1/*
2 * Copyright (c) 2000-2014 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/socket.h>
85#include <sys/socketvar.h>
86#include <sys/stat.h>
87#include <sys/ioctl.h>
88#include <sys/fcntl.h>
89#include <sys/malloc.h>
90#include <sys/mman.h>
91#include <sys/syslog.h>
92#include <sys/unistd.h>
93#include <sys/resourcevar.h>
94#include <sys/aio_kern.h>
95#include <sys/ev.h>
96#include <kern/locks.h>
97#include <sys/uio_internal.h>
98#include <sys/codesign.h>
99#include <sys/codedir_internal.h>
100
101#include <security/audit/audit.h>
102
103#include <sys/mount_internal.h>
104#include <sys/kdebug.h>
105#include <sys/sysproto.h>
106#include <sys/pipe.h>
107#include <sys/spawn.h>
108#include <kern/kern_types.h>
109#include <kern/kalloc.h>
110#include <libkern/OSAtomic.h>
111
112#include <sys/ubc_internal.h>
113
114#include <kern/ipc_misc.h>
115#include <vm/vm_protos.h>
116
117#include <mach/mach_port.h>
118#include <stdbool.h>
119
120#if CONFIG_PROTECT
121#include <sys/cprotect.h>
122#endif
123#include <hfs/hfs.h>
124
125kern_return_t ipc_object_copyin(ipc_space_t, mach_port_name_t,
126    mach_msg_type_name_t, ipc_port_t *);
127void ipc_port_release_send(ipc_port_t);
128
129struct psemnode;
130struct pshmnode;
131
132static int finishdup(proc_t p,
133    struct filedesc *fdp, int old, int new, int flags, int32_t *retval);
134
135int falloc_locked(proc_t p, struct fileproc **resultfp, int *resultfd, vfs_context_t ctx, int locked);
136void fg_drop(struct fileproc * fp);
137void fg_free(struct fileglob *fg);
138void fg_ref(struct fileproc * fp);
139void fileport_releasefg(struct fileglob *fg);
140
141/* flags for close_internal_locked */
142#define FD_DUP2RESV 1
143
144/* We don't want these exported */
145
146__private_extern__
147int unlink1(vfs_context_t, struct nameidata *, int);
148
149static void _fdrelse(struct proc * p, int fd);
150
151
152extern void file_lock_init(void);
153
154extern kauth_scope_t	kauth_scope_fileop;
155
156/* Conflict wait queue for when selects collide (opaque type) */
157extern struct wait_queue select_conflict_queue;
158
159#define f_flag f_fglob->fg_flag
160#define f_type f_fglob->fg_ops->fo_type
161#define f_msgcount f_fglob->fg_msgcount
162#define f_cred f_fglob->fg_cred
163#define f_ops f_fglob->fg_ops
164#define f_offset f_fglob->fg_offset
165#define f_data f_fglob->fg_data
166#define CHECK_ADD_OVERFLOW_INT64L(x, y) \
167		(((((x) > 0) && ((y) > 0) && ((x) > LLONG_MAX - (y))) || \
168		(((x) < 0) && ((y) < 0) && ((x) < LLONG_MIN - (y)))) \
169		? 1 : 0)
170/*
171 * Descriptor management.
172 */
173struct fmsglist fmsghead;	/* head of list of open files */
174struct fmsglist fmsg_ithead;	/* head of list of open files */
175int nfiles;			/* actual number of open files */
176
177
178lck_grp_attr_t * file_lck_grp_attr;
179lck_grp_t * file_lck_grp;
180lck_attr_t * file_lck_attr;
181
182lck_mtx_t * uipc_lock;
183
184
185/*
186 * check_file_seek_range
187 *
188 * Description: Checks if seek offsets are in the range of 0 to LLONG_MAX.
189 *
190 * Parameters:  fl		Flock structure.
191 *		cur_file_offset	Current offset in the file.
192 *
193 * Returns: 	0 		on Success.
194 *		EOVERFLOW	on overflow.
195 *		EINVAL   	on offset less than zero.
196 */
197
198static int
199check_file_seek_range(struct flock *fl, off_t cur_file_offset)
200{
201	if (fl->l_whence == SEEK_CUR) {
202		/* Check if the start marker is beyond LLONG_MAX. */
203		if (CHECK_ADD_OVERFLOW_INT64L(fl->l_start, cur_file_offset)) {
204			/* Check if start marker is negative */
205			if (fl->l_start < 0) {
206				return EINVAL;
207			}
208			return EOVERFLOW;
209		}
210		/* Check if the start marker is negative. */
211		if (fl->l_start + cur_file_offset < 0) {
212			return EINVAL;
213		}
214		/* Check if end marker is beyond LLONG_MAX. */
215		if ((fl->l_len > 0) && (CHECK_ADD_OVERFLOW_INT64L(fl->l_start +
216			cur_file_offset, fl->l_len - 1))) {
217			return EOVERFLOW;
218		}
219		/* Check if the end marker is negative. */
220		if ((fl->l_len <= 0) && (fl->l_start + cur_file_offset +
221			fl->l_len < 0)) {
222			return EINVAL;
223		}
224	} else if (fl->l_whence == SEEK_SET) {
225		/* Check if the start marker is negative. */
226		if (fl->l_start < 0) {
227			return EINVAL;
228		}
229		/* Check if the end marker is beyond LLONG_MAX. */
230		if ((fl->l_len > 0) &&
231		    CHECK_ADD_OVERFLOW_INT64L(fl->l_start, fl->l_len - 1)) {
232			return EOVERFLOW;
233		}
234		/* Check if the end marker is negative. */
235		if ((fl->l_len < 0) &&  fl->l_start + fl->l_len < 0) {
236			return EINVAL;
237		}
238	}
239	return 0;
240}
241
242
243/*
244 * file_lock_init
245 *
246 * Description:	Initialize the file lock group and the uipc and flist locks
247 *
248 * Parameters:	(void)
249 *
250 * Returns:	void
251 *
252 * Notes:	Called at system startup from bsd_init().
253 */
254void
255file_lock_init(void)
256{
257	/* allocate file lock group attribute and group */
258	file_lck_grp_attr= lck_grp_attr_alloc_init();
259
260	file_lck_grp = lck_grp_alloc_init("file",  file_lck_grp_attr);
261
262	/* Allocate file lock attribute */
263	file_lck_attr = lck_attr_alloc_init();
264
265	uipc_lock = lck_mtx_alloc_init(file_lck_grp, file_lck_attr);
266}
267
268
269/*
270 * proc_fdlock, proc_fdlock_spin
271 *
272 * Description:	Lock to control access to the per process struct fileproc
273 *		and struct filedesc
274 *
275 * Parameters:	p				Process to take the lock on
276 *
277 * Returns:	void
278 *
279 * Notes:	The lock is initialized in forkproc() and destroyed in
280 *		reap_child_process().
281 */
282void
283proc_fdlock(proc_t p)
284{
285	lck_mtx_lock(&p->p_fdmlock);
286}
287
288void
289proc_fdlock_spin(proc_t p)
290{
291	lck_mtx_lock_spin(&p->p_fdmlock);
292}
293
294void
295proc_fdlock_assert(proc_t p, int assertflags)
296{
297	lck_mtx_assert(&p->p_fdmlock, assertflags);
298}
299
300
301/*
302 * proc_fdunlock
303 *
304 * Description:	Unlock the lock previously locked by a call to proc_fdlock()
305 *
306 * Parameters:	p				Process to drop the lock on
307 *
308 * Returns:	void
309 */
310void
311proc_fdunlock(proc_t p)
312{
313	lck_mtx_unlock(&p->p_fdmlock);
314}
315
316
317/*
318 * System calls on descriptors.
319 */
320
321
322/*
323 * getdtablesize
324 *
325 * Description:	Returns the per process maximum size of the descriptor table
326 *
327 * Parameters:	p				Process being queried
328 *		retval				Pointer to the call return area
329 *
330 * Returns:	0				Success
331 *
332 * Implicit returns:
333 *		*retval (modified)		Size of dtable
334 */
335int
336getdtablesize(proc_t p, __unused struct getdtablesize_args *uap, int32_t *retval)
337{
338	proc_fdlock_spin(p);
339	*retval = min((int)p->p_rlimit[RLIMIT_NOFILE].rlim_cur, maxfiles);
340	proc_fdunlock(p);
341
342	return (0);
343}
344
345
346void
347procfdtbl_reservefd(struct proc * p, int fd)
348{
349	p->p_fd->fd_ofiles[fd] = NULL;
350        p->p_fd->fd_ofileflags[fd] |= UF_RESERVED;
351}
352
353void
354procfdtbl_markclosefd(struct proc * p, int fd)
355{
356        p->p_fd->fd_ofileflags[fd] |= (UF_RESERVED | UF_CLOSING);
357}
358
359void
360procfdtbl_releasefd(struct proc * p, int fd, struct fileproc * fp)
361{
362	if (fp != NULL)
363        	p->p_fd->fd_ofiles[fd] = fp;
364        p->p_fd->fd_ofileflags[fd] &= ~UF_RESERVED;
365	if ((p->p_fd->fd_ofileflags[fd] & UF_RESVWAIT) == UF_RESVWAIT) {
366		p->p_fd->fd_ofileflags[fd] &= ~UF_RESVWAIT;
367		wakeup(&p->p_fd);
368	}
369}
370
371void
372procfdtbl_waitfd(struct proc * p, int fd)
373{
374        p->p_fd->fd_ofileflags[fd] |= UF_RESVWAIT;
375	msleep(&p->p_fd, &p->p_fdmlock, PRIBIO, "ftbl_waitfd", NULL);
376}
377
378
379void
380procfdtbl_clearfd(struct proc * p, int fd)
381{
382	int waiting;
383
384	waiting = (p->p_fd->fd_ofileflags[fd] & UF_RESVWAIT);
385	p->p_fd->fd_ofiles[fd] = NULL;
386	p->p_fd->fd_ofileflags[fd] = 0;
387	if ( waiting == UF_RESVWAIT) {
388		wakeup(&p->p_fd);
389	}
390}
391
392/*
393 * _fdrelse
394 *
395 * Description:	Inline utility function to free an fd in a filedesc
396 *
397 * Parameters:	fdp				Pointer to filedesc fd lies in
398 *		fd				fd to free
399 *		reserv				fd should be reserved
400 *
401 * Returns:	void
402 *
403 * Locks:	Assumes proc_fdlock for process pointing to fdp is held by
404 *		the caller
405 */
406static void
407_fdrelse(struct proc * p, int fd)
408{
409	struct filedesc *fdp = p->p_fd;
410	int nfd = 0;
411
412	if (fd < fdp->fd_freefile)
413		fdp->fd_freefile = fd;
414#if DIAGNOSTIC
415	if (fd > fdp->fd_lastfile)
416 		panic("fdrelse: fd_lastfile inconsistent");
417#endif
418	procfdtbl_clearfd(p, fd);
419
420	while ((nfd = fdp->fd_lastfile) > 0 &&
421			fdp->fd_ofiles[nfd] == NULL &&
422			!(fdp->fd_ofileflags[nfd] & UF_RESERVED))
423		fdp->fd_lastfile--;
424}
425
426
427int
428fd_rdwr(
429	int fd,
430	enum uio_rw rw,
431	uint64_t base,
432	int64_t len,
433	enum uio_seg segflg,
434	off_t	offset,
435	int	io_flg,
436	int64_t *aresid)
437{
438        struct fileproc *fp;
439	proc_t	p;
440        int error = 0;
441	int flags = 0;
442	int spacetype;
443	uio_t auio = NULL;
444	char uio_buf[ UIO_SIZEOF(1) ];
445	struct vfs_context context = *(vfs_context_current());
446	bool wrote_some = false;
447
448	p = current_proc();
449
450        error = fp_lookup(p, fd, &fp, 0);
451        if (error)
452                return(error);
453
454	if (fp->f_type != DTYPE_VNODE && fp->f_type != DTYPE_PIPE && fp->f_type != DTYPE_SOCKET) {
455		error = EINVAL;
456		goto out;
457	}
458	if (rw == UIO_WRITE && !(fp->f_flag & FWRITE)) {
459                error = EBADF;
460		goto out;
461	}
462
463	if (rw == UIO_READ && !(fp->f_flag & FREAD)) {
464    		error = EBADF;
465    		goto out;
466	}
467
468	context.vc_ucred = fp->f_fglob->fg_cred;
469
470	if (UIO_SEG_IS_USER_SPACE(segflg))
471		spacetype = proc_is64bit(p) ? UIO_USERSPACE64 : UIO_USERSPACE32;
472	else
473		spacetype = UIO_SYSSPACE;
474
475	auio = uio_createwithbuffer(1, offset, spacetype, rw, &uio_buf[0], sizeof(uio_buf));
476
477	uio_addiov(auio, base, len);
478
479	if ( !(io_flg & IO_APPEND))
480		flags = FOF_OFFSET;
481
482	if (rw == UIO_WRITE) {
483		user_ssize_t orig_resid = uio_resid(auio);
484		error = fo_write(fp, auio, flags, &context);
485		wrote_some = uio_resid(auio) < orig_resid;
486	} else
487		error = fo_read(fp, auio, flags, &context);
488
489	if (aresid)
490		*aresid = uio_resid(auio);
491	else {
492		if (uio_resid(auio) && error == 0)
493			error = EIO;
494	}
495out:
496        if (wrote_some)
497                fp_drop_written(p, fd, fp);
498        else
499                fp_drop(p, fd, fp, 0);
500
501	return error;
502}
503
504
505
506/*
507 * dup
508 *
509 * Description:	Duplicate a file descriptor.
510 *
511 * Parameters:	p				Process performing the dup
512 *		uap->fd				The fd to dup
513 *		retval				Pointer to the call return area
514 *
515 * Returns:	0				Success
516 *		!0				Errno
517 *
518 * Implicit returns:
519 *		*retval (modified)		The new descriptor
520 */
521int
522dup(proc_t p, struct dup_args *uap, int32_t *retval)
523{
524	struct filedesc *fdp = p->p_fd;
525	int old = uap->fd;
526	int new, error;
527	struct fileproc *fp;
528
529	proc_fdlock(p);
530	if ( (error = fp_lookup(p, old, &fp, 1)) ) {
531		proc_fdunlock(p);
532		return(error);
533	}
534	if (FP_ISGUARDED(fp, GUARD_DUP)) {
535		error = fp_guard_exception(p, old, fp, kGUARD_EXC_DUP);
536		(void) fp_drop(p, old, fp, 1);
537		proc_fdunlock(p);
538		return (error);
539	}
540	if ( (error = fdalloc(p, 0, &new)) ) {
541		fp_drop(p, old, fp, 1);
542		proc_fdunlock(p);
543		return (error);
544	}
545	error = finishdup(p, fdp, old, new, 0, retval);
546	fp_drop(p, old, fp, 1);
547	proc_fdunlock(p);
548
549	return (error);
550}
551
552/*
553 * dup2
554 *
555 * Description:	Duplicate a file descriptor to a particular value.
556 *
557 * Parameters:	p				Process performing the dup
558 *		uap->from			The fd to dup
559 *		uap->to				The fd to dup it to
560 *		retval				Pointer to the call return area
561 *
562 * Returns:	0				Success
563 *		!0				Errno
564 *
565 * Implicit returns:
566 *		*retval (modified)		The new descriptor
567 */
568int
569dup2(proc_t p, struct dup2_args *uap, int32_t *retval)
570{
571	struct filedesc *fdp = p->p_fd;
572	int old = uap->from, new = uap->to;
573	int i, error;
574	struct fileproc *fp, *nfp;
575
576	proc_fdlock(p);
577
578startover:
579	if ( (error = fp_lookup(p, old, &fp, 1)) ) {
580		proc_fdunlock(p);
581		return(error);
582	}
583	if (FP_ISGUARDED(fp, GUARD_DUP)) {
584		error = fp_guard_exception(p, old, fp, kGUARD_EXC_DUP);
585		(void) fp_drop(p, old, fp, 1);
586		proc_fdunlock(p);
587		return (error);
588	}
589	if (new < 0 ||
590		(rlim_t)new >= p->p_rlimit[RLIMIT_NOFILE].rlim_cur ||
591	    new >= maxfiles) {
592		fp_drop(p, old, fp, 1);
593		proc_fdunlock(p);
594		return (EBADF);
595	}
596	if (old == new) {
597		fp_drop(p, old, fp, 1);
598		*retval = new;
599		proc_fdunlock(p);
600		return (0);
601	}
602	if (new < 0 || new >= fdp->fd_nfiles) {
603		if ( (error = fdalloc(p, new, &i)) ) {
604			fp_drop(p, old, fp, 1);
605			proc_fdunlock(p);
606			return (error);
607		}
608		if (new != i) {
609			fdrelse(p, i);
610			goto closeit;
611		}
612	} else {
613closeit:
614		while ((fdp->fd_ofileflags[new] & UF_RESERVED) == UF_RESERVED)  {
615				fp_drop(p, old, fp, 1);
616				procfdtbl_waitfd(p, new);
617#if DIAGNOSTIC
618				proc_fdlock_assert(p, LCK_MTX_ASSERT_OWNED);
619#endif
620				goto startover;
621		}
622
623		if ((fdp->fd_ofiles[new] != NULL) &&
624		    ((error = fp_lookup(p, new, &nfp, 1)) == 0)) {
625			fp_drop(p, old, fp, 1);
626			if (FP_ISGUARDED(nfp, GUARD_CLOSE)) {
627				error = fp_guard_exception(p,
628				    new, nfp, kGUARD_EXC_CLOSE);
629				(void) fp_drop(p, new, nfp, 1);
630				proc_fdunlock(p);
631				return (error);
632			}
633			(void)close_internal_locked(p, new, nfp, FD_DUP2RESV);
634#if DIAGNOSTIC
635			proc_fdlock_assert(p, LCK_MTX_ASSERT_OWNED);
636#endif
637			procfdtbl_clearfd(p, new);
638			goto startover;
639		} else  {
640#if DIAGNOSTIC
641			if (fdp->fd_ofiles[new] != NULL)
642				panic("dup2: no ref on fileproc %d", new);
643#endif
644			procfdtbl_reservefd(p, new);
645		}
646
647#if DIAGNOSTIC
648		proc_fdlock_assert(p, LCK_MTX_ASSERT_OWNED);
649#endif
650
651	}
652#if DIAGNOSTIC
653	if (fdp->fd_ofiles[new] != 0)
654		panic("dup2: overwriting fd_ofiles with new %d", new);
655	if ((fdp->fd_ofileflags[new] & UF_RESERVED) == 0)
656		panic("dup2: unreserved fileflags with new %d", new);
657#endif
658	error = finishdup(p, fdp, old, new, 0, retval);
659	fp_drop(p, old, fp, 1);
660	proc_fdunlock(p);
661
662	return(error);
663}
664
665
666/*
667 * fcntl
668 *
669 * Description:	The file control system call.
670 *
671 * Parameters:	p				Process performing the fcntl
672 *		uap->fd				The fd to operate against
673 *		uap->cmd			The command to perform
674 *		uap->arg			Pointer to the command argument
675 *		retval				Pointer to the call return area
676 *
677 * Returns:	0				Success
678 *		!0				Errno (see fcntl_nocancel)
679 *
680 * Implicit returns:
681 *		*retval (modified)		fcntl return value (if any)
682 *
683 * Notes:	This system call differs from fcntl_nocancel() in that it
684 *		tests for cancellation prior to performing a potentially
685 *		blocking operation.
686 */
687int
688fcntl(proc_t p, struct fcntl_args *uap, int32_t *retval)
689{
690	__pthread_testcancel(1);
691	return(fcntl_nocancel(p, (struct fcntl_nocancel_args *)uap, retval));
692}
693
694
695/*
696 * fcntl_nocancel
697 *
698 * Description:	A non-cancel-testing file control system call.
699 *
700 * Parameters:	p				Process performing the fcntl
701 *		uap->fd				The fd to operate against
702 *		uap->cmd			The command to perform
703 *		uap->arg			Pointer to the command argument
704 *		retval				Pointer to the call return area
705 *
706 * Returns:	0				Success
707 *		EINVAL
708 *	fp_lookup:EBADF				Bad file descriptor
709 * [F_DUPFD]
710 *	fdalloc:EMFILE
711 *	fdalloc:ENOMEM
712 *	finishdup:EBADF
713 *	finishdup:ENOMEM
714 * [F_SETOWN]
715 *		ESRCH
716 * [F_SETLK]
717 *		EBADF
718 *		EOVERFLOW
719 *	copyin:EFAULT
720 *	vnode_getwithref:???
721 *	VNOP_ADVLOCK:???
722 *	msleep:ETIMEDOUT
723 * [F_GETLK]
724 *		EBADF
725 *		EOVERFLOW
726 *	copyin:EFAULT
727 *	copyout:EFAULT
728 *	vnode_getwithref:???
729 *	VNOP_ADVLOCK:???
730 * [F_PREALLOCATE]
731 *		EBADF
732 *		EINVAL
733 *	copyin:EFAULT
734 *	copyout:EFAULT
735 *	vnode_getwithref:???
736 *	VNOP_ALLOCATE:???
737 * [F_SETSIZE,F_RDADVISE]
738 *		EBADF
739 *	copyin:EFAULT
740 *	vnode_getwithref:???
741 * [F_RDAHEAD,F_NOCACHE]
742 *		EBADF
743 *	vnode_getwithref:???
744 * [???]
745 *
746 * Implicit returns:
747 *		*retval (modified)		fcntl return value (if any)
748 */
749int
750fcntl_nocancel(proc_t p, struct fcntl_nocancel_args *uap, int32_t *retval)
751{
752	int fd = uap->fd;
753	struct filedesc *fdp = p->p_fd;
754	struct fileproc *fp;
755	char *pop;
756	struct vnode *vp = NULLVP;	/* for AUDIT_ARG() at end */
757	int i, tmp, error, error2, flg = F_POSIX;
758	struct flock fl;
759	struct flocktimeout fltimeout;
760	struct timespec *timeout = NULL;
761	struct vfs_context context;
762	off_t offset;
763	int newmin;
764	daddr64_t lbn, bn;
765	unsigned int fflag;
766	user_addr_t argp;
767	boolean_t is64bit;
768
769	AUDIT_ARG(fd, uap->fd);
770	AUDIT_ARG(cmd, uap->cmd);
771
772	proc_fdlock(p);
773	if ( (error = fp_lookup(p, fd, &fp, 1)) ) {
774		proc_fdunlock(p);
775		return(error);
776	}
777	context.vc_thread = current_thread();
778	context.vc_ucred = fp->f_cred;
779
780	is64bit = proc_is64bit(p);
781	if (is64bit) {
782		argp = uap->arg;
783	}
784	else {
785		/*
786		 * Since the arg parameter is defined as a long but may be
787		 * either a long or a pointer we must take care to handle
788		 * sign extension issues.  Our sys call munger will sign
789		 * extend a long when we are called from a 32-bit process.
