entry. Deallocates port on failure. */
5271 name = ipc_port_copyout_send(fileport, get_task_ipcspace(p->task));
5272 if (!MACH_PORT_VALID(name)) {
5273 err = EINVAL;
5274 goto out;
5275 }
5276
5277 err = copyout(&name, user_portaddr, sizeof(mach_port_name_t));
5278 if (err != 0) {
5279 goto out;
5280 }
5281
5282 /* Tag the fileglob for debugging purposes */
5283 lck_mtx_lock_spin(&fg->fg_lock);
5284 fg->fg_lflags |= FG_PORTMADE;
5285 lck_mtx_unlock(&fg->fg_lock);
5286
5287 fp_drop(p, fd, fp, 0);
5288
5289 return 0;
5290
5291 out_unlock:
5292 proc_fdunlock(p);
5293 out:
5294 if (MACH_PORT_VALID(name)) {
5295 /* Don't care if another thread races us to deallocate the entry */
5296 (void) mach_port_deallocate(get_task_ipcspace(p->task), name);
5297 }
5298
5299 if (fp != FILEPROC_NULL) {
5300 fp_drop(p, fd, fp, 0);
5301 }
5302
5303 return err;
5304 }
5305
5306 void
5307 fileport_releasefg(struct fileglob *fg)
5308 {
5309 (void)closef_locked(NULL, fg, PROC_NULL);
5310
5311 return;
5312 }
5313
5314
5315 /*
5316 * fileport_makefd
5317 *
5318 * Description: Obtain the file descriptor for a given Mach send right.
5319 *
5320 * Parameters: p Process calling fileport
5321 * uap->port Name of send right to file port.
5322 *
5323 * Returns: 0 Success
5324 * EINVAL Invalid Mach port name, or port is not for a file.
5325 * fdalloc:EMFILE
5326 * fdalloc:ENOMEM Unable to allocate fileproc or extend file table.
5327 *
5328 * Implicit returns:
5329 * *retval (modified) The new descriptor
5330 */
5331 int
5332 fileport_makefd(proc_t p, struct fileport_makefd_args *uap, int32_t *retval)
5333 {
5334 struct fileglob *fg;
5335 struct fileproc *fp = FILEPROC_NULL;
5336 ipc_port_t port = IPC_PORT_NULL;
5337 mach_port_name_t send = uap->port;
5338 kern_return_t res;
5339 int fd;
5340 int err;
5341
5342 res = ipc_object_copyin(get_task_ipcspace(p->task),
5343 send, MACH_MSG_TYPE_COPY_SEND, &port);
5344
5345 if (res != KERN_SUCCESS) {
5346 err = EINVAL;
5347 goto out;
5348 }
5349
5350 fg = fileport_port_to_fileglob(port);
5351 if (fg == NULL) {
5352 err = EINVAL;
5353 goto out;
5354 }
5355
5356 fp = fileproc_alloc_init(NULL);
5357 if (fp == FILEPROC_NULL) {
5358 err = ENOMEM;
5359 goto out;
5360 }
5361
5362 fp->f_fglob = fg;
5363 fg_ref(fp);
5364
5365 proc_fdlock(p);
5366 err = fdalloc(p, 0, &fd);
5367 if (err != 0) {
5368 proc_fdunlock(p);
5369 goto out;
5370 }
5371 *fdflags(p, fd) |= UF_EXCLOSE;
5372
5373 procfdtbl_releasefd(p, fd, fp);
5374 proc_fdunlock(p);
5375
5376 *retval = fd;
5377 err = 0;
5378 out:
5379 if ((fp != NULL) && (0 != err)) {
5380 fileproc_free(fp);
5381 }
5382
5383 if (IPC_PORT_NULL != port) {
5384 ipc_port_release_send(port);
5385 }
5386
5387 return err;
5388 }
5389
5390
5391 /*
5392 * dupfdopen
5393 *
5394 * Description: Duplicate the specified descriptor to a free descriptor;
5395 * this is the second half of fdopen(), above.