790		 * Since we can never have an address greater than 32-bits
791		 * from a 32-bit process we lop off the top 32-bits to avoid
792		 * getting the wrong address
793		 */
794		argp = CAST_USER_ADDR_T((uint32_t)uap->arg);
795	}
796
797	pop = &fdp->fd_ofileflags[fd];
798
799#if CONFIG_MACF
800	error = mac_file_check_fcntl(proc_ucred(p), fp->f_fglob, uap->cmd,
801	    uap->arg);
802	if (error)
803		goto out;
804#endif
805
806	switch (uap->cmd) {
807
808	case F_DUPFD:
809	case F_DUPFD_CLOEXEC:
810		if (FP_ISGUARDED(fp, GUARD_DUP)) {
811			error = fp_guard_exception(p, fd, fp, kGUARD_EXC_DUP);
812			goto out;
813		}
814		newmin = CAST_DOWN_EXPLICIT(int, uap->arg); /* arg is an int, so we won't lose bits */
815		AUDIT_ARG(value32, newmin);
816		if ((u_int)newmin >= p->p_rlimit[RLIMIT_NOFILE].rlim_cur ||
817		    newmin >= maxfiles) {
818			error = EINVAL;
819			goto out;
820		}
821		if ( (error = fdalloc(p, newmin, &i)) )
822			goto out;
823		error = finishdup(p, fdp, fd, i,
824		    uap->cmd == F_DUPFD_CLOEXEC ? UF_EXCLOSE : 0, retval);
825		goto out;
826
827	case F_GETFD:
828		*retval = (*pop & UF_EXCLOSE)? FD_CLOEXEC : 0;
829		error = 0;
830		goto out;
831
832	case F_SETFD:
833		AUDIT_ARG(value32, uap->arg);
834		if (uap->arg & FD_CLOEXEC)
835			*pop |= UF_EXCLOSE;
836		else {
837			if (FILEPROC_TYPE(fp) == FTYPE_GUARDED) {
838				error = fp_guard_exception(p,
839				    fd, fp, kGUARD_EXC_NOCLOEXEC);
840				goto out;
841			}
842			*pop &= ~UF_EXCLOSE;
843		}
844		error = 0;
845		goto out;
846
847	case F_GETFL:
848		*retval = OFLAGS(fp->f_flag);
849		error = 0;
850		goto out;
851
852	case F_SETFL:
853		fp->f_flag &= ~FCNTLFLAGS;
854		tmp = CAST_DOWN_EXPLICIT(int, uap->arg); /* arg is an int, so we won't lose bits */
855		AUDIT_ARG(value32, tmp);
856		fp->f_flag |= FFLAGS(tmp) & FCNTLFLAGS;
857		tmp = fp->f_flag & FNONBLOCK;
858		error = fo_ioctl(fp, FIONBIO, (caddr_t)&tmp, &context);
859		if (error)
860			goto out;
861		tmp = fp->f_flag & FASYNC;
862		error = fo_ioctl(fp, FIOASYNC, (caddr_t)&tmp, &context);
863		if (!error)
864			goto out;
865		fp->f_flag &= ~FNONBLOCK;
866		tmp = 0;
867		(void)fo_ioctl(fp, FIONBIO, (caddr_t)&tmp, &context);
868		goto out;
869
870	case F_GETOWN:
871		if (fp->f_type == DTYPE_SOCKET) {
872			*retval = ((struct socket *)fp->f_data)->so_pgid;
873			error = 0;
874			goto out;
875		}
876		error = fo_ioctl(fp, (int)TIOCGPGRP, (caddr_t)retval, &context);
877		*retval = -*retval;
878		goto out;
879
880	case F_SETOWN:
881		tmp = CAST_DOWN_EXPLICIT(pid_t, uap->arg); /* arg is an int, so we won't lose bits */
882		AUDIT_ARG(value32, tmp);
883		if (fp->f_type == DTYPE_SOCKET) {
884			((struct socket *)fp->f_data)->so_pgid = tmp;
885			error =0;
886			goto out;
887		}
888		if (fp->f_type == DTYPE_PIPE) {
889			error =  fo_ioctl(fp, TIOCSPGRP, (caddr_t)&tmp, &context);
890			goto out;
891		}
892
893		if (tmp <= 0) {
894			tmp = -tmp;
895		} else {
896			proc_t p1 = proc_find(tmp);
897			if (p1 == 0) {
898				error = ESRCH;
899				goto out;
900			}
901			tmp = (int)p1->p_pgrpid;
902			proc_rele(p1);
903		}
904		error =  fo_ioctl(fp, (int)TIOCSPGRP, (caddr_t)&tmp, &context);
905		goto out;
906
907	case F_SETNOSIGPIPE:
908		tmp = CAST_DOWN_EXPLICIT(int, uap->arg);
909		if (fp->f_type == DTYPE_SOCKET) {
910#if SOCKETS
911			error = sock_setsockopt((struct socket *)fp->f_data,
912			    SOL_SOCKET, SO_NOSIGPIPE, &tmp, sizeof (tmp));
913#else
914			error = EINVAL;
915#endif
916		} else {
917			struct fileglob *fg = fp->f_fglob;
918
919			lck_mtx_lock_spin(&fg->fg_lock);
920			if (tmp)
921				fg->fg_lflags |= FG_NOSIGPIPE;
922			else
923				fg->fg_lflags &= FG_NOSIGPIPE;
924			lck_mtx_unlock(&fg->fg_lock);
925			error = 0;
926		}
927		goto out;
928
929	case F_GETNOSIGPIPE:
930		if (fp->f_type == DTYPE_SOCKET) {
931#if SOCKETS
932			int retsize = sizeof (*retval);
933			error = sock_getsockopt((struct socket *)fp->f_data,
934			    SOL_SOCKET, SO_NOSIGPIPE, retval, &retsize);
935#else
936			error = EINVAL;
937#endif
938		} else {
939			*retval = (fp->f_fglob->fg_lflags & FG_NOSIGPIPE) ?
940				1 : 0;
941			error = 0;
942		}
943		goto out;
944
945	case F_SETLKWTIMEOUT:
946	case F_SETLKW:
947		flg |= F_WAIT;
948		/* Fall into F_SETLK */
949
950	case F_SETLK:
951		if (fp->f_type != DTYPE_VNODE) {
952			error = EBADF;
953			goto out;
954		}
955		vp = (struct vnode *)fp->f_data;
956
957		fflag = fp->f_flag;
958		offset = fp->f_offset;
959		proc_fdunlock(p);
960
961		/* Copy in the lock structure */
962		if (uap->cmd == F_SETLKWTIMEOUT) {
963			error = copyin(argp, (caddr_t) &fltimeout, sizeof(fltimeout));
964			if (error) {
965				goto outdrop;
966			}
967			fl = fltimeout.fl;
968			timeout = &fltimeout.timeout;
969		} else {
970			error = copyin(argp, (caddr_t)&fl, sizeof(fl));
971			if (error) {
972				goto outdrop;
973			}
974		}
975
976		/* Check starting byte and ending byte for EOVERFLOW in SEEK_CUR */
977		/* and ending byte for EOVERFLOW in SEEK_SET */
978		error = check_file_seek_range(&fl, offset);
979		if (error) {
980			goto outdrop;
981		}
982
983		if ( (error = vnode_getwithref(vp)) ) {
984			goto outdrop;
985		}
986		if (fl.l_whence == SEEK_CUR)
987			fl.l_start += offset;
988
989#if CONFIG_MACF
990		error = mac_file_check_lock(proc_ucred(p), fp->f_fglob,
991		    F_SETLK, &fl);
992		if (error) {
993			(void)vnode_put(vp);
994			goto outdrop;
995		}
996#endif
997		switch (fl.l_type) {
998
999		case F_RDLCK:
1000			if ((fflag & FREAD) == 0) {
1001				(void)vnode_put(vp);
1002				error = EBADF;
1003				goto outdrop;
1004			}
1005			// XXX UInt32 unsafe for LP64 kernel
1006			OSBitOrAtomic(P_LADVLOCK, &p->p_ladvflag);
1007			error = VNOP_ADVLOCK(vp, (caddr_t)p, F_SETLK, &fl, flg, &context, timeout);
1008			(void)vnode_put(vp);
1009			goto outdrop;
1010
1011		case F_WRLCK:
1012			if ((fflag & FWRITE) == 0) {
1013				(void)vnode_put(vp);
1014				error = EBADF;
1015				goto outdrop;
1016			}
1017			// XXX UInt32 unsafe for LP64 kernel
1018			OSBitOrAtomic(P_LADVLOCK, &p->p_ladvflag);
1019			error = VNOP_ADVLOCK(vp, (caddr_t)p, F_SETLK, &fl, flg, &context, timeout);
1020			(void)vnode_put(vp);
1021			goto outdrop;
1022
1023		case F_UNLCK:
1024			error = VNOP_ADVLOCK(vp, (caddr_t)p, F_UNLCK, &fl,
1025				F_POSIX, &context, timeout);
1026			(void)vnode_put(vp);
1027			goto outdrop;
1028
1029		default:
1030			(void)vnode_put(vp);
1031			error = EINVAL;
1032			goto outdrop;
1033		}
1034
1035	case F_GETLK:
1036		if (fp->f_type != DTYPE_VNODE) {
1037			error = EBADF;
1038			goto out;
1039		}
1040		vp = (struct vnode *)fp->f_data;
1041
1042		offset = fp->f_offset;
1043		proc_fdunlock(p);
1044
1045		/* Copy in the lock structure */
1046		error = copyin(argp, (caddr_t)&fl, sizeof(fl));
1047		if (error)
1048			goto outdrop;
1049
1050		/* Check starting byte and ending byte for EOVERFLOW in SEEK_CUR */
1051		/* and ending byte for EOVERFLOW in SEEK_SET */
1052		error = check_file_seek_range(&fl, offset);
1053		if (error) {
1054			goto outdrop;
1055		}
1056
1057		if ((fl.l_whence == SEEK_SET) && (fl.l_start < 0)) {
1058			error = EINVAL;
1059			goto outdrop;
1060		}
1061
1062		switch (fl.l_type) {
1063		case F_RDLCK:
1064		case F_UNLCK:
1065		case F_WRLCK:
1066			break;
1067		default:
1068			error = EINVAL;
1069			goto outdrop;
1070		}
1071
1072		switch (fl.l_whence) {
1073		case SEEK_CUR:
1074		case SEEK_SET:
1075		case SEEK_END:
1076			break;
1077		default:
1078			error = EINVAL;
1079			goto outdrop;
1080		}
1081
1082		if ( (error = vnode_getwithref(vp)) == 0 ) {
1083			if (fl.l_whence == SEEK_CUR)
1084			        fl.l_start += offset;
1085
1086#if CONFIG_MACF
1087			error = mac_file_check_lock(proc_ucred(p), fp->f_fglob,
1088			    uap->cmd, &fl);
1089			if (error == 0)
1090#endif
1091			error = VNOP_ADVLOCK(vp, (caddr_t)p, uap->cmd, &fl, F_POSIX, &context, NULL);
1092
1093			(void)vnode_put(vp);
1094
1095			if (error == 0)
1096				error = copyout((caddr_t)&fl, argp, sizeof(fl));
1097		}
1098		goto outdrop;
1099
1100	case F_PREALLOCATE: {
1101		fstore_t alloc_struct;    /* structure for allocate command */
1102		u_int32_t alloc_flags = 0;
1103
1104		if (fp->f_type != DTYPE_VNODE) {
1105			error = EBADF;
1106			goto out;
1107		}
1108
1109		vp = (struct vnode *)fp->f_data;
1110		proc_fdunlock(p);
1111
1112		/* make sure that we have write permission */
1113		if ((fp->f_flag & FWRITE) == 0) {
1114			error = EBADF;
1115			goto outdrop;
1116		}
1117
1118		error = copyin(argp, (caddr_t)&alloc_struct, sizeof(alloc_struct));
1119		if (error)
1120			goto outdrop;
1121
1122		/* now set the space allocated to 0 */
1123		alloc_struct.fst_bytesalloc = 0;
1124
1125		/*
1126		 * Do some simple parameter checking
1127		 */
1128
1129		/* set up the flags */
1130
1131		alloc_flags |= PREALLOCATE;
1132
1133		if (alloc_struct.fst_flags & F_ALLOCATECONTIG)
1134			alloc_flags |= ALLOCATECONTIG;
1135
1136		if (alloc_struct.fst_flags & F_ALLOCATEALL)
1137			alloc_flags |= ALLOCATEALL;
1138
1139		/*
1140		 * Do any position mode specific stuff.  The only
1141		 * position mode  supported now is PEOFPOSMODE
1142		 */
1143
1144		switch (alloc_struct.fst_posmode) {
1145
1146		case F_PEOFPOSMODE:
1147			if (alloc_struct.fst_offset != 0) {
1148				error = EINVAL;
1149				goto outdrop;
1150			}
1151
1152			alloc_flags |= ALLOCATEFROMPEOF;
1153			break;
1154
1155		case F_VOLPOSMODE:
1156			if (alloc_struct.fst_offset <= 0) {
1157				error = EINVAL;
1158				goto outdrop;
1159			}
1160
1161			alloc_flags |= ALLOCATEFROMVOL;
1162			break;
1163
1164		default: {
1165			error = EINVAL;
1166			goto outdrop;
1167			}
1168		}
1169		if ( (error = vnode_getwithref(vp)) == 0 ) {
1170		        /*
1171			 * call allocate to get the space
1172			 */
1173		        error = VNOP_ALLOCATE(vp,alloc_struct.fst_length,alloc_flags,
1174					      &alloc_struct.fst_bytesalloc, alloc_struct.fst_offset,
1175					      &context);
1176			(void)vnode_put(vp);
1177
1178			error2 = copyout((caddr_t)&alloc_struct, argp, sizeof(alloc_struct));
1179
1180			if (error == 0)
1181				error = error2;
1182		}
1183		goto outdrop;
1184
1185		}
1186	case F_SETSIZE:
1187		if (fp->f_type != DTYPE_VNODE) {
1188			error = EBADF;
1189			goto out;
1190		}
1191		vp = (struct vnode *)fp->f_data;
1192		proc_fdunlock(p);
1193
1194		error = copyin(argp, (caddr_t)&offset, sizeof (off_t));
1195		if (error)
1196			goto outdrop;
1197		AUDIT_ARG(value64, offset);
1198
1199		error = vnode_getwithref(vp);
1200		if (error)
1201			goto outdrop;
1202
1203#if CONFIG_MACF
1204		error = mac_vnode_check_truncate(&context,
1205		    fp->f_fglob->fg_cred, vp);
1206		if (error) {
1207			(void)vnode_put(vp);
1208			goto outdrop;
1209		}
1210#endif
1211		/*
1212		 * Make sure that we are root.  Growing a file
1213		 * without zero filling the data is a security hole
1214		 * root would have access anyway so we'll allow it
1215		 */
1216		if (!kauth_cred_issuser(kauth_cred_get())) {
1217			error = EACCES;
1218		} else {
1219			/*
1220			 * set the file size
1221			 */
1222			error = vnode_setsize(vp, offset, IO_NOZEROFILL,
1223			    &context);
1224		}
1225
1226		(void)vnode_put(vp);
1227		goto outdrop;
1228
1229	case F_RDAHEAD:
1230		if (fp->f_type != DTYPE_VNODE) {
1231			error = EBADF;
1232			goto out;
1233		}
1234		if (uap->arg)
1235		        fp->f_fglob->fg_flag &= ~FNORDAHEAD;
1236		else
1237		        fp->f_fglob->fg_flag |= FNORDAHEAD;
1238
1239		goto out;
1240
1241	case F_NOCACHE:
1242		if (fp->f_type != DTYPE_VNODE) {
1243			error = EBADF;
1244			goto out;
1245		}
1246		if (uap->arg)
1247		        fp->f_fglob->fg_flag |= FNOCACHE;
1248		else
1249		        fp->f_fglob->fg_flag &= ~FNOCACHE;
1250
1251		goto out;
1252
1253	case F_NODIRECT:
1254		if (fp->f_type != DTYPE_VNODE) {
1255			error = EBADF;
1256			goto out;
1257		}
1258		if (uap->arg)
1259		        fp->f_fglob->fg_flag |= FNODIRECT;
1260		else
1261		        fp->f_fglob->fg_flag &= ~FNODIRECT;
1262
1263		goto out;
1264
1265	case F_SINGLE_WRITER:
1266		if (fp->f_type != DTYPE_VNODE) {
1267			error = EBADF;
1268			goto out;
1269		}
1270		if (uap->arg)
1271		        fp->f_fglob->fg_flag |= FSINGLE_WRITER;
1272		else
1273		        fp->f_fglob->fg_flag &= ~FSINGLE_WRITER;
1274
1275		goto out;
1276
1277	case F_GLOBAL_NOCACHE:
1278	        if (fp->f_type != DTYPE_VNODE) {
1279		        error = EBADF;
1280			goto out;
1281		}
1282		vp = (struct vnode *)fp->f_data;
1283		proc_fdunlock(p);
1284
1285		if ( (error = vnode_getwithref(vp)) == 0 ) {
1286
1287		        *retval = vnode_isnocache(vp);
1288
1289			if (uap->arg)
1290			        vnode_setnocache(vp);
1291			else
1292			        vnode_clearnocache(vp);
1293
1294			(void)vnode_put(vp);
1295		}
1296		goto outdrop;
1297
1298	case F_CHECK_OPENEVT:
1299	        if (fp->f_type != DTYPE_VNODE) {
1300		        error = EBADF;
1301			goto out;
1302		}
1303		vp = (struct vnode *)fp->f_data;
1304		proc_fdunlock(p);
1305
1306		if ( (error = vnode_getwithref(vp)) == 0 ) {
1307
1308		        *retval = vnode_is_openevt(vp);
1309
1310			if (uap->arg)
1311			        vnode_set_openevt(vp);
1312			else
1313			        vnode_clear_openevt(vp);
1314
1315			(void)vnode_put(vp);
1316		}
1317		goto outdrop;
1318
1319	case F_RDADVISE: {
1320		struct radvisory ra_struct;
1321
1322		if (fp->f_type != DTYPE_VNODE) {
1323			error = EBADF;
1324			goto out;
1325		}
1326		vp = (struct vnode *)fp->f_data;
1327		proc_fdunlock(p);
1328
1329		if ( (error = copyin(argp, (caddr_t)&ra_struct, sizeof(ra_struct))) )
1330			goto outdrop;
1331		if ( (error = vnode_getwithref(vp)) == 0 ) {
1332		        error = VNOP_IOCTL(vp, F_RDADVISE, (caddr_t)&ra_struct, 0, &context);
1333
1334			(void)vnode_put(vp);
1335		}
1336		goto outdrop;
1337		}
1338
1339        case F_FLUSH_DATA:
1340
1341                if (fp->f_type != DTYPE_VNODE) {
1342                        error = EBADF;
1343                        goto out;
1344                }
1345                vp = (struct vnode *)fp->f_data;
1346                proc_fdunlock(p);
1347
1348                if ( (error = vnode_getwithref(vp)) == 0 ) {
1349                        error = cluster_push(vp, 0);
1350
1351                        (void)vnode_put(vp);
1352                }
1353                goto outdrop;
1354
1355	case F_LOG2PHYS:
1356	case F_LOG2PHYS_EXT: {
1357		struct log2phys l2p_struct;    /* structure for allocate command */
1358		int devBlockSize;
1359
1360		off_t file_offset = 0;
1361		size_t a_size = 0;
1362		size_t run = 0;
1363
1364		if (uap->cmd == F_LOG2PHYS_EXT) {
1365			error = copyin(argp, (caddr_t)&l2p_struct, sizeof(l2p_struct));
1366			if (error)
1367				goto out;
1368			file_offset = l2p_struct.l2p_devoffset;
1369		} else {
1370			file_offset = fp->f_offset;
1371		}
1372		if (fp->f_type != DTYPE_VNODE) {
1373			error = EBADF;
1374			goto out;
1375		}
1376		vp = (struct vnode *)fp->f_data;
1377		proc_fdunlock(p);
1378		if ( (error = vnode_getwithref(vp)) ) {
1379			goto outdrop;
1380		}
1381		error = VNOP_OFFTOBLK(vp, file_offset, &lbn);
1382		if (error) {
1383			(void)vnode_put(vp);
1384			goto outdrop;
1385		}
1386		error = VNOP_BLKTOOFF(vp, lbn, &offset);
1387		if (error) {
1388			(void)vnode_put(vp);
1389			goto outdrop;
1390		}
1391		devBlockSize = vfs_devblocksize(vnode_mount(vp));
1392		if (uap->cmd == F_LOG2PHYS_EXT) {
1393#if defined(__LP64__)
1394			a_size = l2p_struct.l2p_contigbytes;
1395#else
1396			if ((l2p_struct.l2p_contigbytes > SIZE_MAX) || (l2p_struct.l2p_contigbytes < 0)) {
1397				/* size_t is 32-bit on a 32-bit kernel, therefore
1398				 * assigning l2p_contigbytes to a_size may have
1399				 * caused integer overflow.  We, therefore, return
1400				 * an error here instead of calculating incorrect
1401				 * value.
1402				 */
1403				printf ("fcntl: F_LOG2PHYS_EXT: l2p_contigbytes=%lld will overflow, returning error\n", l2p_struct.l2p_contigbytes);
1404				error = EFBIG;
1405				goto outdrop;
1406			} else {
1407				a_size = l2p_struct.l2p_contigbytes;
1408			}
1409#endif
1410		} else {
1411			a_size = devBlockSize;
1412		}
1413
1414		error = VNOP_BLOCKMAP(vp, offset, a_size, &bn, &run, NULL, 0, &context);
1415
1416		(void)vnode_put(vp);
1417
1418		if (!error) {
1419			l2p_struct.l2p_flags = 0;	/* for now */
1420			if (uap->cmd == F_LOG2PHYS_EXT) {
1421				l2p_struct.l2p_contigbytes = run - (file_offset - offset);
1422			} else {
1423				l2p_struct.l2p_contigbytes = 0;	/* for now */
1424			}
1425
1426			/*
1427			 * The block number being -1 suggests that the file offset is not backed
1428			 * by any real blocks on-disk.  As a result, just let it be passed back up wholesale.
1429			 */
1430			if (bn == -1) {
1431				/* Don't multiply it by the block size */
1432				l2p_struct.l2p_devoffset = bn;
1433			}
1434			else {
1435				l2p_struct.l2p_devoffset = bn * devBlockSize;
1436				l2p_struct.l2p_devoffset += file_offset - offset;
1437			}
1438			error = copyout((caddr_t)&l2p_struct, argp, sizeof(l2p_struct));
1439		}
1440		goto outdrop;
1441		}
1442	case F_GETPATH: {
1443		char *pathbufp;
1444		int pathlen;
1445
1446		if (fp->f_type != DTYPE_VNODE) {
1447			error = EBADF;
1448			goto out;
1449		}
1450		vp = (struct vnode *)fp->f_data;
1451		proc_fdunlock(p);
1452
1453		pathlen = MAXPATHLEN;
1454		MALLOC(pathbufp, char *, pathlen, M_TEMP, M_WAITOK);
1455		if (pathbufp == NULL) {
1456			error = ENOMEM;
1457			goto outdrop;
1458		}
1459		if ( (error = vnode_getwithref(vp)) == 0 ) {
1460		        error = vn_getpath(vp, pathbufp, &pathlen);
1461		        (void)vnode_put(vp);
1462
1463			if (error == 0)
1464			        error = copyout((caddr_t)pathbufp, argp, pathlen);
1465		}
1466		FREE(pathbufp, M_TEMP);
1467		goto outdrop;
1468	}
1469
1470	case F_PATHPKG_CHECK: {
1471		char *pathbufp;
1472		size_t pathlen;
1473
1474		if (fp->f_type != DTYPE_VNODE) {
1475		        error = EBADF;
1476			goto out;
1477		}
1478		vp = (struct vnode *)fp->f_data;
1479		proc_fdunlock(p);
1480
1481		pathlen = MAXPATHLEN;
1482		pathbufp = kalloc(MAXPATHLEN);
1483
1484		if ( (error = copyinstr(argp, pathbufp, MAXPATHLEN, &pathlen)) == 0 ) {
1485		        if ( (error = vnode_getwithref(vp)) == 0 ) {
1486				AUDIT_ARG(text, pathbufp);
1487			        error = vn_path_package_check(vp, pathbufp, pathlen, retval);
1488
1489				(void)vnode_put(vp);
1490			}
1491		}
1492		kfree(pathbufp, MAXPATHLEN);
1493		goto outdrop;
1494	}
1495
1496	case F_CHKCLEAN:   // used by regression tests to see if all dirty pages got cleaned by fsync()
1497	case F_FULLFSYNC:  // fsync + flush the journal + DKIOCSYNCHRONIZECACHE
1498	case F_FREEZE_FS:  // freeze all other fs operations for the fs of this fd
1499	case F_THAW_FS: {  // thaw all frozen fs operations for the fs of this fd
1500		if (fp->f_type != DTYPE_VNODE) {
1501			error = EBADF;
1502			goto out;
1503		}
1504		vp = (struct vnode *)fp->f_data;
1505		proc_fdunlock(p);
1506
1507		if ( (error = vnode_getwithref(vp)) == 0 ) {
1508		        error = VNOP_IOCTL(vp, uap->cmd, (caddr_t)NULL, 0, &context);
1509
1510			(void)vnode_put(vp);
1511		}
1512		break;
1513	}
1514
1515	/*
1516	 * SPI (private) for opening a file starting from a dir fd
1517	 */
1518	case F_OPENFROM: {
1519		struct user_fopenfrom fopen;
1520		struct vnode_attr va;
1521		struct nameidata nd;
1522		int cmode;
1523
1524		/* Check if this isn't a valid file descriptor */
1525		if ((fp->f_type != DTYPE_VNODE) ||
1526		    (fp->f_flag & FREAD) == 0) {
1527			error = EBADF;
1528			goto out;
1529		}
1530		vp = (struct vnode *)fp->f_data;
1531		proc_fdunlock(p);
1532
1533		if (vnode_getwithref(vp)) {
1534			error = ENOENT;
1535			goto outdrop;
1536		}
1537
1538		/* Only valid for directories */
1539		if (vp->v_type != VDIR) {
1540			vnode_put(vp);
1541			error = ENOTDIR;
1542			goto outdrop;
1543		}
1544
1545		/* Get flags, mode and pathname arguments. */
1546		if (IS_64BIT_PROCESS(p)) {
1547			error = copyin(argp, &fopen, sizeof(fopen));
1548		} else {
1549			struct user32_fopenfrom fopen32;
1550
1551			error = copyin(argp, &fopen32, sizeof(fopen32));
1552			fopen.o_flags = fopen32.o_flags;
1553			fopen.o_mode = fopen32.o_mode;
1554			fopen.o_pathname = CAST_USER_ADDR_T(fopen32.o_pathname);
1555		}
1556		if (error) {
1557			vnode_put(vp);
1558			goto outdrop;
1559		}
1560		AUDIT_ARG(fflags, fopen.o_flags);
1561		AUDIT_ARG(mode, fopen.o_mode);
1562		VATTR_INIT(&va);
1563		/* Mask off all but regular access permissions */
1564		cmode = ((fopen.o_mode &~ fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT;
1565		VATTR_SET(&va, va_mode, cmode & ACCESSPERMS);
1566
1567		/* Start the lookup relative to the file descriptor's vnode. */
1568		NDINIT(&nd, LOOKUP, OP_OPEN, USEDVP | FOLLOW | AUDITVNPATH1, UIO_USERSPACE,
1569		       fopen.o_pathname, &context);
1570		nd.ni_dvp = vp;
1571
1572		error = open1(&context, &nd, fopen.o_flags, &va,
1573			      fileproc_alloc_init, NULL, retval);
1574
1575		vnode_put(vp);
1576		break;
1577	}
1578	/*
1579	 * SPI (private) for unlinking a file starting from a dir fd
1580	 */
1581	case F_UNLINKFROM: {
1582		struct nameidata nd;
1583		user_addr_t pathname;
1584
1585		/* Check if this isn't a valid file descriptor */
1586		if ((fp->f_type != DTYPE_VNODE) ||
1587		    (fp->f_flag & FREAD) == 0) {
1588			error = EBADF;
1589			goto out;
1590		}
1591		vp = (struct vnode *)fp->f_data;
1592		proc_fdunlock(p);
1593
1594		if (vnode_getwithref(vp)) {
1595			error = ENOENT;
1596			goto outdrop;
1597		}
1598
1599		/* Only valid for directories */
1600		if (vp->v_type != VDIR) {
1601			vnode_put(vp);
1602			error = ENOTDIR;
1603			goto outdrop;
1604		}
1605
1606		/* Get flags, mode and pathname arguments. */
1607		if (IS_64BIT_PROCESS(p)) {
1608			pathname = (user_addr_t)argp;
1609		} else {
1610			pathname = CAST_USER_ADDR_T(argp);
1611		}
1612
1613		/* Start the lookup relative to the file descriptor's vnode. */
1614		NDINIT(&nd, DELETE, OP_UNLINK, USEDVP | AUDITVNPATH1, UIO_USERSPACE,
1615		       pathname, &context);
1616		nd.ni_dvp = vp;
1617
1618		error = unlink1(&context, &nd, 0);
1619
1620		vnode_put(vp);
1621		break;
1622
1623	}
1624
1625	case F_ADDSIGS:
1626	case F_ADDFILESIGS:
1627	{
1628		struct user_fsignatures fs;
1629		kern_return_t kr;
1630		vm_offset_t kernel_blob_addr;
1631		vm_size_t kernel_blob_size;
1632
1633		if (fp->f_type != DTYPE_VNODE) {
1634			error = EBADF;
1635			goto out;
1636		}
1637		vp = (struct vnode *)fp->f_data;
1638		proc_fdunlock(p);
1639		error = vnode_getwithref(vp);
1640		if (error)
1641			goto outdrop;
1642
1643		if (IS_64BIT_PROCESS(p)) {
1644			error = copyin(argp, &fs, sizeof (fs));
1645		} else {
1646			struct user32_fsignatures fs32;
1647
1648			error = copyin(argp, &fs32, sizeof (fs32));
1649			fs.fs_file_start = fs32.fs_file_start;
1650			fs.fs_blob_start = CAST_USER_ADDR_T(fs32.fs_blob_start);
1651			fs.fs_blob_size = fs32.fs_blob_size;
1652		}
1653
1654		if (error) {
1655			vnode_put(vp);
1656			goto outdrop;
1657		}
1658
1659		if(ubc_cs_blob_get(vp, CPU_TYPE_ANY, fs.fs_file_start))
1660		{
1661			vnode_put(vp);
1662			goto outdrop;
1663		}
1664/*
1665 * An arbitrary limit, to prevent someone from mapping in a 20GB blob.  This should cover
1666 * our use cases for the immediate future, but note that at the time of this commit, some
1667 * platforms are nearing 2MB blob sizes (with a prior soft limit of 2.5MB).
1668 *
1669 * We should consider how we can manage this more effectively; the above means that some
1670 * platforms are using megabytes of memory for signing data; it merely hasn't crossed the
1671 * threshold considered ridiculous at the time of this change.
1672 */
1673#define CS_MAX_BLOB_SIZE (10ULL * 1024ULL * 1024ULL)
1674		if (fs.fs_blob_size > CS_MAX_BLOB_SIZE) {
1675			error = E2BIG;
1676			vnode_put(vp);
1677			goto outdrop;
1678		}
1679
1680		kernel_blob_size = CAST_DOWN(vm_size_t, fs.fs_blob_size);
1681		kr = ubc_cs_blob_allocate(&kernel_blob_addr, &kernel_blob_size);
1682		if (kr != KERN_SUCCESS) {
1683			error = ENOMEM;
1684			vnode_put(vp);
1685			goto outdrop;
1686		}
1687
1688		if(uap->cmd == F_ADDSIGS) {
1689			error = copyin(fs.fs_blob_start,
1690				       (void *) kernel_blob_addr,
1691				       kernel_blob_size);
1692		} else /* F_ADDFILESIGS */ {
1693			error = vn_rdwr(UIO_READ,
1694					vp,
1695					(caddr_t) kernel_blob_addr,
1696					kernel_blob_size,
1697					 fs.fs_file_start + fs.fs_blob_start,
1698					UIO_SYSSPACE,
1699					0,
1700					kauth_cred_get(),
1701					0,
1702					p);
1703		}
1704
1705		if (error) {
1706			ubc_cs_blob_deallocate(kernel_blob_addr,
1707					       kernel_blob_size);
1708			vnode_put(vp);
1709			goto outdrop;
1710		}
1711
1712		error = ubc_cs_blob_add(
1713			vp,
1714			CPU_TYPE_ANY,	/* not for a specific architecture */
1715			fs.fs_file_start,
1716			kernel_blob_addr,
1717			kernel_blob_size);
1718		if (error) {
1719			ubc_cs_blob_deallocate(kernel_blob_addr,
1720					       kernel_blob_size);
1721		} else {
1722			/* ubc_blob_add() has consumed "kernel_blob_addr" */
1723#if CHECK_CS_VALIDATION_BITMAP
1724			ubc_cs_validation_bitmap_allocate( vp );
1725#endif
1726		}
1727
1728		(void) vnode_put(vp);
1729		break;
1730	}
1731	case F_FINDSIGS: {
1732#ifdef SECURE_KERNEL
1733		error = ENOTSUP;
1734#else /* !SECURE_KERNEL */
1735		off_t offsetMacho;
1736
1737		if (fp->f_type != DTYPE_VNODE) {
1738			error = EBADF;
1739			goto out;
1740		}
1741		vp = (struct vnode *)fp->f_data;
1742		proc_fdunlock(p);
1743		error = vnode_getwithref(vp);
1744		if (error)
1745			goto outdrop;
1746
1747		error = copyin(argp, &offsetMacho, sizeof(offsetMacho));
1748		if (error) {
1749			(void)vnode_put(vp);
1750			goto outdrop;
1751		}
1752
1753#if CONFIG_MACF
1754		error = mac_vnode_find_sigs(p, vp, offsetMacho);
1755#else
1756		error = EPERM;
1757#endif
1758		if (error) {
1759			(void)vnode_put(vp);
1760			goto outdrop;
1761		}
1762#endif /* SECURE_KERNEL */
1763		break;
1764	}
1765#if CONFIG_PROTECT
1766	case F_GETPROTECTIONCLASS: {
1767		int class = 0;
1768
1769		if (fp->f_type != DTYPE_VNODE) {
1770			error = EBADF;
1771			goto out;
1772		}
1773		vp = (struct vnode *)fp->f_data;
1774
1775		proc_fdunlock(p);
1776
1777		if (vnode_getwithref(vp)) {
1778			error = ENOENT;
1779			goto outdrop;
1780		}
1781
1782		error = cp_vnode_getclass (vp, &class);
1783		if (error == 0) {
1784			*retval = class;
1785		}
1786
1787		vnode_put(vp);
1788		break;
1789	}
1790
1791	case F_SETPROTECTIONCLASS: {
1792		/* tmp must be a valid PROTECTION_CLASS_* */
1793		tmp = CAST_DOWN_EXPLICIT(uint32_t, uap->arg);
1794
1795		if (fp->f_type != DTYPE_VNODE) {
1796			error = EBADF;
1797			goto out;
1798		}
1799		vp = (struct vnode *)fp->f_data;
1800
1801		proc_fdunlock(p);
1802
1803		if (vnode_getwithref(vp)) {
1804			error = ENOENT;
1805			goto outdrop;
1806		}
1807
1808		/* Only go forward if you have write access */
1809		vfs_context_t ctx = vfs_context_current();
1810		if(vnode_authorize(vp, NULLVP, (KAUTH_VNODE_ACCESS | KAUTH_VNODE_WRITE_DATA), ctx) != 0) {
1811			vnode_put(vp);
1812			error = EBADF;
1813			goto outdrop;
1814		}
1815		error = cp_vnode_setclass (vp, tmp);
1816		vnode_put(vp);
1817		break;
1818	}
1819
1820	case F_TRANSCODEKEY: {
1821
1822		char *backup_keyp = NULL;
1823		unsigned backup_key_len = CP_MAX_WRAPPEDKEYSIZE;
1824
1825		if (fp->f_type != DTYPE_VNODE) {
1826			error = EBADF;
1827			goto out;
1828		}
1829
1830		vp = (struct vnode *)fp->f_data;
1831		proc_fdunlock(p);
1832
1833		if (vnode_getwithref(vp)) {
1834			error = ENOENT;
1835			goto outdrop;
1836		}
1837
1838		MALLOC(backup_keyp, char *, backup_key_len, M_TEMP, M_WAITOK);
1839		if (backup_keyp == NULL) {
1840			error = ENOMEM;
1841			goto outdrop;
1842		}
1843
1844		error = cp_vnode_transcode (vp, backup_keyp, &backup_key_len);
1845		vnode_put(vp);
1846
1847		if (error == 0) {
1848			error = copyout((caddr_t)backup_keyp, argp, backup_key_len);
1849			*retval = backup_key_len;
1850		}
1851
1852		FREE(backup_keyp, M_TEMP);
1853
1854		break;
1855	}
1856
1857	case F_GETPROTECTIONLEVEL:  {
1858		uint32_t cp_version = 0;
1859
1860		if (fp->f_type != DTYPE_VNODE) {
1861			error = EBADF;
1862			goto out;
1863		}
1864
1865		vp = (struct vnode*) fp->f_data;
1866		proc_fdunlock (p);
1867
1868		if (vnode_getwithref(vp)) {
1869			error = ENOENT;
1870			goto outdrop;
1871		}
1872
1873		/*
1874		 * if cp_get_major_vers fails, error will be set to proper errno
1875		 * and cp_version will still be 0.
1876		 */
1877
1878		error = cp_get_root_major_vers (vp, &cp_version);
1879		*retval = cp_version;
1880
1881		vnode_put (vp);
1882		break;
1883	}
1884
1885	case F_GETDEFAULTPROTLEVEL:  {
1886		uint32_t cp_default = 0;
1887
1888		if (fp->f_type != DTYPE_VNODE) {
1889			error = EBADF;
1890			goto out;
1891		}
1892
1893		vp = (struct vnode*) fp->f_data;
1894		proc_fdunlock (p);
1895
1896		if (vnode_getwithref(vp)) {
1897			error = ENOENT;
1898			goto outdrop;
1899		}
1900
1901		/*
1902		 * if cp_get_major_vers fails, error will be set to proper errno
1903		 * and cp_version will still be 0.
1904		 */
1905
1906		error = cp_get_default_level(vp, &cp_default);
1907		*retval = cp_default;
1908
1909		vnode_put (vp);
1910		break;
1911	}
1912
1913
1914#endif /* CONFIG_PROTECT */
1915
1916	case F_MOVEDATAEXTENTS: {
1917		struct fileproc *fp2 = NULL;
1918		struct vnode *src_vp = NULLVP;
1919		struct vnode *dst_vp = NULLVP;
1920		/* We need to grab the 2nd FD out of the argments before moving on. */
1921		int fd2 = CAST_DOWN_EXPLICIT(int32_t, uap->arg);
1922
1923		if (fp->f_type != DTYPE_VNODE) {
1924			error = EBADF;
1925			goto out;
1926		}
1927
1928		/* For now, special case HFS+ only, since this is SPI. */
1929		src_vp = (struct vnode *)fp->f_data;
1930		if (src_vp->v_tag != VT_HFS) {
1931			error = EINVAL;
1932			goto out;
1933		}
1934
1935		/*
1936		 * Get the references before we start acquiring iocounts on the vnodes,
1937		 * while we still hold the proc fd lock
1938		 */
1939		if ( (error = fp_lookup(p, fd2, &fp2, 1)) ) {
1940			error = EBADF;
1941			goto out;
1942		}
1943		if (fp2->f_type != DTYPE_VNODE) {
1944			fp_drop(p, fd2, fp2, 1);
1945			error = EBADF;
1946			goto out;
1947		}
1948		dst_vp = (struct vnode *)fp2->f_data;
1949		if (dst_vp->v_tag != VT_HFS) {
1950			fp_drop(p, fd2, fp2, 1);
1951			error = EINVAL;
1952			goto out;
1953		}
1954
1955#if CONFIG_MACF
1956		/* Re-do MAC checks against the new FD, pass in a fake argument */
1957		error = mac_file_check_fcntl(proc_ucred(p), fp2->f_fglob, uap->cmd, 0);
1958		if (error) {
1959			fp_drop(p, fd2, fp2, 1);
1960			goto out;
1961		}
1962#endif
1963		/* Audit the 2nd FD */
1964		AUDIT_ARG(fd, fd2);
1965
1966		proc_fdunlock(p);
1967
1968		if (vnode_getwithref(src_vp)) {
1969			fp_drop(p, fd2, fp2, 0);
1970			error = ENOENT;
1971			goto outdrop;
1972		}
1973		if (vnode_getwithref(dst_vp)) {
1974			vnode_put (src_vp);
1975			fp_drop(p, fd2, fp2, 0);
1976			error = ENOENT;
1977			goto outdrop;
1978		}
1979
1980		/*
1981		 * Basic asserts; validate they are not the same and that
1982		 * both live on the same filesystem.
1983		 */
1984		if (dst_vp == src_vp) {
1985			vnode_put (src_vp);
1986			vnode_put (dst_vp);
1987			fp_drop (p, fd2, fp2, 0);
1988			error = EINVAL;
1989			goto outdrop;
1990		}
1991
1992		if (dst_vp->v_mount != src_vp->v_mount) {
1993			vnode_put (src_vp);
1994			vnode_put (dst_vp);
1995			fp_drop (p, fd2, fp2, 0);
1996			error = EXDEV;
1997			goto outdrop;
1998		}
1999
2000		/* Now we have a legit pair of FDs.  Go to work */
2001
2002		/* Now check for write access to the target files */
2003		if(vnode_authorize(src_vp, NULLVP,
2004						   (KAUTH_VNODE_ACCESS | KAUTH_VNODE_WRITE_DATA), &context) != 0) {
2005			vnode_put(src_vp);
2006			vnode_put(dst_vp);
2007			fp_drop(p, fd2, fp2, 0);
2008			error = EBADF;
2009			goto outdrop;
2010		}
2011
2012		if(vnode_authorize(dst_vp, NULLVP,
2013						   (KAUTH_VNODE_ACCESS | KAUTH_VNODE_WRITE_DATA), &context) != 0) {
2014			vnode_put(src_vp);
2015			vnode_put(dst_vp);
2016			fp_drop(p, fd2, fp2, 0);
2017			error = EBADF;
2018			goto outdrop;
2019		}
2020
2021		/* Verify that both vps point to files and not directories */
2022		if ( !vnode_isreg(src_vp) || !vnode_isreg(dst_vp)) {
2023			error = EINVAL;
2024			vnode_put (src_vp);
2025			vnode_put (dst_vp);
2026			fp_drop (p, fd2, fp2, 0);
2027			goto outdrop;
2028		}
2029
2030		/*
2031		 * The exchangedata syscall handler passes in 0 for the flags to VNOP_EXCHANGE.
2032		 * We'll pass in our special bit indicating that the new behavior is expected
2033		 */
2034
2035		error = VNOP_EXCHANGE(src_vp, dst_vp, FSOPT_EXCHANGE_DATA_ONLY, &context);
2036
2037		vnode_put (src_vp);
2038		vnode_put (dst_vp);
2039		fp_drop(p, fd2, fp2, 0);
2040		break;
2041	}
2042
2043	/*
2044	 * SPI for making a file compressed.