5396 *
5397 * Parameters: fdp filedesc pointer to fill in
5398 * indx fd to dup to
5399 * dfd fd to dup from
5400 * mode mode to set on new fd
5401 * error command code
5402 *
5403 * Returns: 0 Success
5404 * EBADF Source fd is bad
5405 * EACCES Requested mode not allowed
5406 * !0 'error', if not ENODEV or
5407 * ENXIO
5408 *
5409 * Notes: XXX This is not thread safe; see fdopen() above
5410 */
5411 int
5412 dupfdopen(struct filedesc *fdp, int indx, int dfd, int flags, int error)
5413 {
5414 struct fileproc *wfp;
5415 struct fileproc *fp;
5416 #if CONFIG_MACF
5417 int myerror;
5418 #endif
5419 proc_t p = current_proc();
5420
5421 /*
5422 * If the to-be-dup'd fd number is greater than the allowed number
5423 * of file descriptors, or the fd to be dup'd has already been
5424 * closed, reject. Note, check for new == old is necessary as
5425 * falloc could allocate an already closed to-be-dup'd descriptor
5426 * as the new descriptor.
5427 */
5428 proc_fdlock(p);
5429
5430 fp = fdp->fd_ofiles[indx];
5431 if (dfd < 0 || dfd >= fdp->fd_nfiles ||
5432 (wfp = fdp->fd_ofiles[dfd]) == NULL || wfp == fp ||
5433 (fdp->fd_ofileflags[dfd] & UF_RESERVED)) {
5434
5435 proc_fdunlock(p);
5436 return (EBADF);
5437 }
5438 #if CONFIG_MACF
5439 myerror = mac_file_check_dup(proc_ucred(p), wfp->f_fglob, dfd);
5440 if (myerror) {
5441 proc_fdunlock(p);
5442 return (myerror);
5443 }
5444 #endif
5445 /*
5446 * There are two cases of interest here.
5447 *
5448 * For ENODEV simply dup (dfd) to file descriptor
5449 * (indx) and return.
5450 *
5451 * For ENXIO steal away the file structure from (dfd) and
5452 * store it in (indx). (dfd) is effectively closed by
5453 * this operation.
5454 *
5455 * Any other error code is just returned.
5456 */
5457 switch (error) {
5458 case ENODEV:
5459 if (FP_ISGUARDED(wfp, GUARD_DUP)) {
5460 proc_fdunlock(p);
5461 return (EPERM);
5462 }
5463
5464 /*
5465 * Check that the mode the file is being opened for is a
5466 * subset of the mode of the existing descriptor.
5467 */
5468 if (((flags & (FREAD|FWRITE)) | wfp->f_flag) != wfp->f_flag) {
5469 proc_fdunlock(p);
5470 return (EACCES);
5471 }
5472 if (indx > fdp->fd_lastfile)
5473 fdp->fd_lastfile = indx;
5474 (void)fg_ref(wfp);
5475
5476 if (fp->f_fglob)
5477 fg_free(fp->f_fglob);
5478 fp->f_fglob = wfp->f_fglob;
5479
5480 fdp->fd_ofileflags[indx] = fdp->fd_ofileflags[dfd] |
5481 (flags & O_CLOEXEC) ? UF_EXCLOSE : 0;
5482
5483 proc_fdunlock(p);
5484 return (0);
5485
5486 default:
5487 proc_fdunlock(p);
5488 return (error);
5489 }
5490 /* NOTREACHED */
5491 }
5492
5493
5494 /*
5495 * fg_ref
5496 *
5497 * Description: Add a reference to a fileglob by fileproc
5498 *
5499 * Parameters: fp fileproc containing fileglob
5500 * pointer
5501 *
5502 * Returns: void
5503 *
5504 * Notes: XXX Should use OSAddAtomic?