2045	 */
2046	case F_MAKECOMPRESSED: {
2047		uint32_t gcounter = CAST_DOWN_EXPLICIT(uint32_t, uap->arg);
2048
2049		if (fp->f_type != DTYPE_VNODE) {
2050			error = EBADF;
2051			goto out;
2052		}
2053
2054		vp = (struct vnode*) fp->f_data;
2055		proc_fdunlock (p);
2056
2057		/* get the vnode */
2058		if (vnode_getwithref(vp)) {
2059			error = ENOENT;
2060			goto outdrop;
2061		}
2062
2063		/* Is it a file? */
2064		if ((vnode_isreg(vp) == 0) && (vnode_islnk(vp) == 0)) {
2065			vnode_put(vp);
2066			error = EBADF;
2067			goto outdrop;
2068		}
2069
2070		/* invoke ioctl to pass off to FS */
2071		/* Only go forward if you have write access */
2072		vfs_context_t ctx = vfs_context_current();
2073		if(vnode_authorize(vp, NULLVP, (KAUTH_VNODE_ACCESS | KAUTH_VNODE_WRITE_DATA), ctx) != 0) {
2074			vnode_put(vp);
2075			error = EBADF;
2076			goto outdrop;
2077		}
2078
2079		error = VNOP_IOCTL(vp, uap->cmd, (caddr_t)&gcounter, 0, &context);
2080
2081		vnode_put (vp);
2082		break;
2083	}
2084
2085	/*
2086	 * SPI (private) for indicating to a filesystem that subsequent writes to
2087	 * the open FD will written to the Fastflow.
2088	 */
2089	case F_SET_GREEDY_MODE:
2090		/* intentionally drop through to the same handler as F_SETSTATIC.
2091		 * both fcntls should pass the argument and their selector into VNOP_IOCTL.
2092		 */
2093
2094	/*
2095	 * SPI (private) for indicating to a filesystem that subsequent writes to
2096	 * the open FD will represent static content.
2097	 */
2098	case F_SETSTATICCONTENT: {
2099		caddr_t ioctl_arg = NULL;
2100
2101		if (uap->arg) {
2102			ioctl_arg = (caddr_t) 1;
2103		}
2104
2105		if (fp->f_type != DTYPE_VNODE) {
2106			error = EBADF;
2107			goto out;
2108		}
2109		vp = (struct vnode *)fp->f_data;
2110		proc_fdunlock(p);
2111
2112		error = vnode_getwithref(vp);
2113		if (error) {
2114			error = ENOENT;
2115			goto outdrop;
2116		}
2117
2118		/* Only go forward if you have write access */
2119		vfs_context_t ctx = vfs_context_current();
2120		if(vnode_authorize(vp, NULLVP, (KAUTH_VNODE_ACCESS | KAUTH_VNODE_WRITE_DATA), ctx) != 0) {
2121			vnode_put(vp);
2122			error = EBADF;
2123			goto outdrop;
2124		}
2125
2126		error = VNOP_IOCTL(vp, uap->cmd, ioctl_arg, 0, &context);
2127		(void)vnode_put(vp);
2128
2129		break;
2130	}
2131
2132	/*
2133	 * SPI (private) for indicating to the lower level storage driver that the
2134	 * subsequent writes should be of a particular IO type (burst, greedy, static),
2135	 * or other flavors that may be necessary.
2136	 */
2137	case F_SETIOTYPE: {
2138		caddr_t param_ptr;
2139		uint32_t param;
2140
2141		if (uap->arg) {
2142			/* extract 32 bits of flags from userland */
2143			param_ptr = (caddr_t) uap->arg;
2144			param = (uint32_t) param_ptr;
2145		}
2146		else {
2147			/* If no argument is specified, error out */
2148			error = EINVAL;
2149			goto out;
2150		}
2151
2152		/*
2153		 * Validate the different types of flags that can be specified:
2154		 * all of them are mutually exclusive for now.
2155		 */
2156		switch (param) {
2157			case F_IOTYPE_ISOCHRONOUS:
2158				break;
2159
2160			default:
2161				error = EINVAL;
2162				goto out;
2163		}
2164
2165
2166		if (fp->f_type != DTYPE_VNODE) {
2167			error = EBADF;
2168			goto out;
2169		}
2170		vp = (struct vnode *)fp->f_data;
2171		proc_fdunlock(p);
2172
2173		error = vnode_getwithref(vp);
2174		if (error) {
2175			error = ENOENT;
2176			goto outdrop;
2177		}
2178
2179		/* Only go forward if you have write access */
2180		vfs_context_t ctx = vfs_context_current();
2181		if(vnode_authorize(vp, NULLVP, (KAUTH_VNODE_ACCESS | KAUTH_VNODE_WRITE_DATA), ctx) != 0) {
2182			vnode_put(vp);
2183			error = EBADF;
2184			goto outdrop;
2185		}
2186
2187		error = VNOP_IOCTL(vp, uap->cmd, param_ptr, 0, &context);
2188		(void)vnode_put(vp);
2189
2190		break;
2191	}
2192
2193
2194	/*
2195	 * Extract the CodeDirectory of the vnode associated with
2196	 * the file descriptor and copy it back to user space
2197	 */
2198	case F_GETCODEDIR: {
2199		struct user_fcodeblobs args;
2200
2201		if (fp->f_type != DTYPE_VNODE) {
2202			error = EBADF;
2203			goto out;
2204		}
2205
2206		vp = (struct vnode *)fp->f_data;
2207		proc_fdunlock(p);
2208
2209		if ((fp->f_flag & FREAD) == 0) {
2210			error = EBADF;
2211			goto outdrop;
2212		}
2213
2214		if (IS_64BIT_PROCESS(p)) {
2215			struct user64_fcodeblobs args64;
2216
2217			error = copyin(argp, &args64, sizeof(args64));
2218			if (error)
2219				goto outdrop;
2220
2221			args.f_cd_hash = args64.f_cd_hash;
2222			args.f_hash_size = args64.f_hash_size;
2223			args.f_cd_buffer = args64.f_cd_buffer;
2224			args.f_cd_size = args64.f_cd_size;
2225			args.f_out_size = args64.f_out_size;
2226			args.f_arch = args64.f_arch;
2227		} else {
2228			struct user32_fcodeblobs args32;
2229
2230			error = copyin(argp, &args32, sizeof(args32));
2231			if (error)
2232				goto outdrop;
2233
2234			args.f_cd_hash = CAST_USER_ADDR_T(args32.f_cd_hash);
2235			args.f_hash_size = args32.f_hash_size;
2236			args.f_cd_buffer = CAST_USER_ADDR_T(args32.f_cd_buffer);
2237			args.f_cd_size = args32.f_cd_size;
2238			args.f_out_size = CAST_USER_ADDR_T(args32.f_out_size);
2239			args.f_arch = args32.f_arch;
2240		}
2241
2242		if (vp->v_ubcinfo == NULL) {
2243			error = EINVAL;
2244			goto outdrop;
2245		}
2246
2247		struct cs_blob *t_blob = vp->v_ubcinfo->cs_blobs;
2248
2249		/*
2250		 * This call fails if there is no cs_blob corresponding to the
2251		 * vnode, or if there are multiple cs_blobs present, and the caller
2252		 * did not specify which cpu_type they want the cs_blob for
2253		 */
2254		if (t_blob == NULL) {
2255			error = ENOENT; /* there is no codesigning blob for this process */
2256			goto outdrop;
2257		} else if (args.f_arch == 0 && t_blob->csb_next != NULL) {
2258			error = ENOENT; /* too many architectures and none specified */
2259			goto outdrop;
2260		}
2261
2262		/* If the user specified an architecture, find the right blob */
2263		if (args.f_arch != 0) {
2264			while (t_blob) {
2265				if (t_blob->csb_cpu_type == args.f_arch)
2266					break;
2267				t_blob = t_blob->csb_next;
2268			}
2269			/* The cpu_type the user requested could not be found */
2270			if (t_blob == NULL) {
2271				error = ENOENT;
2272				goto outdrop;
2273			}
2274		}
2275
2276		const CS_SuperBlob *super_blob = (void *)t_blob->csb_mem_kaddr;
2277		const CS_CodeDirectory *cd = findCodeDirectory(super_blob,
2278	 						 (char *) super_blob,
2279							 (char *) super_blob + t_blob->csb_mem_size);
2280		if (cd == NULL) {
2281			error = ENOENT;
2282			goto outdrop;
2283		}
2284
2285		uint64_t buffer_size = ntohl(cd->length);
2286
2287		if (buffer_size > UINT_MAX) {
2288			error = ERANGE;
2289			goto outdrop;
2290		}
2291
2292		error = copyout(&buffer_size, args.f_out_size, sizeof(unsigned int));
2293		if (error)
2294			goto outdrop;
2295
2296		if (sizeof(t_blob->csb_sha1) > args.f_hash_size ||
2297					buffer_size > args.f_cd_size) {
2298			error = ERANGE;
2299			goto outdrop;
2300		}
2301
2302		error = copyout(t_blob->csb_sha1, args.f_cd_hash, sizeof(t_blob->csb_sha1));
2303		if (error)
2304			goto outdrop;
2305		error = copyout(cd, args.f_cd_buffer, buffer_size);
2306		if (error)
2307			goto outdrop;
2308
2309		break;
2310	}
2311
2312	/*
2313	 * Set the vnode pointed to by 'fd'
2314	 * and tag it as the (potentially future) backing store
2315	 * for another filesystem
2316	 */
2317	case F_SETBACKINGSTORE: {
2318		if (fp->f_type != DTYPE_VNODE) {
2319			error = EBADF;
2320			goto out;
2321		}
2322
2323		vp = (struct vnode *)fp->f_data;
2324
2325		if (vp->v_tag != VT_HFS) {
2326			error = EINVAL;
2327			goto out;
2328		}
2329		proc_fdunlock(p);
2330
2331		if (vnode_getwithref(vp)) {
2332			error = ENOENT;
2333			goto outdrop;
2334		}
2335
2336		/* only proceed if you have write access */
2337		vfs_context_t ctx = vfs_context_current();
2338		if(vnode_authorize(vp, NULLVP, (KAUTH_VNODE_ACCESS | KAUTH_VNODE_WRITE_DATA), ctx) != 0) {
2339			vnode_put(vp);
2340			error = EBADF;
2341			goto outdrop;
2342		}
2343
2344
2345		/* If arg != 0, set, otherwise unset */
2346		if (uap->arg) {
2347			error = VNOP_IOCTL (vp, uap->cmd, (caddr_t)1, 0, &context);
2348		}
2349		else {
2350			error = VNOP_IOCTL (vp, uap->cmd, (caddr_t)NULL, 0, &context);
2351		}
2352
2353		vnode_put(vp);
2354		break;
2355	}
2356
2357	/*
2358	 * like F_GETPATH, but special semantics for
2359	 * the mobile time machine handler.
2360	 */
2361	case F_GETPATH_MTMINFO: {
2362		char *pathbufp;
2363		int pathlen;
2364
2365		if (fp->f_type != DTYPE_VNODE) {
2366			error = EBADF;
2367			goto out;
2368		}
2369		vp = (struct vnode *)fp->f_data;
2370		proc_fdunlock(p);
2371
2372		pathlen = MAXPATHLEN;
2373		MALLOC(pathbufp, char *, pathlen, M_TEMP, M_WAITOK);
2374		if (pathbufp == NULL) {
2375			error = ENOMEM;
2376			goto outdrop;
2377		}
2378		if ( (error = vnode_getwithref(vp)) == 0 ) {
2379			int backingstore = 0;
2380
2381			/* Check for error from vn_getpath before moving on */
2382			if ((error = vn_getpath(vp, pathbufp, &pathlen)) == 0) {
2383				if (vp->v_tag == VT_HFS) {
2384					error = VNOP_IOCTL (vp, uap->cmd, (caddr_t) &backingstore, 0, &context);
2385				}
2386				(void)vnode_put(vp);
2387
2388				if (error == 0) {
2389					error = copyout((caddr_t)pathbufp, argp, pathlen);
2390				}
2391				if (error == 0) {
2392					/*
2393					 * If the copyout was successful, now check to ensure
2394					 * that this vnode is not a BACKINGSTORE vnode.  mtmd
2395					 * wants the path regardless.
2396					 */
2397					if (backingstore) {
2398						error = EBUSY;
2399					}
2400				}
2401			} else
2402				(void)vnode_put(vp);
2403		}
2404		FREE(pathbufp, M_TEMP);
2405		goto outdrop;
2406	}
2407
2408	default:
2409		/*
2410		 * This is an fcntl() that we d not recognize at this level;
2411		 * if this is a vnode, we send it down into the VNOP_IOCTL
2412		 * for this vnode; this can include special devices, and will
2413		 * effectively overload fcntl() to send ioctl()'s.
2414		 */
2415		if((uap->cmd & IOC_VOID) && (uap->cmd & IOC_INOUT)){
2416            error = EINVAL;
2417			goto out;
2418		}
2419
2420		/* Catch any now-invalid fcntl() selectors */
2421		switch (uap->cmd) {
2422			case F_MARKDEPENDENCY:
2423				error = EINVAL;
2424				goto out;
2425			default:
2426				break;
2427		}
2428
2429		if (fp->f_type != DTYPE_VNODE) {
2430			error = EBADF;
2431			goto out;
2432		}
2433		vp = (struct vnode *)fp->f_data;
2434		proc_fdunlock(p);
2435
2436		if ( (error = vnode_getwithref(vp)) == 0 ) {
2437#define STK_PARAMS 128
2438			char stkbuf[STK_PARAMS];
2439			unsigned int size;
2440			caddr_t data, memp;
2441			/*
2442			 * For this to work properly, we have to copy in the
2443			 * ioctl() cmd argument if there is one; we must also
2444			 * check that a command parameter, if present, does
2445			 * not exceed the maximum command length dictated by
2446			 * the number of bits we have available in the command
2447			 * to represent a structure length.  Finally, we have
2448			 * to copy the results back out, if it is that type of
2449			 * ioctl().
2450			 */
2451			size = IOCPARM_LEN(uap->cmd);
2452			if (size > IOCPARM_MAX) {
2453				(void)vnode_put(vp);
2454				error = EINVAL;
2455				break;
2456			}
2457
2458			memp = NULL;
2459			if (size > sizeof (stkbuf)) {
2460				if ((memp = (caddr_t)kalloc(size)) == 0) {
2461					(void)vnode_put(vp);
2462					error = ENOMEM;
2463					goto outdrop;
2464				}
2465				data = memp;
2466			} else {
2467				data = &stkbuf[0];
2468			}
2469
2470			if (uap->cmd & IOC_IN) {
2471				if (size) {
2472					/* structure */
2473					error = copyin(argp, data, size);
2474					if (error) {
2475						(void)vnode_put(vp);
2476						if (memp)
2477							kfree(memp, size);
2478						goto outdrop;
2479					}
2480
2481					/* Bzero the section beyond that which was needed */
2482					if (size <= sizeof(stkbuf)) {
2483						bzero ( (((uint8_t*)data) + size), (sizeof(stkbuf) - size));
2484					}
2485				} else {
2486					/* int */
2487					if (is64bit) {
2488						*(user_addr_t *)data = argp;
2489					} else {
2490						*(uint32_t *)data = (uint32_t)argp;
2491					}
2492				};
2493			} else if ((uap->cmd & IOC_OUT) && size) {
2494				/*
2495				 * Zero the buffer so the user always
2496				 * gets back something deterministic.
2497				 */
2498				bzero(data, size);
2499			} else if (uap->cmd & IOC_VOID) {
2500				if (is64bit) {
2501				    *(user_addr_t *)data = argp;
2502				} else {
2503				    *(uint32_t *)data = (uint32_t)argp;
2504				}
2505			}
2506
2507			error = VNOP_IOCTL(vp, uap->cmd, CAST_DOWN(caddr_t, data), 0, &context);
2508
2509			(void)vnode_put(vp);
2510
2511			/* Copy any output data to user */
2512			if (error == 0 && (uap->cmd & IOC_OUT) && size)
2513				error = copyout(data, argp, size);
2514			if (memp)
2515				kfree(memp, size);
2516		}
2517		break;
2518	}
2519
2520outdrop:
2521	AUDIT_ARG(vnpath_withref, vp, ARG_VNODE1);
2522	fp_drop(p, fd, fp, 0);
2523	return(error);
2524out:
2525	fp_drop(p, fd, fp, 1);
2526	proc_fdunlock(p);
2527	return(error);
2528}
2529
2530
2531/*
2532 * finishdup
2533 *
2534 * Description:	Common code for dup, dup2, and fcntl(F_DUPFD).
2535 *
2536 * Parameters:	p				Process performing the dup
2537 *		old				The fd to dup
2538 *		new				The fd to dup it to
2539 *		fd_flags			Flags to augment the new fd
2540 *		retval				Pointer to the call return area
2541 *
2542 * Returns:	0				Success
2543 *		EBADF
2544 *		ENOMEM
2545 *
2546 * Implicit returns:
2547 *		*retval (modified)		The new descriptor
2548 *
2549 * Locks:	Assumes proc_fdlock for process pointing to fdp is held by
2550 *		the caller
2551 *
2552 * Notes:	This function may drop and reacquire this lock; it is unsafe
2553 *		for a caller to assume that other state protected by the lock
2554 *		has not been subsequently changed out from under it.
2555 */
2556int
2557finishdup(proc_t p,
2558    struct filedesc *fdp, int old, int new, int fd_flags, int32_t *retval)
2559{
2560	struct fileproc *nfp;
2561	struct fileproc *ofp;
2562#if CONFIG_MACF
2563	int error;
2564#endif
2565
2566#if DIAGNOSTIC
2567	proc_fdlock_assert(p, LCK_MTX_ASSERT_OWNED);
2568#endif
2569	if ((ofp = fdp->fd_ofiles[old]) == NULL ||
2570	    (fdp->fd_ofileflags[old] & UF_RESERVED)) {
2571		fdrelse(p, new);
2572		return (EBADF);
2573	}
2574	fg_ref(ofp);
2575
2576#if CONFIG_MACF
2577	error = mac_file_check_dup(proc_ucred(p), ofp->f_fglob, new);
2578	if (error) {
2579		fg_drop(ofp);
2580		fdrelse(p, new);
2581		return (error);
2582	}
2583#endif
2584
2585	proc_fdunlock(p);
2586
2587	nfp = fileproc_alloc_init(NULL);
2588
2589	proc_fdlock(p);
2590
2591	if (nfp == NULL) {
2592		fg_drop(ofp);
2593		fdrelse(p, new);
2594		return (ENOMEM);
2595	}
2596
2597	nfp->f_fglob = ofp->f_fglob;
2598
2599#if DIAGNOSTIC
2600	if (fdp->fd_ofiles[new] != 0)
2601		panic("finishdup: overwriting fd_ofiles with new %d", new);
2602	if ((fdp->fd_ofileflags[new] & UF_RESERVED) == 0)
2603		panic("finishdup: unreserved fileflags with new %d", new);
2604#endif
2605
2606	if (new > fdp->fd_lastfile)
2607		fdp->fd_lastfile = new;
2608	*fdflags(p, new) |= fd_flags;
2609	procfdtbl_releasefd(p, new, nfp);
2610	*retval = new;
2611	return (0);
2612}
2613
2614
2615/*
2616 * close
2617 *
2618 * Description:	The implementation of the close(2) system call
2619 *
2620 * Parameters:	p			Process in whose per process file table
2621 *					the close is to occur
2622 *		uap->fd			fd to be closed
2623 *		retval			<unused>
2624 *
2625 * Returns:	0			Success
2626 *	fp_lookup:EBADF			Bad file descriptor
2627 *      fp_guard_exception:???          Guarded file descriptor
2628 *	close_internal:EBADF
2629 *	close_internal:??? 		Anything returnable by a per-fileops
2630 *					close function
2631 */
2632int
2633close(proc_t p, struct close_args *uap, int32_t *retval)
2634{
2635	__pthread_testcancel(1);
2636	return(close_nocancel(p, (struct close_nocancel_args *)uap, retval));
2637}
2638
2639
2640int
2641close_nocancel(proc_t p, struct close_nocancel_args *uap, __unused int32_t *retval)
2642{
2643	struct fileproc *fp;
2644	int fd = uap->fd;
2645	int error;
2646
2647	AUDIT_SYSCLOSE(p, fd);
2648
2649	proc_fdlock(p);
2650
2651	if ( (error = fp_lookup(p,fd,&fp, 1)) ) {
2652		proc_fdunlock(p);
2653		return(error);
2654	}
2655
2656	if (FP_ISGUARDED(fp, GUARD_CLOSE)) {
2657		error = fp_guard_exception(p, fd, fp, kGUARD_EXC_CLOSE);
2658		(void) fp_drop(p, fd, fp, 1);
2659		proc_fdunlock(p);
2660		return (error);
2661	}
2662
2663	error = close_internal_locked(p, fd, fp, 0);
2664
2665	proc_fdunlock(p);
2666
2667	return (error);
2668}
2669
2670
2671/*
2672 * close_internal_locked
2673 *
2674 * Close a file descriptor.
2675 *
2676 * Parameters:	p			Process in whose per process file table
2677 *					the close is to occur
2678 *		fd			fd to be closed
2679 *		fp			fileproc associated with the fd
2680 *
2681 * Returns:	0			Success
2682 *		EBADF			fd already in close wait state
2683 *	closef_locked:??? 		Anything returnable by a per-fileops
2684 *					close function
2685 *
2686 * Locks:	Assumes proc_fdlock for process is held by the caller and returns
2687 *		with lock held
2688 *
2689 * Notes:	This function may drop and reacquire this lock; it is unsafe
2690 *		for a caller to assume that other state protected by the lock
2691 *		has not been subsequently changed out from under it.
2692 */
2693int
2694close_internal_locked(proc_t p, int fd, struct fileproc *fp, int flags)
2695{
2696	struct filedesc *fdp = p->p_fd;
2697	int error =0;
2698	int resvfd = flags & FD_DUP2RESV;
2699
2700
2701#if DIAGNOSTIC
2702	proc_fdlock_assert(p, LCK_MTX_ASSERT_OWNED);
2703#endif
2704
2705	/* Keep people from using the filedesc while we are closing it */
2706	procfdtbl_markclosefd(p, fd);
2707
2708
2709	if ((fp->f_flags & FP_CLOSING) == FP_CLOSING) {
2710		panic("close_internal_locked: being called on already closing fd");
2711	}
2712
2713
2714#if DIAGNOSTIC
2715	if ((fdp->fd_ofileflags[fd] & UF_RESERVED) == 0)
2716		panic("close_internal: unreserved fileflags with fd %d", fd);
2717#endif
2718
2719	fp->f_flags |= FP_CLOSING;
2720
2721	if ( (fp->f_flags & FP_AIOISSUED) || kauth_authorize_fileop_has_listeners() ) {
2722
2723	        proc_fdunlock(p);
2724
2725		if ( (fp->f_type == DTYPE_VNODE) && kauth_authorize_fileop_has_listeners() ) {
2726		        /*
2727			 * call out to allow 3rd party notification of close.
2728			 * Ignore result of kauth_authorize_fileop call.
2729			 */
2730		        if (vnode_getwithref((vnode_t)fp->f_data) == 0) {
2731		        	u_int	fileop_flags = 0;
2732		        	if ((fp->f_flags & FP_WRITTEN) != 0)
2733		        		fileop_flags |= KAUTH_FILEOP_CLOSE_MODIFIED;
2734			        kauth_authorize_fileop(fp->f_fglob->fg_cred, KAUTH_FILEOP_CLOSE,
2735						       (uintptr_t)fp->f_data, (uintptr_t)fileop_flags);
2736				vnode_put((vnode_t)fp->f_data);
2737			}
2738		}
2739		if (fp->f_flags & FP_AIOISSUED)
2740		        /*
2741			 * cancel all async IO requests that can be cancelled.
2742			 */
2743		        _aio_close( p, fd );
2744
2745		proc_fdlock(p);
2746	}
2747
2748	if (fd < fdp->fd_knlistsize)
2749		knote_fdclose(p, fd);
2750
2751	if (fp->f_flags & FP_WAITEVENT)
2752		(void)waitevent_close(p, fp);
2753
2754	fileproc_drain(p, fp);
2755
2756	if (resvfd == 0) {
2757		_fdrelse(p, fd);
2758	} else {
2759		procfdtbl_reservefd(p, fd);
2760	}
2761
2762	error = closef_locked(fp, fp->f_fglob, p);
2763	if ((fp->f_flags & FP_WAITCLOSE) == FP_WAITCLOSE)
2764		wakeup(&fp->f_flags);
2765	fp->f_flags &= ~(FP_WAITCLOSE | FP_CLOSING);
2766
2767	proc_fdunlock(p);
2768
2769	fileproc_free(fp);
2770
2771	proc_fdlock(p);
2772
2773#if DIAGNOSTIC
2774	if (resvfd != 0) {
2775		if ((fdp->fd_ofileflags[fd] & UF_RESERVED) == 0)
2776			panic("close with reserved fd returns with freed fd:%d: proc: %p", fd, p);
2777	}
2778#endif
2779
2780	return(error);
2781}
2782
2783
2784/*
2785 * fstat1
2786 *
2787 * Description:	Return status information about a file descriptor.
2788 *
2789 * Parameters:	p				The process doing the fstat
2790 *		fd				The fd to stat
2791 *		ub				The user stat buffer
2792 *		xsecurity			The user extended security
2793 *						buffer, or 0 if none
2794 *		xsecurity_size			The size of xsecurity, or 0
2795 *						if no xsecurity
2796 *		isstat64			Flag to indicate 64 bit version
2797 *						for inode size, etc.
2798 *
2799 * Returns:	0				Success
2800 *		EBADF
2801 *		EFAULT
2802 *	fp_lookup:EBADF				Bad file descriptor
2803 *	vnode_getwithref:???
2804 *	copyout:EFAULT
2805 *	vnode_getwithref:???
2806 *	vn_stat:???
2807 *	soo_stat:???
2808 *	pipe_stat:???
2809 *	pshm_stat:???
2810 *	kqueue_stat:???
2811 *
2812 * Notes:	Internal implementation for all other fstat() related
2813 *		functions
2814 *
2815 *		XXX switch on node type is bogus; need a stat in struct
2816 *		XXX fileops instead.
2817 */
2818static int
2819fstat1(proc_t p, int fd, user_addr_t ub, user_addr_t xsecurity, user_addr_t xsecurity_size, int isstat64)
2820{
2821	struct fileproc *fp;
2822	union {
2823		struct stat sb;
2824		struct stat64 sb64;
2825	} source;
2826	union {
2827		struct user64_stat user64_sb;
2828		struct user32_stat user32_sb;
2829		struct user64_stat64 user64_sb64;
2830		struct user32_stat64 user32_sb64;
2831	} dest;
2832	int error, my_size;
2833	file_type_t type;
2834	caddr_t data;
2835	kauth_filesec_t fsec;
2836	user_size_t xsecurity_bufsize;
2837	vfs_context_t ctx = vfs_context_current();
2838	void * sbptr;
2839
2840
2841	AUDIT_ARG(fd, fd);
2842
2843	if ((error = fp_lookup(p, fd, &fp, 0)) != 0) {
2844		return(error);
2845	}
2846	type = fp->f_type;
2847	data = fp->f_data;
2848	fsec = KAUTH_FILESEC_NONE;
2849
2850	sbptr = (void *)&source;
2851
2852	switch (type) {
2853
2854	case DTYPE_VNODE:
2855		if ((error = vnode_getwithref((vnode_t)data)) == 0) {
2856			/*
2857			 * If the caller has the file open, and is not
2858			 * requesting extended security information, we are
2859			 * going to let them get the basic stat information.
2860			 */
2861			if (xsecurity == USER_ADDR_NULL) {
2862				error = vn_stat_noauth((vnode_t)data, sbptr, NULL, isstat64, ctx);
2863			} else {
2864				error = vn_stat((vnode_t)data, sbptr, &fsec, isstat64, ctx);
2865			}
2866
2867			AUDIT_ARG(vnpath, (struct vnode *)data, ARG_VNODE1);
2868			(void)vnode_put((vnode_t)data);
2869		}
2870		break;
2871
2872#if SOCKETS
2873	case DTYPE_SOCKET:
2874		error = soo_stat((struct socket *)data, sbptr, isstat64);
2875		break;
2876#endif /* SOCKETS */
2877
2878	case DTYPE_PIPE:
2879		error = pipe_stat((void *)data, sbptr, isstat64);
2880		break;
2881
2882	case DTYPE_PSXSHM:
2883		error = pshm_stat((void *)data, sbptr, isstat64);
2884		break;
2885
2886	case DTYPE_KQUEUE:
2887		error = kqueue_stat((void *)data, sbptr, isstat64, p);
2888		break;
2889
2890	default:
2891		error = EBADF;
2892		goto out;
2893	}
2894	if (error == 0) {
2895		caddr_t sbp;
2896
2897		if (isstat64 != 0) {
2898			source.sb64.st_lspare = 0;
2899			source.sb64.st_qspare[0] = 0LL;
2900			source.sb64.st_qspare[1] = 0LL;
2901
2902			if (IS_64BIT_PROCESS(current_proc())) {
2903				munge_user64_stat64(&source.sb64, &dest.user64_sb64);
2904				my_size = sizeof(dest.user64_sb64);
2905				sbp = (caddr_t)&dest.user64_sb64;
2906			} else {
2907				munge_user32_stat64(&source.sb64, &dest.user32_sb64);
2908				my_size = sizeof(dest.user32_sb64);
2909				sbp = (caddr_t)&dest.user32_sb64;
2910			}
2911		} else {
2912			source.sb.st_lspare = 0;
2913			source.sb.st_qspare[0] = 0LL;
2914			source.sb.st_qspare[1] = 0LL;
2915			if (IS_64BIT_PROCESS(current_proc())) {
2916				munge_user64_stat(&source.sb, &dest.user64_sb);
2917				my_size = sizeof(dest.user64_sb);
2918				sbp = (caddr_t)&dest.user64_sb;
2919			} else {
2920				munge_user32_stat(&source.sb, &dest.user32_sb);
2921				my_size = sizeof(dest.user32_sb);
2922				sbp = (caddr_t)&dest.user32_sb;
2923			}
2924		}
2925
2926		error = copyout(sbp, ub, my_size);
2927	}
2928
2929	/* caller wants extended security information? */
2930	if (xsecurity != USER_ADDR_NULL) {
2931
2932		/* did we get any? */
2933		 if (fsec == KAUTH_FILESEC_NONE) {
2934			if (susize(xsecurity_size, 0) != 0) {
2935				error = EFAULT;
2936				goto out;
2937			}
2938		} else {
2939			/* find the user buffer size */
2940			xsecurity_bufsize = fusize(xsecurity_size);
2941
2942			/* copy out the actual data size */
2943			if (susize(xsecurity_size, KAUTH_FILESEC_COPYSIZE(fsec)) != 0) {
2944				error = EFAULT;
2945				goto out;
2946			}
2947
2948			/* if the caller supplied enough room, copy out to it */
2949			if (xsecurity_bufsize >= KAUTH_FILESEC_COPYSIZE(fsec))
2950				error = copyout(fsec, xsecurity, KAUTH_FILESEC_COPYSIZE(fsec));
2951		}
2952	}
2953out:
2954	fp_drop(p, fd, fp, 0);
2955	if (fsec != NULL)
2956		kauth_filesec_free(fsec);
2957	return (error);
2958}
2959
2960
2961/*
2962 * fstat_extended
2963 *
2964 * Description:	Extended version of fstat supporting returning extended
2965 *		security information
2966 *
2967 * Parameters:	p				The process doing the fstat
2968 *		uap->fd				The fd to stat
2969 *		uap->ub				The user stat buffer
2970 *		uap->xsecurity			The user extended security
2971 *						buffer, or 0 if none
2972 *		uap->xsecurity_size		The size of xsecurity, or 0
2973 *
2974 * Returns:	0				Success
2975 *		!0				Errno (see fstat1)
2976 */
2977int
2978fstat_extended(proc_t p, struct fstat_extended_args *uap, __unused int32_t *retval)
2979{
2980	return(fstat1(p, uap->fd, uap->ub, uap->xsecurity, uap->xsecurity_size, 0));
2981}
2982
2983
2984/*
2985 * fstat
2986 *
2987 * Description:	Get file status for the file associated with fd
2988 *
2989 * Parameters:	p				The process doing the fstat
2990 *		uap->fd				The fd to stat
2991 *		uap->ub				The user stat buffer
2992 *
2993 * Returns:	0				Success
2994 *		!0				Errno (see fstat1)
2995 */
2996int
2997fstat(proc_t p, register struct fstat_args *uap, __unused int32_t *retval)
2998{
2999	return(fstat1(p, uap->fd, uap->ub, 0, 0, 0));
3000}
3001
3002
3003/*
3004 * fstat64_extended
3005 *
3006 * Description:	Extended version of fstat64 supporting returning extended
3007 *		security information
3008 *
3009 * Parameters:	p				The process doing the fstat
3010 *		uap->fd				The fd to stat
3011 *		uap->ub				The user stat buffer
3012 *		uap->xsecurity			The user extended security
3013 *						buffer, or 0 if none
3014 *		uap->xsecurity_size		The size of xsecurity, or 0
3015 *
3016 * Returns:	0				Success
3017 *		!0				Errno (see fstat1)
3018 */
3019int
3020fstat64_extended(proc_t p, struct fstat64_extended_args *uap, __unused int32_t *retval)
3021{
3022	return(fstat1(p, uap->fd, uap->ub, uap->xsecurity, uap->xsecurity_size, 1));
3023}
3024
3025
3026/*
3027 * fstat64
3028 *
3029 * Description:	Get 64 bit version of the file status for the file associated
3030 *		with fd
3031 *
3032 * Parameters:	p				The process doing the fstat
3033 *		uap->fd				The fd to stat
3034 *		uap->ub				The user stat buffer
3035 *
3036 * Returns:	0				Success
3037 *		!0				Errno (see fstat1)
3038 */
3039int
3040fstat64(proc_t p, register struct fstat64_args *uap, __unused int32_t *retval)
3041{
3042	return(fstat1(p, uap->fd, uap->ub, 0, 0, 1));
3043}
3044
3045
3046/*
3047 * fpathconf
3048 *
3049 * Description:	Return pathconf information about a file descriptor.
3050 *
3051 * Parameters:	p				Process making the request
3052 *		uap->fd				fd to get information about
3053 *		uap->name			Name of information desired
3054 *		retval				Pointer to the call return area
3055 *
3056 * Returns:	0				Success
3057 *		EINVAL
3058 *	fp_lookup:EBADF				Bad file descriptor
3059 *	vnode_getwithref:???
3060 *	vn_pathconf:???
3061 *
3062 * Implicit returns:
3063 *		*retval (modified)		Returned information (numeric)
3064 */
3065int
3066fpathconf(proc_t p, struct fpathconf_args *uap, int32_t *retval)
3067{
3068	int fd = uap->fd;
3069	struct fileproc *fp;
3070	struct vnode *vp;
3071	int error = 0;
3072	file_type_t type;
3073	caddr_t data;
3074
3075
3076	AUDIT_ARG(fd, uap->fd);
3077	if ( (error = fp_lookup(p, fd, &fp, 0)) )
3078		return(error);
3079	type = fp->f_type;
3080	data = fp->f_data;
3081
3082	switch (type) {
3083
3084	case DTYPE_SOCKET:
3085		if (uap->name != _PC_PIPE_BUF) {
3086			error = EINVAL;
3087			goto out;
3088		}
3089		*retval = PIPE_BUF;
3090		error = 0;
3091		goto out;
3092
3093	case DTYPE_PIPE:
3094		if (uap->name != _PC_PIPE_BUF) {
3095			error = EINVAL;
3096			goto out;
3097		}
3098		*retval = PIPE_BUF;
3099		error = 0;
3100		goto out;
3101
3102	case DTYPE_VNODE:
3103		vp = (struct vnode *)data;
3104
3105		if ( (error = vnode_getwithref(vp)) == 0) {
3106		        AUDIT_ARG(vnpath, vp, ARG_VNODE1);
3107
3108			error = vn_pathconf(vp, uap->name, retval, vfs_context_current());
3109
3110			(void)vnode_put(vp);
3111		}
3112		goto out;
3113
3114	default:
3115		error = EINVAL;
3116		goto out;
3117
3118	}
3119	/*NOTREACHED*/
3120out:
3121	fp_drop(p, fd, fp, 0);
3122	return(error);
3123}
3124
3125/*
3126 * Statistics counter for the number of times a process calling fdalloc()
3127 * has resulted in an expansion of the per process open file table.
3128 *
3129 * XXX This would likely be of more use if it were per process
3130 */
3131int fdexpand;
3132
3133
3134/*
3135 * fdalloc
3136 *
3137 * Description:	Allocate a file descriptor for the process.
3138 *
3139 * Parameters:	p				Process to allocate the fd in
3140 *		want				The fd we would prefer to get
3141 *		result				Pointer to fd we got
3142 *
3143 * Returns:	0				Success
3144 *		EMFILE
3145 *		ENOMEM
3146 *
3147 * Implicit returns:
3148 *		*result (modified)		The fd which was allocated
3149 */
3150int
3151fdalloc(proc_t p, int want, int *result)
3152{
3153	struct filedesc *fdp = p->p_fd;
3154	int i;
3155	int lim, last, numfiles, oldnfiles;
3156	struct fileproc **newofiles, **ofiles;
3157	char *newofileflags;
3158
3159	/*
3160	 * Search for a free descriptor starting at the higher
3161	 * of want or fd_freefile.  If that fails, consider
3162	 * expanding the ofile array.
3163	 */
3164#if DIAGNOSTIC
3165	proc_fdlock_assert(p, LCK_MTX_ASSERT_OWNED);
3166#endif
3167
3168	lim = min((int)p->p_rlimit[RLIMIT_NOFILE].rlim_cur, maxfiles);
3169	for (;;) {
3170		last = min(fdp->fd_nfiles, lim);
3171		if ((i = want) < fdp->fd_freefile)
3172			i = fdp->fd_freefile;
3173		for (; i < last; i++) {
3174			if (fdp->fd_ofiles[i] == NULL && !(fdp->fd_ofileflags[i] & UF_RESERVED)) {
3175				procfdtbl_reservefd(p, i);
3176				if (i > fdp->fd_lastfile)
3177					fdp->fd_lastfile = i;
3178				if (want <= fdp->fd_freefile)
3179					fdp->fd_freefile = i;
3180				*result = i;
3181				return (0);
3182			}
3183		}
3184
3185		/*
3186		 * No space in current array.  Expand?
3187		 */
3188		if (fdp->fd_nfiles >= lim)
3189			return (EMFILE);
3190		if (fdp->fd_nfiles < NDEXTENT)
3191			numfiles = NDEXTENT;
3192		else
3193			numfiles = 2 * fdp->fd_nfiles;
3194		/* Enforce lim */
3195		if (numfiles > lim)
3196			numfiles = lim;
3197		proc_fdunlock(p);
3198		MALLOC_ZONE(newofiles, struct fileproc **,
3199				numfiles * OFILESIZE, M_OFILETABL, M_WAITOK);
3200		proc_fdlock(p);
3201		if (newofiles == NULL) {
3202			return (ENOMEM);
3203		}
3204		if (fdp->fd_nfiles >= numfiles) {
3205			FREE_ZONE(newofiles, numfiles * OFILESIZE, M_OFILETABL);
3206			continue;
3207		}
3208		newofileflags = (char *) &newofiles[numfiles];
3209		/*
3210		 * Copy the existing ofile and ofileflags arrays
3211		 * and zero the new portion of each array.
3212		 */
3213		oldnfiles = fdp->fd_nfiles;
3214		(void) memcpy(newofiles, fdp->fd_ofiles,
3215				oldnfiles * sizeof(*fdp->fd_ofiles));
3216		(void) memset(&newofiles[oldnfiles], 0,
3217				(numfiles - oldnfiles) * sizeof(*fdp->fd_ofiles));
3218
3219		(void) memcpy(newofileflags, fdp->fd_ofileflags,
3220				oldnfiles * sizeof(*fdp->fd_ofileflags));
3221		(void) memset(&newofileflags[oldnfiles], 0,
3222				(numfiles - oldnfiles) *
3223						sizeof(*fdp->fd_ofileflags));
3224		ofiles = fdp->fd_ofiles;
3225		fdp->fd_ofiles = newofiles;
3226		fdp->fd_ofileflags = newofileflags;
3227		fdp->fd_nfiles = numfiles;
3228		FREE_ZONE(ofiles, oldnfiles * OFILESIZE, M_OFILETABL);
3229		fdexpand++;
3230	}
3231}
3232
3233
3234/*
3235 * fdavail
3236 *
3237 * Description:	Check to see whether n user file descriptors are available
3238 *		to the process p.
3239 *
3240 * Parameters:	p				Process to check in
3241 *		n				The number of fd's desired
3242 *
3243 * Returns:	0				No
3244 *		1				Yes
3245 *
3246 * Locks:	Assumes proc_fdlock for process is held by the caller
3247 *
3248 * Notes:	The answer only remains valid so long as the proc_fdlock is
3249 *		held by the caller.
3250 */
3251int
3252fdavail(proc_t p, int n)
3253{
3254	struct filedesc *fdp = p->p_fd;
3255	struct fileproc **fpp;
3256	char *flags;
3257	int i, lim;
3258
3259	lim = min((int)p->p_rlimit[RLIMIT_NOFILE].rlim_cur, maxfiles);
3260	if ((i = lim - fdp->fd_nfiles) > 0 && (n -= i) <= 0)
3261		return (1);
3262	fpp = &fdp->fd_ofiles[fdp->fd_freefile];
3263	flags = &fdp->fd_ofileflags[fdp->fd_freefile];
3264	for (i = fdp->fd_nfiles - fdp->fd_freefile; --i >= 0; fpp++, flags++)
3265		if (*fpp == NULL && !(*flags & UF_RESERVED) && --n <= 0)
3266			return (1);
3267	return (0);
3268}
3269
3270
3271/*
3272 * fdrelse
3273 *
3274 * Description:	Legacy KPI wrapper function for _fdrelse
3275 *
3276 * Parameters:	p				Process in which fd lives
3277 *		fd				fd to free
3278 *
3279 * Returns:	void
3280 *
3281 * Locks:	Assumes proc_fdlock for process is held by the caller
3282 */
3283void
3284fdrelse(proc_t p, int fd)
3285{
3286	_fdrelse(p, fd);
3287}
3288
3289
3290/*
3291 * fdgetf_noref
3292 *
3293 * Description:	Get the fileproc pointer for the given fd from the per process
3294 *		open file table without taking an explicit reference on it.
3295 *
3296 * Parameters:	p				Process containing fd
3297 *		fd				fd to obtain fileproc for
3298 *		resultfp			Pointer to pointer return area
3299 *
3300 * Returns:	0				Success
3301 *		EBADF
3302 *
3303 * Implicit returns:
3304 *		*resultfp (modified)		Pointer to fileproc pointer
3305 *
3306 * Locks:	Assumes proc_fdlock for process is held by the caller
3307 *
3308 * Notes:	Because there is no reference explicitly taken, the returned
3309 *		fileproc pointer is only valid so long as the proc_fdlock
3310 *		remains held by the caller.