5505 */
5506 void
5507 fg_ref(struct fileproc * fp)
5508 {
5509 struct fileglob *fg;
5510
5511 fg = fp->f_fglob;
5512
5513 lck_mtx_lock_spin(&fg->fg_lock);
5514
5515 #if DIAGNOSTIC
5516 if ((fp->f_flags & ~((unsigned int)FP_VALID_FLAGS)) != 0)
5517 panic("fg_ref: invalid bits on fp %p", fp);
5518
5519 if (fg->fg_count == 0)
5520 panic("fg_ref: adding fgcount to zeroed fg: fp %p fg %p",
5521 fp, fg);
5522 #endif
5523 fg->fg_count++;
5524 lck_mtx_unlock(&fg->fg_lock);
5525 }
5526
5527
5528 /*
5529 * fg_drop
5530 *
5531 * Description: Remove a reference to a fileglob by fileproc
5532 *
5533 * Parameters: fp fileproc containing fileglob
5534 * pointer
5535 *
5536 * Returns: void
5537 *
5538 * Notes: XXX Should use OSAddAtomic?
5539 */
5540 void
5541 fg_drop(struct fileproc * fp)
5542 {
5543 struct fileglob *fg;
5544
5545 fg = fp->f_fglob;
5546 lck_mtx_lock_spin(&fg->fg_lock);
5547 fg->fg_count--;
5548 lck_mtx_unlock(&fg->fg_lock);
5549 }
5550
5551 #if SOCKETS
5552 /*
5553 * fg_insertuipc_mark
5554 *
5555 * Description: Mark fileglob for insertion onto message queue if needed
5556 * Also takes fileglob reference
5557 *
5558 * Parameters: fg Fileglob pointer to insert
5559 *
5560 * Returns: true, if the fileglob needs to be inserted onto msg queue
5561 *
5562 * Locks: Takes and drops fg_lock, potentially many times
5563 */
5564 boolean_t
5565 fg_insertuipc_mark(struct fileglob * fg)
5566 {
5567 boolean_t insert = FALSE;
5568
5569 lck_mtx_lock_spin(&fg->fg_lock);
5570 while (fg->fg_lflags & FG_RMMSGQ) {
5571 lck_mtx_convert_spin(&fg->fg_lock);
5572
5573 fg->fg_lflags |= FG_WRMMSGQ;
5574 msleep(&fg->fg_lflags, &fg->fg_lock, 0, "fg_insertuipc", NULL);
5575 }
5576
5577 fg->fg_count++;
5578 fg->fg_msgcount++;
5579 if (fg->fg_msgcount == 1) {
5580 fg->fg_lflags |= FG_INSMSGQ;
5581 insert = TRUE;
5582 }
5583 lck_mtx_unlock(&fg->fg_lock);
5584 return (insert);
5585 }
5586
5587 /*
5588 * fg_insertuipc
5589 *
5590 * Description: Insert marked fileglob onto message queue
5591 *
5592 * Parameters: fg Fileglob pointer to insert
5593 *
5594 * Returns: void
5595 *
5596 * Locks: Takes and drops fg_lock & uipc_lock
5597 * DO NOT call this function with proc_fdlock held as unp_gc()
5598 * can potentially try to acquire proc_fdlock, which can result
5599 * in a deadlock if this function is in unp_gc_wait().
5600 */
5601 void
5602 fg_insertuipc(struct fileglob * fg)
5603 {
5604 if (fg->fg_lflags & FG_INSMSGQ) {
5605 lck_mtx_lock_spin(uipc_lock);
5606 unp_gc_wait();
5607 LIST_INSERT_HEAD(&fmsghead, fg, f_msglist);
5608 lck_mtx_unlock(uipc_lock);
5609 lck_mtx_lock(&fg->fg_lock);
5610 fg->fg_lflags &= ~FG_INSMSGQ;
5611 if (fg->fg_lflags & FG_WINSMSGQ) {
5612 fg->fg_lflags &= ~FG_WINSMSGQ;
5613 wakeup(&fg->fg_lflags);
5614 }
5615 lck_mtx_unlock(&fg->fg_lock);
5616 }
5617 }
5618
5619 /*
5620 * fg_removeuipc_mark
5621 *
5622 * Description: Mark the fileglob for removal from message queue if needed
5623 * Also releases fileglob message queue reference
5624 *
5625 * Parameters: fg Fileglob pointer to remove
5626 *
5627 * Returns: true, if the fileglob needs to be removed from msg queue
5628 *
5629 * Locks: Takes and drops fg_lock, potentially many times
5630 */
5631 boolean_t
5632 fg_removeuipc_mark(struct fileglob * fg)
5633 {
5634 boolean_t remove = FALSE;
5635
5636 lck_mtx_lock_spin(&fg->fg_lock);
5637 while (fg->fg_lflags & FG_INSMSGQ) {
5638 lck_mtx_convert_spin(&fg->fg_lock);
5639
5640 fg->fg_lflags |= FG_WINSMSGQ;
5641 msleep(&fg->fg_lflags, &fg->fg_lock, 0, "fg_removeuipc", NULL);
5642 }
5643 fg->fg_msgcount--;
5644 if (fg->fg_msgcount == 0) {
5645 fg->fg_lflags |= FG_RMMSGQ;
5646 remove = TRUE;
5647 }
5648 lck_mtx_unlock(&fg->fg_lock);
5649 return (remove);
5650 }
5651
5652 /*
5653 * fg_removeuipc
5654 *
5655 * Description: Remove marked fileglob from message queue
5656 *
5657 * Parameters: fg Fileglob pointer to remove
5658 *
5659 * Returns: void
5660 *
5661 * Locks: Takes and drops fg_lock & uipc_lock
5662 * DO NOT call this function with proc_fdlock held as unp_gc()
5663 * can potentially try to acquire proc_fdlock, which can result
5664 * in a deadlock if this function is in unp_gc_wait().