3311 */
3312int
3313fdgetf_noref(proc_t p, int fd, struct fileproc **resultfp)
3314{
3315	struct filedesc *fdp = p->p_fd;
3316	struct fileproc *fp;
3317
3318	if (fd < 0 || fd >= fdp->fd_nfiles ||
3319			(fp = fdp->fd_ofiles[fd]) == NULL ||
3320			(fdp->fd_ofileflags[fd] & UF_RESERVED)) {
3321		return (EBADF);
3322	}
3323	if (resultfp)
3324		*resultfp = fp;
3325	return (0);
3326}
3327
3328
3329/*
3330 * fp_getfvp
3331 *
3332 * Description:	Get fileproc and vnode pointer for a given fd from the per
3333 *		process open file table of the specified process, and if
3334 *		successful, increment the f_iocount
3335 *
3336 * Parameters:	p				Process in which fd lives
3337 *		fd				fd to get information for
3338 *		resultfp			Pointer to result fileproc
3339 *						pointer area, or 0 if none
3340 *		resultvp			Pointer to result vnode pointer
3341 *						area, or 0 if none
3342 *
3343 * Returns:	0				Success
3344 *		EBADF				Bad file descriptor
3345 *		ENOTSUP				fd does not refer to a vnode
3346 *
3347 * Implicit returns:
3348 *		*resultfp (modified)		Fileproc pointer
3349 *		*resultvp (modified)		vnode pointer
3350 *
3351 * Notes:	The resultfp and resultvp fields are optional, and may be
3352 *		independently specified as NULL to skip returning information
3353 *
3354 * Locks:	Internally takes and releases proc_fdlock
3355 */
3356int
3357fp_getfvp(proc_t p, int fd, struct fileproc **resultfp, struct vnode **resultvp)
3358{
3359	struct filedesc *fdp = p->p_fd;
3360	struct fileproc *fp;
3361
3362	proc_fdlock_spin(p);
3363	if (fd < 0 || fd >= fdp->fd_nfiles ||
3364			(fp = fdp->fd_ofiles[fd]) == NULL ||
3365			(fdp->fd_ofileflags[fd] & UF_RESERVED)) {
3366		proc_fdunlock(p);
3367		return (EBADF);
3368	}
3369	if (fp->f_type != DTYPE_VNODE) {
3370		proc_fdunlock(p);
3371		return(ENOTSUP);
3372	}
3373	fp->f_iocount++;
3374
3375	if (resultfp)
3376		*resultfp = fp;
3377	if (resultvp)
3378		*resultvp = (struct vnode *)fp->f_data;
3379	proc_fdunlock(p);
3380
3381	return (0);
3382}
3383
3384
3385/*
3386 * fp_getfvpandvid
3387 *
3388 * Description:	Get fileproc, vnode pointer, and vid for a given fd from the
3389 *		per process open file table of the specified process, and if
3390 *		successful, increment the f_iocount
3391 *
3392 * Parameters:	p				Process in which fd lives
3393 *		fd				fd to get information for
3394 *		resultfp			Pointer to result fileproc
3395 *						pointer area, or 0 if none
3396 *		resultvp			Pointer to result vnode pointer
3397 *						area, or 0 if none
3398 *		vidp				Pointer to resuld vid area
3399 *
3400 * Returns:	0				Success
3401 *		EBADF				Bad file descriptor
3402 *		ENOTSUP				fd does not refer to a vnode
3403 *
3404 * Implicit returns:
3405 *		*resultfp (modified)		Fileproc pointer
3406 *		*resultvp (modified)		vnode pointer
3407 *		*vidp				vid value
3408 *
3409 * Notes:	The resultfp and resultvp fields are optional, and may be
3410 *		independently specified as NULL to skip returning information
3411 *
3412 * Locks:	Internally takes and releases proc_fdlock
3413 */
3414int
3415fp_getfvpandvid(proc_t p, int fd, struct fileproc **resultfp,
3416		struct vnode **resultvp, uint32_t *vidp)
3417{
3418	struct filedesc *fdp = p->p_fd;
3419	struct fileproc *fp;
3420
3421	proc_fdlock_spin(p);
3422	if (fd < 0 || fd >= fdp->fd_nfiles ||
3423			(fp = fdp->fd_ofiles[fd]) == NULL ||
3424			(fdp->fd_ofileflags[fd] & UF_RESERVED)) {
3425		proc_fdunlock(p);
3426		return (EBADF);
3427	}
3428	if (fp->f_type != DTYPE_VNODE) {
3429		proc_fdunlock(p);
3430		return(ENOTSUP);
3431	}
3432	fp->f_iocount++;
3433
3434	if (resultfp)
3435		*resultfp = fp;
3436	if (resultvp)
3437		*resultvp = (struct vnode *)fp->f_data;
3438	if (vidp)
3439		*vidp = (uint32_t)vnode_vid((struct vnode *)fp->f_data);
3440	proc_fdunlock(p);
3441
3442	return (0);
3443}
3444
3445
3446/*
3447 * fp_getfsock
3448 *
3449 * Description:	Get fileproc and socket pointer for a given fd from the
3450 *		per process open file table of the specified process, and if
3451 *		successful, increment the f_iocount
3452 *
3453 * Parameters:	p				Process in which fd lives
3454 *		fd				fd to get information for
3455 *		resultfp			Pointer to result fileproc
3456 *						pointer area, or 0 if none
3457 *		results				Pointer to result socket
3458 *						pointer area, or 0 if none
3459 *
3460 * Returns:	EBADF			The file descriptor is invalid
3461 *		EOPNOTSUPP		The file descriptor is not a socket
3462 *		0			Success
3463 *
3464 * Implicit returns:
3465 *		*resultfp (modified)		Fileproc pointer
3466 *		*results (modified)		socket pointer
3467 *
3468 * Notes:	EOPNOTSUPP should probably be ENOTSOCK; this function is only
3469 *		ever called from accept1().
3470 */
3471int
3472fp_getfsock(proc_t p, int fd, struct fileproc **resultfp,
3473	    struct socket **results)
3474{
3475	struct filedesc *fdp = p->p_fd;
3476	struct fileproc *fp;
3477
3478	proc_fdlock_spin(p);
3479	if (fd < 0 || fd >= fdp->fd_nfiles ||
3480			(fp = fdp->fd_ofiles[fd]) == NULL ||
3481			(fdp->fd_ofileflags[fd] & UF_RESERVED)) {
3482		proc_fdunlock(p);
3483		return (EBADF);
3484	}
3485	if (fp->f_type != DTYPE_SOCKET) {
3486		proc_fdunlock(p);
3487		return(EOPNOTSUPP);
3488	}
3489	fp->f_iocount++;
3490
3491	if (resultfp)
3492		*resultfp = fp;
3493	if (results)
3494		*results = (struct socket *)fp->f_data;
3495	proc_fdunlock(p);
3496
3497	return (0);
3498}
3499
3500
3501/*
3502 * fp_getfkq
3503 *
3504 * Description:	Get fileproc and kqueue pointer for a given fd from the
3505 *		per process open file table of the specified process, and if
3506 *		successful, increment the f_iocount
3507 *
3508 * Parameters:	p				Process in which fd lives
3509 *		fd				fd to get information for
3510 *		resultfp			Pointer to result fileproc
3511 *						pointer area, or 0 if none
3512 *		resultkq			Pointer to result kqueue
3513 *						pointer area, or 0 if none
3514 *
3515 * Returns:	EBADF			The file descriptor is invalid
3516 *		EBADF			The file descriptor is not a socket
3517 *		0			Success
3518 *
3519 * Implicit returns:
3520 *		*resultfp (modified)		Fileproc pointer
3521 *		*resultkq (modified)		kqueue pointer
3522 *
3523 * Notes:	The second EBADF should probably be something else to make
3524 *		the error condition distinct.
3525 */
3526int
3527fp_getfkq(proc_t p, int fd, struct fileproc **resultfp,
3528	  struct kqueue **resultkq)
3529{
3530	struct filedesc *fdp = p->p_fd;
3531	struct fileproc *fp;
3532
3533	proc_fdlock_spin(p);
3534	if ( fd < 0 || fd >= fdp->fd_nfiles ||
3535			(fp = fdp->fd_ofiles[fd]) == NULL ||
3536			(fdp->fd_ofileflags[fd] & UF_RESERVED)) {
3537		proc_fdunlock(p);
3538		return (EBADF);
3539	}
3540	if (fp->f_type != DTYPE_KQUEUE) {
3541		proc_fdunlock(p);
3542		return(EBADF);
3543	}
3544	fp->f_iocount++;
3545
3546	if (resultfp)
3547		*resultfp = fp;
3548	if (resultkq)
3549		*resultkq = (struct kqueue *)fp->f_data;
3550	proc_fdunlock(p);
3551
3552	return (0);
3553}
3554
3555
3556/*
3557 * fp_getfpshm
3558 *
3559 * Description:	Get fileproc and POSIX shared memory pointer for a given fd
3560 *		from the per process open file table of the specified process
3561 *		and if successful, increment the f_iocount
3562 *
3563 * Parameters:	p				Process in which fd lives
3564 *		fd				fd to get information for
3565 *		resultfp			Pointer to result fileproc
3566 *						pointer area, or 0 if none
3567 *		resultpshm			Pointer to result POSIX
3568 *						shared memory pointer
3569 *						pointer area, or 0 if none
3570 *
3571 * Returns:	EBADF			The file descriptor is invalid
3572 *		EBADF			The file descriptor is not a POSIX
3573 *					shared memory area
3574 *		0			Success
3575 *
3576 * Implicit returns:
3577 *		*resultfp (modified)		Fileproc pointer
3578 *		*resultpshm (modified)		POSIX shared memory pointer
3579 *
3580 * Notes:	The second EBADF should probably be something else to make
3581 *		the error condition distinct.
3582 */
3583int
3584fp_getfpshm(proc_t p, int fd, struct fileproc **resultfp,
3585	    struct pshmnode **resultpshm)
3586{
3587	struct filedesc *fdp = p->p_fd;
3588	struct fileproc *fp;
3589
3590	proc_fdlock_spin(p);
3591	if (fd < 0 || fd >= fdp->fd_nfiles ||
3592			(fp = fdp->fd_ofiles[fd]) == NULL ||
3593			(fdp->fd_ofileflags[fd] & UF_RESERVED)) {
3594		proc_fdunlock(p);
3595		return (EBADF);
3596	}
3597	if (fp->f_type != DTYPE_PSXSHM) {
3598
3599		proc_fdunlock(p);
3600		return(EBADF);
3601	}
3602	fp->f_iocount++;
3603
3604	if (resultfp)
3605		*resultfp = fp;
3606	if (resultpshm)
3607		*resultpshm = (struct pshmnode *)fp->f_data;
3608	proc_fdunlock(p);
3609
3610	return (0);
3611}
3612
3613
3614/*
3615 * fp_getfsem
3616 *
3617 * Description:	Get fileproc and POSIX semaphore pointer for a given fd from
3618 *		the per process open file table of the specified process
3619 *		and if successful, increment the f_iocount
3620 *
3621 * Parameters:	p				Process in which fd lives
3622 *		fd				fd to get information for
3623 *		resultfp			Pointer to result fileproc
3624 *						pointer area, or 0 if none
3625 *		resultpsem			Pointer to result POSIX
3626 *						semaphore pointer area, or
3627 *						0 if none
3628 *
3629 * Returns:	EBADF			The file descriptor is invalid
3630 *		EBADF			The file descriptor is not a POSIX
3631 *					semaphore
3632 *		0			Success
3633 *
3634 * Implicit returns:
3635 *		*resultfp (modified)		Fileproc pointer
3636 *		*resultpsem (modified)		POSIX semaphore pointer
3637 *
3638 * Notes:	The second EBADF should probably be something else to make
3639 *		the error condition distinct.
3640 *
3641 *		In order to support unnamed POSIX semaphores, the named
3642 *		POSIX semaphores will have to move out of the per-process
3643 *		open filetable, and into a global table that is shared with
3644 *		unnamed POSIX semaphores, since unnamed POSIX semaphores
3645 *		are typically used by declaring instances in shared memory,
3646 *		and there's no other way to do this without changing the
3647 *		underlying type, which would introduce binary compatibility
3648 *		issues.
3649 */
3650int
3651fp_getfpsem(proc_t p, int fd, struct fileproc **resultfp,
3652	    struct psemnode **resultpsem)
3653{
3654	struct filedesc *fdp = p->p_fd;
3655	struct fileproc *fp;
3656
3657	proc_fdlock_spin(p);
3658	if (fd < 0 || fd >= fdp->fd_nfiles ||
3659			(fp = fdp->fd_ofiles[fd]) == NULL ||
3660			(fdp->fd_ofileflags[fd] & UF_RESERVED)) {
3661		proc_fdunlock(p);
3662		return (EBADF);
3663	}
3664	if (fp->f_type != DTYPE_PSXSEM) {
3665		proc_fdunlock(p);
3666		return(EBADF);
3667	}
3668	fp->f_iocount++;
3669
3670	if (resultfp)
3671		*resultfp = fp;
3672	if (resultpsem)
3673		*resultpsem = (struct psemnode *)fp->f_data;
3674	proc_fdunlock(p);
3675
3676	return (0);
3677}
3678
3679
3680/*
3681 * fp_getfpipe
3682 *
3683 * Description:	Get fileproc and pipe pointer for a given fd from the
3684 *		per process open file table of the specified process
3685 *		and if successful, increment the f_iocount
3686 *
3687 * Parameters:	p				Process in which fd lives
3688 *		fd				fd to get information for
3689 *		resultfp			Pointer to result fileproc
3690 *						pointer area, or 0 if none
3691 *		resultpipe			Pointer to result pipe
3692 *						pointer area, or 0 if none
3693 *
3694 * Returns:	EBADF			The file descriptor is invalid
3695 *		EBADF			The file descriptor is not a socket
3696 *		0			Success
3697 *
3698 * Implicit returns:
3699 *		*resultfp (modified)		Fileproc pointer
3700 *		*resultpipe (modified)		pipe pointer
3701 *
3702 * Notes:	The second EBADF should probably be something else to make
3703 *		the error condition distinct.
3704 */
3705int
3706fp_getfpipe(proc_t p, int fd, struct fileproc **resultfp,
3707	    struct pipe **resultpipe)
3708{
3709	struct filedesc *fdp = p->p_fd;
3710	struct fileproc *fp;
3711
3712	proc_fdlock_spin(p);
3713	if (fd < 0 || fd >= fdp->fd_nfiles ||
3714			(fp = fdp->fd_ofiles[fd]) == NULL ||
3715			(fdp->fd_ofileflags[fd] & UF_RESERVED)) {
3716		proc_fdunlock(p);
3717		return (EBADF);
3718	}
3719	if (fp->f_type != DTYPE_PIPE) {
3720		proc_fdunlock(p);
3721		return(EBADF);
3722	}
3723	fp->f_iocount++;
3724
3725	if (resultfp)
3726		*resultfp = fp;
3727	if (resultpipe)
3728		*resultpipe = (struct pipe *)fp->f_data;
3729	proc_fdunlock(p);
3730
3731	return (0);
3732}
3733
3734/*
3735 * fp_lookup
3736 *
3737 * Description:	Get fileproc pointer for a given fd from the per process
3738 *		open file table of the specified process and if successful,
3739 *		increment the f_iocount
3740 *
3741 * Parameters:	p				Process in which fd lives
3742 *		fd				fd to get information for
3743 *		resultfp			Pointer to result fileproc
3744 *						pointer area, or 0 if none
3745 *		locked				!0 if the caller holds the
3746 *						proc_fdlock, 0 otherwise
3747 *
3748 * Returns:	0			Success
3749 *		EBADF			Bad file descriptor
3750 *
3751 * Implicit returns:
3752 *		*resultfp (modified)		Fileproc pointer
3753 *
3754 * Locks:	If the argument 'locked' is non-zero, then the caller is
3755 *		expected to have taken and held the proc_fdlock; if it is
3756 *		zero, than this routine internally takes and drops this lock.
3757 */
3758int
3759fp_lookup(proc_t p, int fd, struct fileproc **resultfp, int locked)
3760{
3761	struct filedesc *fdp = p->p_fd;
3762	struct fileproc *fp;
3763
3764	if (!locked)
3765		proc_fdlock_spin(p);
3766	if (fd < 0 || fdp == NULL || fd >= fdp->fd_nfiles ||
3767			(fp = fdp->fd_ofiles[fd]) == NULL ||
3768			(fdp->fd_ofileflags[fd] & UF_RESERVED)) {
3769		if (!locked)
3770			proc_fdunlock(p);
3771		return (EBADF);
3772	}
3773	fp->f_iocount++;
3774
3775	if (resultfp)
3776		*resultfp = fp;
3777	if (!locked)
3778		proc_fdunlock(p);
3779
3780	return (0);
3781}
3782
3783
3784/*
3785 * fp_tryswap
3786 *
3787 * Description: Swap the fileproc pointer for a given fd with a new
3788 *		fileproc pointer in the per-process open file table of
3789 *		the specified process.  The fdlock must be held at entry.
3790 *
3791 * Parameters:  p		Process containing the fd
3792 *		fd		The fd of interest
3793 *		nfp		Pointer to the newfp
3794 *
3795 * Returns:	0		Success
3796 *		EBADF		Bad file descriptor
3797 *		EINTR		Interrupted
3798 *		EKEEPLOOKING	f_iocount changed while lock was dropped.
3799 */
3800int
3801fp_tryswap(proc_t p, int fd, struct fileproc *nfp)
3802{
3803	struct fileproc *fp;
3804	int error;
3805
3806	proc_fdlock_assert(p, LCK_MTX_ASSERT_OWNED);
3807
3808	if (0 != (error = fp_lookup(p, fd, &fp, 1)))
3809		return (error);
3810	/*
3811	 * At this point, our caller (change_guardedfd_np) has
3812	 * one f_iocount reference, and we just took another
3813	 * one to begin the replacement.
3814	 */
3815	if (fp->f_iocount < 2) {
3816		panic("f_iocount too small %d", fp->f_iocount);
3817	} else if (2 == fp->f_iocount) {
3818
3819		/* Copy the contents of *fp, preserving the "type" of *nfp */
3820
3821		nfp->f_flags = (nfp->f_flags & FP_TYPEMASK) |
3822			(fp->f_flags & ~FP_TYPEMASK);
3823		nfp->f_iocount = fp->f_iocount;
3824		nfp->f_fglob = fp->f_fglob;
3825		nfp->f_waddr = fp->f_waddr;
3826
3827		p->p_fd->fd_ofiles[fd] = nfp;
3828		(void) fp_drop(p, fd, nfp, 1);
3829	} else {
3830		/*
3831		 * Wait for all other active references to evaporate.
3832		 */
3833		p->p_fpdrainwait = 1;
3834		error = msleep(&p->p_fpdrainwait, &p->p_fdmlock,
3835		    PRIBIO | PCATCH, "tryswap fpdrain", NULL);
3836		if (0 == error) {
3837			/*
3838			 * Return an "internal" errno to trigger a full
3839			 * reevaluation of the change-guard attempt.
3840			 */
3841			error = EKEEPLOOKING;
3842			printf("%s: lookup collision fd %d\n", __func__, fd);
3843		}
3844		(void) fp_drop(p, fd, fp, 1);
3845	}
3846	return (error);
3847}
3848
3849
3850/*
3851 * fp_drop_written
3852 *
3853 * Description:	Set the FP_WRITTEN flag on the fileproc and drop the I/O
3854 *		reference previously taken by calling fp_lookup et. al.
3855 *
3856 * Parameters:	p				Process in which the fd lives
3857 *		fd				fd associated with the fileproc
3858 *		fp				fileproc on which to set the
3859 *						flag and drop the reference
3860 *
3861 * Returns:	0				Success
3862 *	fp_drop:EBADF				Bad file descriptor
3863 *
3864 * Locks:	This function internally takes and drops the proc_fdlock for
3865 *		the supplied process
3866 *
3867 * Notes:	The fileproc must correspond to the fd in the supplied proc
3868 */
3869int
3870fp_drop_written(proc_t p, int fd, struct fileproc *fp)
3871{
3872        int error;
3873
3874	proc_fdlock_spin(p);
3875
3876	fp->f_flags |= FP_WRITTEN;
3877
3878	error = fp_drop(p, fd, fp, 1);
3879
3880	proc_fdunlock(p);
3881
3882	return (error);
3883}
3884
3885
3886/*
3887 * fp_drop_event
3888 *
3889 * Description:	Set the FP_WAITEVENT flag on the fileproc and drop the I/O
3890 *		reference previously taken by calling fp_lookup et. al.
3891 *
3892 * Parameters:	p				Process in which the fd lives
3893 *		fd				fd associated with the fileproc
3894 *		fp				fileproc on which to set the
3895 *						flag and drop the reference
3896 *
3897 * Returns:	0				Success
3898 *	fp_drop:EBADF				Bad file descriptor
3899 *
3900 * Locks:	This function internally takes and drops the proc_fdlock for
3901 *		the supplied process
3902 *
3903 * Notes:	The fileproc must correspond to the fd in the supplied proc
3904 */
3905int
3906fp_drop_event(proc_t p, int fd, struct fileproc *fp)
3907{
3908        int error;
3909
3910	proc_fdlock_spin(p);
3911
3912	fp->f_flags |= FP_WAITEVENT;
3913
3914	error = fp_drop(p, fd, fp, 1);
3915
3916	proc_fdunlock(p);
3917
3918	return (error);
3919}
3920
3921
3922/*
3923 * fp_drop
3924 *
3925 * Description:	Drop the I/O reference previously taken by calling fp_lookup
3926 *		et. al.
3927 *
3928 * Parameters:	p				Process in which the fd lives
3929 *		fd				fd associated with the fileproc
3930 *		fp				fileproc on which to set the
3931 *						flag and drop the reference
3932 *		locked				flag to internally take and
3933 *						drop proc_fdlock if it is not
3934 *						already held by the caller
3935 *
3936 * Returns:	0				Success
3937 *		EBADF				Bad file descriptor
3938 *
3939 * Locks:	This function internally takes and drops the proc_fdlock for
3940 *		the supplied process if 'locked' is non-zero, and assumes that
3941 *		the caller already holds this lock if 'locked' is non-zero.
3942 *
3943 * Notes:	The fileproc must correspond to the fd in the supplied proc
3944 */
3945int
3946fp_drop(proc_t p, int fd, struct fileproc *fp, int locked)
3947{
3948	struct filedesc *fdp = p->p_fd;
3949	int	needwakeup = 0;
3950
3951	if (!locked)
3952		proc_fdlock_spin(p);
3953	 if ((fp == FILEPROC_NULL) && (fd < 0 || fd >= fdp->fd_nfiles ||
3954			(fp = fdp->fd_ofiles[fd]) == NULL ||
3955			((fdp->fd_ofileflags[fd] & UF_RESERVED) &&
3956			 !(fdp->fd_ofileflags[fd] & UF_CLOSING)))) {
3957		if (!locked)
3958			proc_fdunlock(p);
3959		return (EBADF);
3960	}
3961	fp->f_iocount--;
3962
3963	if (fp->f_iocount == 0) {
3964		if (fp->f_flags & FP_SELCONFLICT)
3965			fp->f_flags &= ~FP_SELCONFLICT;
3966
3967		if (p->p_fpdrainwait) {
3968			p->p_fpdrainwait = 0;
3969			needwakeup = 1;
3970		}
3971	}
3972	if (!locked)
3973		proc_fdunlock(p);
3974	if (needwakeup)
3975	        wakeup(&p->p_fpdrainwait);
3976
3977	return (0);
3978}
3979
3980
3981/*
3982 * file_vnode
3983 *
3984 * Description:	Given an fd, look it up in the current process's per process
3985 *		open file table, and return its internal vnode pointer.
3986 *
3987 * Parameters:	fd				fd to obtain vnode from
3988 *		vpp				pointer to vnode return area
3989 *
3990 * Returns:	0				Success
3991 *		EINVAL				The fd does not refer to a
3992 *						vnode fileproc entry
3993 *	fp_lookup:EBADF				Bad file descriptor
3994 *
3995 * Implicit returns:
3996 *		*vpp (modified)			Returned vnode pointer
3997 *
3998 * Locks:	This function internally takes and drops the proc_fdlock for
3999 *		the current process
4000 *
4001 * Notes:	If successful, this function increments the f_iocount on the
4002 *		fd's corresponding fileproc.
4003 *
4004 *		The fileproc referenced is not returned; because of this, care
4005 *		must be taken to not drop the last reference (e.g. by closing
4006 *		the file).  This is inherently unsafe, since the reference may
4007 *		not be recoverable from the vnode, if there is a subsequent
4008 *		close that destroys the associate fileproc.  The caller should
4009 *		therefore retain their own reference on the fileproc so that
4010 *		the f_iocount can be dropped subsequently.  Failure to do this
4011 *		can result in the returned pointer immediately becoming invalid
4012 *		following the call.
4013 *
4014 *		Use of this function is discouraged.
4015 */
4016int
4017file_vnode(int fd, struct vnode **vpp)
4018{
4019	proc_t p = current_proc();
4020	struct fileproc *fp;
4021	int error;
4022
4023	proc_fdlock_spin(p);
4024	if ( (error = fp_lookup(p, fd, &fp, 1)) ) {
4025		proc_fdunlock(p);
4026		return(error);
4027	}
4028	if (fp->f_type != DTYPE_VNODE) {
4029		fp_drop(p, fd, fp,1);
4030		proc_fdunlock(p);
4031		return(EINVAL);
4032	}
4033	if (vpp != NULL)
4034		*vpp = (struct vnode *)fp->f_data;
4035	proc_fdunlock(p);
4036
4037	return(0);
4038}
4039
4040
4041/*
4042 * file_vnode_withvid
4043 *
4044 * Description:	Given an fd, look it up in the current process's per process
4045 *		open file table, and return its internal vnode pointer.
4046 *
4047 * Parameters:	fd				fd to obtain vnode from
4048 *		vpp				pointer to vnode return area
4049 *		vidp				pointer to vid of the returned vnode
4050 *
4051 * Returns:	0				Success
4052 *		EINVAL				The fd does not refer to a
4053 *						vnode fileproc entry
4054 *	fp_lookup:EBADF				Bad file descriptor
4055 *
4056 * Implicit returns:
4057 *		*vpp (modified)			Returned vnode pointer
4058 *
4059 * Locks:	This function internally takes and drops the proc_fdlock for
4060 *		the current process
4061 *
4062 * Notes:	If successful, this function increments the f_iocount on the
4063 *		fd's corresponding fileproc.
4064 *
4065 *		The fileproc referenced is not returned; because of this, care
4066 *		must be taken to not drop the last reference (e.g. by closing
4067 *		the file).  This is inherently unsafe, since the reference may
4068 *		not be recoverable from the vnode, if there is a subsequent
4069 *		close that destroys the associate fileproc.  The caller should
4070 *		therefore retain their own reference on the fileproc so that
4071 *		the f_iocount can be dropped subsequently.  Failure to do this
4072 *		can result in the returned pointer immediately becoming invalid
4073 *		following the call.
4074 *
4075 *		Use of this function is discouraged.
4076 */
4077int
4078file_vnode_withvid(int fd, struct vnode **vpp, uint32_t * vidp)
4079{
4080	proc_t p = current_proc();
4081	struct fileproc *fp;
4082	vnode_t vp;
4083	int error;
4084
4085	proc_fdlock_spin(p);
4086	if ( (error = fp_lookup(p, fd, &fp, 1)) ) {
4087		proc_fdunlock(p);
4088		return(error);
4089	}
4090	if (fp->f_type != DTYPE_VNODE) {
4091		fp_drop(p, fd, fp,1);
4092		proc_fdunlock(p);
4093		return(EINVAL);
4094	}
4095	vp = (struct vnode *)fp->f_data;
4096	if (vpp != NULL)
4097		*vpp = vp;
4098
4099	if ((vidp != NULL) && (vp != NULLVP))
4100		*vidp = (uint32_t)vp->v_id;
4101
4102	proc_fdunlock(p);
4103
4104	return(0);
4105}
4106
4107
4108/*
4109 * file_socket
4110 *
4111 * Description:	Given an fd, look it up in the current process's per process
4112 *		open file table, and return its internal socket pointer.
4113 *
4114 * Parameters:	fd				fd to obtain vnode from
4115 *		sp				pointer to socket return area
4116 *
4117 * Returns:	0				Success
4118 *		ENOTSOCK			Not a socket
4119 *		fp_lookup:EBADF			Bad file descriptor
4120 *
4121 * Implicit returns:
4122 *		*sp (modified)			Returned socket pointer
4123 *
4124 * Locks:	This function internally takes and drops the proc_fdlock for
4125 *		the current process
4126 *
4127 * Notes:	If successful, this function increments the f_iocount on the
4128 *		fd's corresponding fileproc.
4129 *
4130 *		The fileproc referenced is not returned; because of this, care
4131 *		must be taken to not drop the last reference (e.g. by closing
4132 *		the file).  This is inherently unsafe, since the reference may
4133 *		not be recoverable from the socket, if there is a subsequent
4134 *		close that destroys the associate fileproc.  The caller should
4135 *		therefore retain their own reference on the fileproc so that
4136 *		the f_iocount can be dropped subsequently.  Failure to do this
4137 *		can result in the returned pointer immediately becoming invalid
4138 *		following the call.
4139 *
4140 *		Use of this function is discouraged.
4141 */
4142int
4143file_socket(int fd, struct socket **sp)
4144{
4145	proc_t p = current_proc();
4146	struct fileproc *fp;
4147	int error;
4148
4149	proc_fdlock_spin(p);
4150	if ( (error = fp_lookup(p, fd, &fp, 1)) ) {
4151		proc_fdunlock(p);
4152		return(error);
4153	}
4154	if (fp->f_type != DTYPE_SOCKET) {
4155		fp_drop(p, fd, fp,1);
4156		proc_fdunlock(p);
4157		return(ENOTSOCK);
4158	}
4159	*sp = (struct socket *)fp->f_data;
4160	proc_fdunlock(p);
4161
4162	return(0);
4163}
4164
4165
4166/*
4167 * file_flags
4168 *
4169 * Description:	Given an fd, look it up in the current process's per process
4170 *		open file table, and return its fileproc's flags field.
4171 *
4172 * Parameters:	fd				fd whose flags are to be
4173 *						retrieved
4174 *		flags				pointer to flags data area
4175 *
4176 * Returns:	0				Success
4177 *		ENOTSOCK			Not a socket
4178 *		fp_lookup:EBADF			Bad file descriptor
4179 *
4180 * Implicit returns:
4181 *		*flags (modified)		Returned flags field
4182 *
4183 * Locks:	This function internally takes and drops the proc_fdlock for
4184 *		the current process
4185 *
4186 * Notes:	This function will internally increment and decrement the
4187 *		f_iocount of the fileproc as part of its operation.
4188 */
4189int
4190file_flags(int fd, int *flags)
4191{
4192
4193	proc_t p = current_proc();
4194	struct fileproc *fp;
4195	int error;
4196
4197	proc_fdlock_spin(p);
4198	if ( (error = fp_lookup(p, fd, &fp, 1)) ) {
4199		proc_fdunlock(p);
4200		return(error);
4201	}
4202	*flags = (int)fp->f_flag;
4203	fp_drop(p, fd, fp,1);
4204	proc_fdunlock(p);
4205
4206	return(0);
4207}
4208
4209
4210/*
4211 * file_drop
4212 *
4213 * Description:	Drop an iocount reference on an fd, and wake up any waiters
4214 *		for draining (i.e. blocked in fileproc_drain() called during
4215 *		the last attempt to close a file).
4216 *
4217 * Parameters:	fd				fd on which an ioreference is
4218 *						to be dropped
4219 *
4220 * Returns:	0				Success
4221 *		EBADF				Bad file descriptor
4222 *
4223 * Description:	Given an fd, look it up in the current process's per process
4224 *		open file table, and drop it's fileproc's f_iocount by one
4225 *
4226 * Notes:	This is intended as a corresponding operation to the functions
4227 *		file_vnode() and file_socket() operations.
4228 *
4229 *		Technically, the close reference is supposed to be protected
4230 *		by a fileproc_drain(), however, a drain will only block if
4231 *		the fd refers to a character device, and that device has had
4232 *		preparefileread() called on it.  If it refers to something
4233 *		other than a character device, then the drain will occur and
4234 *		block each close attempt, rather than merely the last close.
4235 *
4236 *		Since it's possible for an fd that refers to a character
4237 *		device to have an intermediate close followed by an open to
4238 *		cause a different file to correspond to that descriptor,
4239 *		unless there was a cautionary reference taken on the fileproc,
4240 *		this is an inherently unsafe function.  This happens in the
4241 *		case where multiple fd's in a process refer to the same
4242 *		character device (e.g. stdin/out/err pointing to a tty, etc.).
4243 *
4244 *		Use of this function is discouraged.
4245 */
4246int
4247file_drop(int fd)
4248{
4249	struct fileproc *fp;
4250	proc_t p = current_proc();
4251	int	needwakeup = 0;
4252
4253	proc_fdlock_spin(p);
4254	if (fd < 0 || fd >= p->p_fd->fd_nfiles ||
4255			(fp = p->p_fd->fd_ofiles[fd]) == NULL ||
4256			((p->p_fd->fd_ofileflags[fd] & UF_RESERVED) &&
4257			 !(p->p_fd->fd_ofileflags[fd] & UF_CLOSING))) {
4258		proc_fdunlock(p);
4259		return (EBADF);
4260	}
4261	fp->f_iocount --;
4262
4263	if (fp->f_iocount == 0) {
4264		if (fp->f_flags & FP_SELCONFLICT)
4265			fp->f_flags &= ~FP_SELCONFLICT;
4266
4267		if (p->p_fpdrainwait) {
4268			p->p_fpdrainwait = 0;
4269			needwakeup = 1;
4270		}
4271	}
4272	proc_fdunlock(p);
4273
4274	if (needwakeup)
4275	        wakeup(&p->p_fpdrainwait);
4276	return(0);
4277}
4278
4279
4280static int falloc_withalloc_locked(proc_t, struct fileproc **, int *,
4281    vfs_context_t, struct fileproc * (*)(void *), void *, int);
4282
4283/*
4284 * falloc
4285 *
4286 * Description:	Allocate an entry in the per process open file table and
4287 *		return the corresponding fileproc and fd.
4288 *
4289 * Parameters:	p				The process in whose open file
4290 *						table the fd is to be allocated
4291 *		resultfp			Pointer to fileproc pointer
4292 *						return area
4293 *		resultfd			Pointer to fd return area
4294 *		ctx				VFS context
4295 *
4296 * Returns:	0				Success
4297 *	falloc:ENFILE				Too many open files in system
4298 *	falloc:EMFILE				Too many open files in process
4299 *	falloc:ENOMEM				M_FILEPROC or M_FILEGLOB zone
4300 *						exhausted
4301 *
4302 * Implicit returns:
4303 *		*resultfd (modified)		Returned fileproc pointer
4304 *		*resultfd (modified)		Returned fd
4305 *
4306 * Locks:	This function takes and drops the proc_fdlock; if this lock
4307 *		is already held, use falloc_locked() instead.
4308 *
4309 * Notes:	This function takes separate process and context arguments
4310 *		solely to support kern_exec.c; otherwise, it would take
4311 *		neither, and expect falloc_locked() to use the
4312 *		vfs_context_current() routine internally.
4313 */
4314int
4315falloc(proc_t p, struct fileproc **resultfp, int *resultfd, vfs_context_t ctx)
4316{
4317	return (falloc_withalloc(p, resultfp, resultfd, ctx,
4318	    fileproc_alloc_init, NULL));
4319}
4320
4321/*
4322 * Like falloc, but including the fileproc allocator and create-args
4323 */
4324int
4325falloc_withalloc(proc_t p, struct fileproc **resultfp, int *resultfd,
4326    vfs_context_t ctx, fp_allocfn_t fp_zalloc, void *arg)
4327{
4328	int error;
4329
4330	proc_fdlock(p);
4331	error = falloc_withalloc_locked(p,
4332	    resultfp, resultfd, ctx, fp_zalloc, arg, 1);
4333	proc_fdunlock(p);
4334
4335	return (error);
4336}
4337
4338/*
4339 * "uninitialized" ops -- ensure fg->fg_ops->fo_type always exists
4340 */
4341static const struct fileops uninitops;
4342
4343/*
4344 * falloc_locked
4345 *
4346 * Create a new open file structure and allocate
4347 * a file descriptor for the process that refers to it.
4348 *
4349 * Returns:	0			Success
4350 *
4351 * Description:	Allocate an entry in the per process open file table and
4352 *		return the corresponding fileproc and fd.
4353 *
4354 * Parameters:	p				The process in whose open file
4355 *						table the fd is to be allocated
4356 *		resultfp			Pointer to fileproc pointer
4357 *						return area
4358 *		resultfd			Pointer to fd return area
4359 *		ctx				VFS context
4360 *		locked				Flag to indicate whether the
4361 *						caller holds proc_fdlock
4362 *
4363 * Returns:	0				Success
4364 *		ENFILE				Too many open files in system
4365 *		fdalloc:EMFILE			Too many open files in process
4366 *		ENOMEM				M_FILEPROC or M_FILEGLOB zone
4367 *						exhausted
4368 *	fdalloc:ENOMEM
4369 *
4370 * Implicit returns:
4371 *		*resultfd (modified)		Returned fileproc pointer
4372 *		*resultfd (modified)		Returned fd
4373 *
4374 * Locks:	If the parameter 'locked' is zero, this function takes and
4375 *		drops the proc_fdlock; if non-zero, the caller must hold the
4376 *		lock.
4377 *
4378 * Notes:	If you intend to use a non-zero 'locked' parameter, use the
4379 *		utility function falloc() instead.
4380 *
4381 *		This function takes separate process and context arguments
4382 *		solely to support kern_exec.c; otherwise, it would take
4383 *		neither, and use the vfs_context_current() routine internally.
4384 */
4385int
4386falloc_locked(proc_t p, struct fileproc **resultfp, int *resultfd,
4387	      vfs_context_t ctx, int locked)
4388{
4389	return (falloc_withalloc_locked(p, resultfp, resultfd, ctx,
4390	    fileproc_alloc_init, NULL, locked));
4391}
4392
4393static int
4394falloc_withalloc_locked(proc_t p, struct fileproc **resultfp, int *resultfd,
4395    vfs_context_t ctx, fp_allocfn_t fp_zalloc, void *crarg,
4396    int locked)
4397{
4398	struct fileproc *fp;
4399	struct fileglob *fg;
4400	int error, nfd;
4401
4402	if (!locked)
4403		proc_fdlock(p);
4404	if ( (error = fdalloc(p, 0, &nfd)) ) {
4405		if (!locked)
4406			proc_fdunlock(p);
4407		return (error);
4408	}
4409	if (nfiles >= maxfiles) {
4410		if (!locked)
4411			proc_fdunlock(p);
4412		tablefull("file");
4413		return (ENFILE);
4414	}
4415#if CONFIG_MACF
4416	error = mac_file_check_create(proc_ucred(p));
4417	if (error) {
4418		if (!locked)
4419			proc_fdunlock(p);
4420		return (error);
4421	}
4422#endif
4423
4424	/*
4425	 * Allocate a new file descriptor.
4426	 * If the process has file descriptor zero open, add to the list
4427	 * of open files at that point, otherwise put it at the front of
4428	 * the list of open files.
4429	 */
4430	proc_fdunlock(p);
4431
4432	fp = (*fp_zalloc)(crarg);
4433	if (fp == NULL) {
4434		if (locked)
4435			proc_fdlock(p);
4436		return (ENOMEM);
4437	}
4438	MALLOC_ZONE(fg, struct fileglob *, sizeof(struct fileglob), M_FILEGLOB, M_WAITOK);
4439	if (fg == NULL) {
4440		fileproc_free(fp);
4441		if (locked)
4442			proc_fdlock(p);
4443		return (ENOMEM);
4444	}
4445	bzero(fg, sizeof(struct fileglob));
4446	lck_mtx_init(&fg->fg_lock, file_lck_grp, file_lck_attr);
4447
4448	fp->f_iocount = 1;
4449	fg->fg_count = 1;
4450	fg->fg_ops = &uninitops;
4451	fp->f_fglob = fg;
4452#if CONFIG_MACF
4453	mac_file_label_init(fg);
4454#endif
4455
4456	kauth_cred_ref(ctx->vc_ucred);
4457
4458	proc_fdlock(p);
4459
4460	fp->f_cred = ctx->vc_ucred;
4461
4462#if CONFIG_MACF
4463	mac_file_label_associate(fp->f_cred, fg);
4464#endif
4465
4466	OSAddAtomic(1, &nfiles);
4467
4468	p->p_fd->fd_ofiles[nfd] = fp;
4469
4470	if (!locked)
4471		proc_fdunlock(p);
4472
4473	if (resultfp)
4474		*resultfp = fp;
4475	if (resultfd)
4476		*resultfd = nfd;
4477
4478	return (0);
4479}
4480
4481
4482/*
4483 * fg_free
4484 *
4485 * Description:	Free a file structure; drop the global open file count, and
4486 *		drop the credential reference, if the fileglob has one, and
4487 *		destroy the instance mutex before freeing
4488 *
4489 * Parameters:	fg				Pointer to fileglob to be
4490 *						freed
4491 *
4492 * Returns:	void
4493 */
4494void
4495fg_free(struct fileglob *fg)
4496{
4497	OSAddAtomic(-1, &nfiles);
4498
4499	if (fg->fg_vn_data) {
4500		fg_vn_data_free(fg->fg_vn_data);
4501		fg->fg_vn_data = NULL;
4502	}
4503
4504	if (IS_VALID_CRED(fg->fg_cred)) {
4505		kauth_cred_unref(&fg->fg_cred);
4506	}
4507	lck_mtx_destroy(&fg->fg_lock, file_lck_grp);
4508
4509#if CONFIG_MACF
4510	mac_file_label_destroy(fg);
4511#endif
4512	FREE_ZONE(fg, sizeof *fg, M_FILEGLOB);
4513}
4514
4515
4516/*
4517 * fdexec
4518 *
4519 * Description:	Perform close-on-exec processing for all files in a process
4520 *		that are either marked as close-on-exec, or which were in the
4521 *		process of being opened at the time of the execve
4522 *
4523 *		Also handles the case (via posix_spawn()) where -all-
4524 *		files except those marked with "inherit" as treated as
4525 *		close-on-exec.
4526 *
4527 * Parameters:	p				Pointer to process calling
4528 *						execve
4529 *
4530 * Returns:	void
4531 *
4532 * Locks:	This function internally takes and drops proc_fdlock()
4533 *
4534 */
4535void
4536fdexec(proc_t p, short flags)
4537{
4538	struct filedesc *fdp = p->p_fd;
4539	int i;
4540	boolean_t cloexec_default = (flags & POSIX_SPAWN_CLOEXEC_DEFAULT) != 0;
4541
4542	proc_fdlock(p);
4543	for (i = fdp->fd_lastfile; i >= 0; i--) {
4544
4545		struct fileproc *fp = fdp->fd_ofiles[i];
4546		char *flagp = &fdp->fd_ofileflags[i];
4547
4548		if (fp && cloexec_default) {
4549			/*
4550			 * Reverse the usual semantics of file descriptor
4551			 * inheritance - all of them should be closed
4552			 * except files marked explicitly as "inherit" and
4553			 * not marked close-on-exec.
4554			 */
4555			if ((*flagp & (UF_EXCLOSE|UF_INHERIT)) != UF_INHERIT)
4556				*flagp |= UF_EXCLOSE;
4557			*flagp &= ~UF_INHERIT;
4558		}
4559
4560		if (
4561		    ((*flagp & (UF_RESERVED|UF_EXCLOSE)) == UF_EXCLOSE)
4562#if CONFIG_MACF
4563		    || (fp && mac_file_check_inherit(proc_ucred(p), fp->f_fglob))
4564#endif
4565		) {
4566                        if (i < fdp->fd_knlistsize)
4567                                knote_fdclose(p, i);
4568			procfdtbl_clearfd(p, i);
4569			if (i == fdp->fd_lastfile && i > 0)
4570				fdp->fd_lastfile--;
4571			if (i < fdp->fd_freefile)
4572				fdp->fd_freefile = i;
4573
4574			/*
4575			 * Wait for any third party viewers (e.g., lsof)
4576			 * to release their references to this fileproc.
4577			 */
4578			while (fp->f_iocount > 0) {
4579				p->p_fpdrainwait = 1;
4580				msleep(&p->p_fpdrainwait, &p->p_fdmlock, PRIBIO,
4581				    "fpdrain", NULL);
4582			}
4583
4584			closef_locked(fp, fp->f_fglob, p);
4585
4586			fileproc_free(fp);
4587		}
4588	}
4589	proc_fdunlock(p);
4590}
4591
4592
4593/*
4594 * fdcopy
4595 *
4596 * Description:	Copy a filedesc structure.  This is normally used as part of
4597 *		forkproc() when forking a new process, to copy the per process
4598 *		open file table over to the new process.
4599 *
4600 * Parameters:	p				Process whose open file table
4601 *						is to be copied (parent)
4602 *		uth_cdir			Per thread current working
4603 *						cirectory, or NULL
4604 *
4605 * Returns:	NULL				Copy failed
4606 *		!NULL				Pointer to new struct filedesc
4607 *
4608 * Locks:	This function internally takes and drops proc_fdlock()
4609 *
4610 * Notes:	Files are copied directly, ignoring the new resource limits
4611 *		for the process that's being copied into.  Since the descriptor
4612 *		references are just additional references, this does not count
4613 *		against the number of open files on the system.