5665 */
5666 void
5667 fg_removeuipc(struct fileglob * fg)
5668 {
5669 if (fg->fg_lflags & FG_RMMSGQ) {
5670 lck_mtx_lock_spin(uipc_lock);
5671 unp_gc_wait();
5672 LIST_REMOVE(fg, f_msglist);
5673 lck_mtx_unlock(uipc_lock);
5674 lck_mtx_lock(&fg->fg_lock);
5675 fg->fg_lflags &= ~FG_RMMSGQ;
5676 if (fg->fg_lflags & FG_WRMMSGQ) {
5677 fg->fg_lflags &= ~FG_WRMMSGQ;
5678 wakeup(&fg->fg_lflags);
5679 }
5680 lck_mtx_unlock(&fg->fg_lock);
5681 }
5682 }
5683 #endif /* SOCKETS */
5684
5685 /*
5686 * fo_read
5687 *
5688 * Description: Generic fileops read indirected through the fileops pointer
5689 * in the fileproc structure
5690 *
5691 * Parameters: fp fileproc structure pointer
5692 * uio user I/O structure pointer
5693 * flags FOF_ flags
5694 * ctx VFS context for operation
5695 *
5696 * Returns: 0 Success
5697 * !0 Errno from read
5698 */
5699 int
5700 fo_read(struct fileproc *fp, struct uio *uio, int flags, vfs_context_t ctx)
5701 {
5702 return ((*fp->f_ops->fo_read)(fp, uio, flags, ctx));
5703 }
5704
5705
5706 /*
5707 * fo_write
5708 *
5709 * Description: Generic fileops write indirected through the fileops pointer
5710 * in the fileproc structure
5711 *
5712 * Parameters: fp fileproc structure pointer
5713 * uio user I/O structure pointer
5714 * flags FOF_ flags
5715 * ctx VFS context for operation
5716 *
5717 * Returns: 0 Success
5718 * !0 Errno from write
5719 */
5720 int
5721 fo_write(struct fileproc *fp, struct uio *uio, int flags, vfs_context_t ctx)
5722 {
5723 return((*fp->f_ops->fo_write)(fp, uio, flags, ctx));
5724 }
5725
5726
5727 /*
5728 * fo_ioctl
5729 *
5730 * Description: Generic fileops ioctl indirected through the fileops pointer
5731 * in the fileproc structure
5732 *
5733 * Parameters: fp fileproc structure pointer
5734 * com ioctl command
5735 * data pointer to internalized copy
5736 * of user space ioctl command
5737 * parameter data in kernel space
5738 * ctx VFS context for operation
5739 *
5740 * Returns: 0 Success
5741 * !0 Errno from ioctl
5742 *
5743 * Locks: The caller is assumed to have held the proc_fdlock; this
5744 * function releases and reacquires this lock. If the caller
5745 * accesses data protected by this lock prior to calling this
5746 * function, it will need to revalidate/reacquire any cached
5747 * protected data obtained prior to the call.
5748 */
5749 int
5750 fo_ioctl(struct fileproc *fp, u_long com, caddr_t data, vfs_context_t ctx)
5751 {
5752 int error;
5753
5754 proc_fdunlock(vfs_context_proc(ctx));
5755 error = (*fp->f_ops->fo_ioctl)(fp, com, data, ctx);
5756 proc_fdlock(vfs_context_proc(ctx));
5757 return(error);
5758 }
5759
5760
5761 /*
5762 * fo_select
5763 *
5764 * Description: Generic fileops select indirected through the fileops pointer
5765 * in the fileproc structure
5766 *
5767 * Parameters: fp fileproc structure pointer
5768 * which select which
5769 * wql pointer to wait queue list
5770 * ctx VFS context for operation
5771 *
5772 * Returns: 0 Success
5773 * !0 Errno from select
5774 */
5775 int
5776 fo_select(struct fileproc *fp, int which, void *wql, vfs_context_t ctx)
5777 {
5778 return((*fp->f_ops->fo_select)(fp, which, wql, ctx));
5779 }
5780
5781
5782 /*
5783 * fo_close
5784 *
5785 * Description: Generic fileops close indirected through the fileops pointer
5786 * in the fileproc structure
5787 *
5788 * Parameters: fp fileproc structure pointer for
5789 * file to close
5790 * ctx VFS context for operation
5791 *
5792 * Returns: 0 Success
5793 * !0 Errno from close
5794 */
5795 int
5796 fo_close(struct fileglob *fg, vfs_context_t ctx)
5797 {
5798 return((*fg->fg_ops->fo_close)(fg, ctx));
5799 }
5800
5801
5802 /*
5803 * fo_kqfilter
5804 *
5805 * Description: Generic fileops kqueue filter indirected through the fileops
5806 * pointer in the fileproc structure
5807 *
5808 * Parameters: fp fileproc structure pointer
5809 * kn pointer to knote to filter on
5810 * ctx VFS context for operation
5811 *
5812 * Returns: (kn->kn_flags & EV_ERROR) error in kn->kn_data
5813 * 0 Filter is not active
5814 * !0 Filter is active
5815 */
5816 int
5817 fo_kqfilter(struct fileproc *fp, struct knote *kn, vfs_context_t ctx)
5818 {
5819 return ((*fp->f_ops->fo_kqfilter)(fp, kn, ctx));
5820 }
5821
5822 /*
5823 * The ability to send a file descriptor to another
5824 * process is opt-in by file type.
5825 */
5826 boolean_t
5827 file_issendable(proc_t p, struct fileproc *fp)
5828 {
5829 proc_fdlock_assert(p, LCK_MTX_ASSERT_OWNED);
5830
5831 switch (fp->f_type) {
5832 case DTYPE_VNODE:
5833 case DTYPE_SOCKET:
5834 case DTYPE_PIPE:
5835 case DTYPE_PSXSHM:
5836 return (0 == (fp->f_fglob->fg_lflags & FG_CONFINED));
5837 default:
5838 /* DTYPE_KQUEUE, DTYPE_FSEVENTS, DTYPE_PSXSEM */
5839 return FALSE;
5840 }
5841 }
5842
5843
5844 struct fileproc *
5845 fileproc_alloc_init(__unused void *arg)
5846 {
5847 struct fileproc *fp;
5848
5849 MALLOC_ZONE(fp, struct fileproc *, sizeof (*fp), M_FILEPROC, M_WAITOK);
5850 if (fp)
5851 bzero(fp, sizeof (*fp));
5852
5853 return (fp);
5854 }
5855
5856 void
5857 fileproc_free(struct fileproc *fp)
5858 {
5859 switch (FILEPROC_TYPE(fp)) {
5860 case FTYPE_SIMPLE:
5861 FREE_ZONE(fp, sizeof (*fp), M_FILEPROC);
5862 break;
5863 case FTYPE_GUARDED:
5864 guarded_fileproc_free(fp);
5865 break;
5866 default:
5867 panic("%s: corrupt fp %p flags %x", __func__, fp, fp->f_flags);
5868 }
5869 }
mirror of XNU