4614 *
4615 *		The struct filedesc includes the current working directory,
4616 *		and the current root directory, if the process is chroot'ed.
4617 *
4618 *		If the exec was called by a thread using a per thread current
4619 *		working directory, we inherit the working directory from the
4620 *		thread making the call, rather than from the process.
4621 *
4622 *		In the case of a failure to obtain a reference, for most cases,
4623 *		the file entry will be silently dropped.  There's an exception
4624 *		for the case of a chroot dir, since a failure to to obtain a
4625 *		reference there would constitute an "escape" from the chroot
4626 *		environment, which must not be allowed.  In that case, we will
4627 *		deny the execve() operation, rather than allowing the escape.
4628 */
4629struct filedesc *
4630fdcopy(proc_t p, vnode_t uth_cdir)
4631{
4632	struct filedesc *newfdp, *fdp = p->p_fd;
4633	int i;
4634	struct fileproc *ofp, *fp;
4635	vnode_t	v_dir;
4636
4637	MALLOC_ZONE(newfdp, struct filedesc *,
4638			sizeof(*newfdp), M_FILEDESC, M_WAITOK);
4639	if (newfdp == NULL)
4640		return(NULL);
4641
4642	proc_fdlock(p);
4643
4644	/*
4645	 * the FD_CHROOT flag will be inherited via this copy
4646	 */
4647	(void) memcpy(newfdp, fdp, sizeof(*newfdp));
4648
4649	/*
4650	 * If we are running with per-thread current working directories,
4651	 * inherit the new current working directory from the current thread
4652	 * instead, before we take our references.
4653	 */
4654	if (uth_cdir != NULLVP)
4655		newfdp->fd_cdir = uth_cdir;
4656
4657	/*
4658	 * For both fd_cdir and fd_rdir make sure we get
4659	 * a valid reference... if we can't, than set
4660	 * set the pointer(s) to NULL in the child... this
4661	 * will keep us from using a non-referenced vp
4662	 * and allows us to do the vnode_rele only on
4663	 * a properly referenced vp
4664	 */
4665	if ( (v_dir = newfdp->fd_cdir) ) {
4666	        if (vnode_getwithref(v_dir) == 0) {
4667		        if ( (vnode_ref(v_dir)) )
4668			        newfdp->fd_cdir = NULL;
4669			vnode_put(v_dir);
4670		} else
4671		        newfdp->fd_cdir = NULL;
4672	}
4673	if (newfdp->fd_cdir == NULL && fdp->fd_cdir) {
4674	        /*
4675		 * we couldn't get a new reference on
4676		 * the current working directory being
4677		 * inherited... we might as well drop
4678		 * our reference from the parent also
4679		 * since the vnode has gone DEAD making
4680		 * it useless... by dropping it we'll
4681		 * be that much closer to recycling it
4682		 */
4683	        vnode_rele(fdp->fd_cdir);
4684		fdp->fd_cdir = NULL;
4685	}
4686
4687	if ( (v_dir = newfdp->fd_rdir) ) {
4688		if (vnode_getwithref(v_dir) == 0) {
4689			if ( (vnode_ref(v_dir)) )
4690			        newfdp->fd_rdir = NULL;
4691			vnode_put(v_dir);
4692		} else {
4693		        newfdp->fd_rdir = NULL;
4694		}
4695	}
4696	/* Coming from a chroot environment and unable to get a reference... */
4697	if (newfdp->fd_rdir == NULL && fdp->fd_rdir) {
4698	        /*
4699		 * We couldn't get a new reference on
4700		 * the chroot directory being
4701		 * inherited... this is fatal, since
4702		 * otherwise it would constitute an
4703		 * escape from a chroot environment by
4704		 * the new process.
4705		 */
4706		if (newfdp->fd_cdir)
4707		        vnode_rele(newfdp->fd_cdir);
4708		FREE_ZONE(newfdp, sizeof *newfdp, M_FILEDESC);
4709		return(NULL);
4710	}
4711
4712	/*
4713	 * If the number of open files fits in the internal arrays
4714	 * of the open file structure, use them, otherwise allocate
4715	 * additional memory for the number of descriptors currently
4716	 * in use.
4717	 */
4718	if (newfdp->fd_lastfile < NDFILE)
4719		i = NDFILE;
4720	else {
4721		/*
4722		 * Compute the smallest multiple of NDEXTENT needed
4723		 * for the file descriptors currently in use,
4724		 * allowing the table to shrink.
4725		 */
4726		i = newfdp->fd_nfiles;
4727		while (i > 2 * NDEXTENT && i > newfdp->fd_lastfile * 2)
4728			i /= 2;
4729	}
4730	proc_fdunlock(p);
4731
4732	MALLOC_ZONE(newfdp->fd_ofiles, struct fileproc **,
4733				i * OFILESIZE, M_OFILETABL, M_WAITOK);
4734	if (newfdp->fd_ofiles == NULL) {
4735		if (newfdp->fd_cdir)
4736		        vnode_rele(newfdp->fd_cdir);
4737		if (newfdp->fd_rdir)
4738			vnode_rele(newfdp->fd_rdir);
4739
4740		FREE_ZONE(newfdp, sizeof(*newfdp), M_FILEDESC);
4741		return(NULL);
4742	}
4743	(void) memset(newfdp->fd_ofiles, 0, i * OFILESIZE);
4744	proc_fdlock(p);
4745
4746	newfdp->fd_ofileflags = (char *) &newfdp->fd_ofiles[i];
4747	newfdp->fd_nfiles = i;
4748
4749	if (fdp->fd_nfiles > 0) {
4750		struct fileproc **fpp;
4751		char *flags;
4752
4753		(void) memcpy(newfdp->fd_ofiles, fdp->fd_ofiles,
4754					(newfdp->fd_lastfile + 1) * sizeof(*fdp->fd_ofiles));
4755		(void) memcpy(newfdp->fd_ofileflags, fdp->fd_ofileflags,
4756					(newfdp->fd_lastfile + 1) * sizeof(*fdp->fd_ofileflags));
4757
4758		/*
4759		 * kq descriptors cannot be copied.
4760		 */
4761		if (newfdp->fd_knlistsize != -1) {
4762			fpp = &newfdp->fd_ofiles[newfdp->fd_lastfile];
4763			flags = &newfdp->fd_ofileflags[newfdp->fd_lastfile];
4764			for (i = newfdp->fd_lastfile;
4765			    i >= 0; i--, fpp--, flags--) {
4766				if (*flags & UF_RESERVED)
4767					continue;	/* (removed below) */
4768				if (*fpp != NULL && (*fpp)->f_type == DTYPE_KQUEUE) {
4769					*fpp = NULL;
4770					*flags = 0;
4771					if (i < newfdp->fd_freefile)
4772						newfdp->fd_freefile = i;
4773				}
4774				if (*fpp == NULL && i == newfdp->fd_lastfile && i > 0)
4775					newfdp->fd_lastfile--;
4776			}
4777			newfdp->fd_knlist = NULL;
4778			newfdp->fd_knlistsize = -1;
4779			newfdp->fd_knhash = NULL;
4780			newfdp->fd_knhashmask = 0;
4781		}
4782		fpp = newfdp->fd_ofiles;
4783		flags = newfdp->fd_ofileflags;
4784
4785		for (i = newfdp->fd_lastfile + 1; --i >= 0; fpp++, flags++)
4786			if ((ofp = *fpp) != NULL &&
4787			    0 == (*flags & (UF_FORKCLOSE|UF_RESERVED))) {
4788#if DEBUG
4789				if (FILEPROC_TYPE(ofp) != FTYPE_SIMPLE)
4790					panic("complex fileproc");
4791#endif
4792				fp = fileproc_alloc_init(NULL);
4793				if (fp == NULL) {
4794					/*
4795					 * XXX no room to copy, unable to
4796					 * XXX safely unwind state at present
4797					 */
4798					*fpp = NULL;
4799				} else {
4800					fp->f_flags |=
4801					    (ofp->f_flags & ~FP_TYPEMASK);
4802					fp->f_fglob = ofp->f_fglob;
4803					(void)fg_ref(fp);
4804					*fpp = fp;
4805				}
4806			} else {
4807				if (i < newfdp->fd_freefile)
4808					newfdp->fd_freefile = i;
4809				*fpp = NULL;
4810				*flags = 0;
4811			}
4812	}
4813
4814	proc_fdunlock(p);
4815	return (newfdp);
4816}
4817
4818
4819/*
4820 * fdfree
4821 *
4822 * Description:	Release a filedesc (per process open file table) structure;
4823 *		this is done on process exit(), or from forkproc_free() if
4824 *		the fork fails for some reason subsequent to a successful
4825 *		call to fdcopy()
4826 *
4827 * Parameters:	p				Pointer to process going away
4828 *
4829 * Returns:	void
4830 *
4831 * Locks:	This function internally takes and drops proc_fdlock()
4832 */
4833void
4834fdfree(proc_t p)
4835{
4836	struct filedesc *fdp;
4837	struct fileproc *fp;
4838	int i;
4839
4840	proc_fdlock(p);
4841
4842	if (p == kernproc || NULL == (fdp = p->p_fd)) {
4843	        proc_fdunlock(p);
4844		return;
4845	}
4846
4847	extern struct filedesc filedesc0;
4848
4849	if (&filedesc0 == fdp)
4850		panic("filedesc0");
4851
4852	if (fdp->fd_nfiles > 0 && fdp->fd_ofiles) {
4853	        for (i = fdp->fd_lastfile; i >= 0; i--) {
4854			if ((fp = fdp->fd_ofiles[i]) != NULL) {
4855
4856			  if (fdp->fd_ofileflags[i] & UF_RESERVED)
4857			    	panic("fdfree: found fp with UF_RESERVED");
4858
4859				procfdtbl_reservefd(p, i);
4860
4861				if (i < fdp->fd_knlistsize)
4862					knote_fdclose(p, i);
4863				if (fp->f_flags & FP_WAITEVENT)
4864					(void)waitevent_close(p, fp);
4865				(void) closef_locked(fp, fp->f_fglob, p);
4866				fileproc_free(fp);
4867			}
4868		}
4869		FREE_ZONE(fdp->fd_ofiles, fdp->fd_nfiles * OFILESIZE, M_OFILETABL);
4870		fdp->fd_ofiles = NULL;
4871		fdp->fd_nfiles = 0;
4872	}
4873
4874	proc_fdunlock(p);
4875
4876	if (fdp->fd_cdir)
4877	        vnode_rele(fdp->fd_cdir);
4878	if (fdp->fd_rdir)
4879		vnode_rele(fdp->fd_rdir);
4880
4881	proc_fdlock_spin(p);
4882	p->p_fd = NULL;
4883	proc_fdunlock(p);
4884
4885	if (fdp->fd_knlist)
4886		FREE(fdp->fd_knlist, M_KQUEUE);
4887	if (fdp->fd_knhash)
4888		FREE(fdp->fd_knhash, M_KQUEUE);
4889
4890	FREE_ZONE(fdp, sizeof(*fdp), M_FILEDESC);
4891}
4892
4893/*
4894 * closef_locked
4895 *
4896 * Description:	Internal form of closef; called with proc_fdlock held
4897 *
4898 * Parameters:	fp			Pointer to fileproc for fd
4899 *		fg			Pointer to fileglob for fd
4900 *		p			Pointer to proc structure
4901 *
4902 * Returns:	0			Success
4903 *	closef_finish:??? 		Anything returnable by a per-fileops
4904 *					close function
4905 *
4906 * Note:	Decrements reference count on file structure; if this was the
4907 *		last reference, then closef_finish() is called
4908 *
4909 *		p and fp are allowed to  be NULL when closing a file that was
4910 *		being passed in a message (but only if we are called when this
4911 *		is NOT the last reference).
4912 */
4913int
4914closef_locked(struct fileproc *fp, struct fileglob *fg, proc_t p)
4915{
4916	struct vnode *vp;
4917	struct flock lf;
4918	struct vfs_context context;
4919	int error;
4920
4921	if (fg == NULL) {
4922		return (0);
4923	}
4924
4925	/* Set up context with cred stashed in fg */
4926	if (p == current_proc())
4927		context.vc_thread = current_thread();
4928	else
4929		context.vc_thread = NULL;
4930	context.vc_ucred = fg->fg_cred;
4931
4932	/*
4933	 * POSIX record locking dictates that any close releases ALL
4934	 * locks owned by this process.  This is handled by setting
4935	 * a flag in the unlock to free ONLY locks obeying POSIX
4936	 * semantics, and not to free BSD-style file locks.
4937	 * If the descriptor was in a message, POSIX-style locks
4938	 * aren't passed with the descriptor.
4939	 */
4940	if (p && (p->p_ladvflag & P_LADVLOCK) &&
4941	    DTYPE_VNODE == FILEGLOB_DTYPE(fg)) {
4942		proc_fdunlock(p);
4943
4944		lf.l_whence = SEEK_SET;
4945		lf.l_start = 0;
4946		lf.l_len = 0;
4947		lf.l_type = F_UNLCK;
4948		vp = (struct vnode *)fg->fg_data;
4949
4950		if ( (error = vnode_getwithref(vp)) == 0 ) {
4951			(void) VNOP_ADVLOCK(vp, (caddr_t)p, F_UNLCK, &lf, F_POSIX, &context, NULL);
4952			(void)vnode_put(vp);
4953		}
4954		proc_fdlock(p);
4955	}
4956	lck_mtx_lock_spin(&fg->fg_lock);
4957	fg->fg_count--;
4958
4959	if (fg->fg_count > 0) {
4960		lck_mtx_unlock(&fg->fg_lock);
4961		return (0);
4962	}
4963#if DIAGNOSTIC
4964	if (fg->fg_count != 0)
4965		panic("fg %p: being freed with bad fg_count (%d)", fg, fg->fg_count);
4966#endif
4967
4968	if (fp && (fp->f_flags & FP_WRITTEN))
4969	        fg->fg_flag |= FWASWRITTEN;
4970
4971	fg->fg_lflags |= FG_TERM;
4972	lck_mtx_unlock(&fg->fg_lock);
4973
4974	if (p)
4975		proc_fdunlock(p);
4976
4977	/* Since we ensure that fg->fg_ops is always initialized,
4978	 * it is safe to invoke fo_close on the fg */
4979	error = fo_close(fg, &context);
4980
4981	fg_free(fg);
4982
4983	if (p)
4984		proc_fdlock(p);
4985
4986	return(error);
4987}
4988
4989
4990/*
4991 * fileproc_drain
4992 *
4993 * Description:	Drain out pending I/O operations
4994 *
4995 * Parameters:	p				Process closing this file
4996 *		fp				fileproc struct for the open
4997 *						instance on the file
4998 *
4999 * Returns:	void
5000 *
5001 * Locks:	Assumes the caller holds the proc_fdlock
5002 *
5003 * Notes:	For character devices, this occurs on the last close of the
5004 *		device; for all other file descriptors, this occurs on each
5005 *		close to prevent fd's from being closed out from under
5006 *		operations currently in progress and blocked
5007 *
5008 * See Also: 	file_vnode(), file_socket(), file_drop(), and the cautions
5009 *		regarding their use and interaction with this function.
5010 */
5011void
5012fileproc_drain(proc_t p, struct fileproc * fp)
5013{
5014	struct vfs_context context;
5015
5016	context.vc_thread = proc_thread(p);	/* XXX */
5017	context.vc_ucred = fp->f_fglob->fg_cred;
5018
5019	fp->f_iocount-- ; /* (the one the close holds) */
5020
5021	while (fp->f_iocount) {
5022
5023	        lck_mtx_convert_spin(&p->p_fdmlock);
5024
5025		if (fp->f_fglob->fg_ops->fo_drain) {
5026			(*fp->f_fglob->fg_ops->fo_drain)(fp, &context);
5027		}
5028		if ((fp->f_flags & FP_INSELECT) == FP_INSELECT) {
5029			if (wait_queue_wakeup_all((wait_queue_t)fp->f_waddr, NULL, THREAD_INTERRUPTED) == KERN_INVALID_ARGUMENT)
5030				panic("bad wait queue for wait_queue_wakeup_all %p", fp->f_waddr);
5031		}
5032		if ((fp->f_flags & FP_SELCONFLICT) == FP_SELCONFLICT) {
5033			if (wait_queue_wakeup_all(&select_conflict_queue, NULL, THREAD_INTERRUPTED) == KERN_INVALID_ARGUMENT)
5034				panic("bad select_conflict_queue");
5035		}
5036		p->p_fpdrainwait = 1;
5037
5038		msleep(&p->p_fpdrainwait, &p->p_fdmlock, PRIBIO, "fpdrain", NULL);
5039
5040	}
5041#if DIAGNOSTIC
5042	if ((fp->f_flags & FP_INSELECT) != 0)
5043		panic("FP_INSELECT set on drained fp");
5044#endif
5045	if ((fp->f_flags & FP_SELCONFLICT) == FP_SELCONFLICT)
5046		fp->f_flags &= ~FP_SELCONFLICT;
5047}
5048
5049
5050/*
5051 * fp_free
5052 *
5053 * Description:	Release the fd and free the fileproc associated with the fd
5054 *		in the per process open file table of the specified process;
5055 *		these values must correspond.
5056 *
5057 * Parameters:	p				Process containing fd
5058 *		fd				fd to be released
5059 *		fp				fileproc to be freed
5060 *
5061 * Returns:	0				Success
5062 *
5063 * Notes:	XXX function should be void - no one interprets the returns
5064 *		XXX code
5065 */
5066int
5067fp_free(proc_t p, int fd, struct fileproc * fp)
5068{
5069        proc_fdlock_spin(p);
5070	fdrelse(p, fd);
5071        proc_fdunlock(p);
5072
5073	fg_free(fp->f_fglob);
5074	fileproc_free(fp);
5075	return(0);
5076}
5077
5078
5079/*
5080 * flock
5081 *
5082 * Description:	Apply an advisory lock on a file descriptor.
5083 *
5084 * Parameters:	p				Process making request
5085 *		uap->fd				fd on which the lock is to be
5086 *						attempted
5087 *		uap->how			(Un)Lock bits, including type
5088 *		retval				Pointer to the call return area
5089 *
5090 * Returns:	0				Success
5091 *	fp_getfvp:EBADF				Bad file descriptor
5092 *	fp_getfvp:ENOTSUP			fd does not refer to a vnode
5093 *	vnode_getwithref:???
5094 *	VNOP_ADVLOCK:???
5095 *
5096 * Implicit returns:
5097 *		*retval (modified)		Size of dtable
5098 *
5099 * Notes:	Just attempt to get a record lock of the requested type on
5100 *		the entire file (l_whence = SEEK_SET, l_start = 0, l_len = 0).
5101 */
5102int
5103flock(proc_t p, struct flock_args *uap, __unused int32_t *retval)
5104{
5105	int fd = uap->fd;
5106	int how = uap->how;
5107	struct fileproc *fp;
5108	struct vnode *vp;
5109	struct flock lf;
5110	vfs_context_t ctx = vfs_context_current();
5111	int error=0;
5112
5113	AUDIT_ARG(fd, uap->fd);
5114	if ( (error = fp_getfvp(p, fd, &fp, &vp)) ) {
5115		return(error);
5116	}
5117	if ( (error = vnode_getwithref(vp)) ) {
5118		goto out1;
5119	}
5120	AUDIT_ARG(vnpath, vp, ARG_VNODE1);
5121
5122	lf.l_whence = SEEK_SET;
5123	lf.l_start = 0;
5124	lf.l_len = 0;
5125	if (how & LOCK_UN) {
5126		lf.l_type = F_UNLCK;
5127		fp->f_flag &= ~FHASLOCK;
5128		error = VNOP_ADVLOCK(vp, (caddr_t)fp->f_fglob, F_UNLCK, &lf, F_FLOCK, ctx, NULL);
5129		goto out;
5130	}
5131	if (how & LOCK_EX)
5132		lf.l_type = F_WRLCK;
5133	else if (how & LOCK_SH)
5134		lf.l_type = F_RDLCK;
5135	else {
5136	        error = EBADF;
5137		goto out;
5138	}
5139#if CONFIG_MACF
5140	error = mac_file_check_lock(proc_ucred(p), fp->f_fglob, F_SETLK, &lf);
5141	if (error)
5142		goto out;
5143#endif
5144	fp->f_flag |= FHASLOCK;
5145	if (how & LOCK_NB) {
5146		error = VNOP_ADVLOCK(vp, (caddr_t)fp->f_fglob, F_SETLK, &lf, F_FLOCK, ctx, NULL);
5147		goto out;
5148	}
5149	error = VNOP_ADVLOCK(vp, (caddr_t)fp->f_fglob, F_SETLK, &lf, F_FLOCK|F_WAIT, ctx, NULL);
5150out:
5151	(void)vnode_put(vp);
5152out1:
5153	fp_drop(p, fd, fp, 0);
5154	return(error);
5155
5156}
5157
5158/*
5159 * fileport_makeport
5160 *
5161 * Description: Obtain a Mach send right for a given file descriptor.
5162 *
5163 * Parameters:	p		Process calling fileport
5164 * 		uap->fd		The fd to reference
5165 * 		uap->portnamep  User address at which to place port name.
5166 *
5167 * Returns:	0		Success.
5168 *     		EBADF		Bad file descriptor.
5169 *     		EINVAL		File descriptor had type that cannot be sent, misc. other errors.
5170 *     		EFAULT		Address at which to store port name is not valid.
5171 *     		EAGAIN		Resource shortage.
5172 *
5173 * Implicit returns:
5174 *		On success, name of send right is stored at user-specified address.
5175 */
5176int
5177fileport_makeport(proc_t p, struct fileport_makeport_args *uap,
5178    __unused int *retval)
5179{
5180	int err;
5181	int fd = uap->fd;
5182	user_addr_t user_portaddr = uap->portnamep;
5183	struct fileproc *fp = FILEPROC_NULL;
5184	struct fileglob *fg = NULL;
5185	ipc_port_t fileport;
5186	mach_port_name_t name = MACH_PORT_NULL;
5187
5188	err = fp_lookup(p, fd, &fp, 0);
5189	if (err != 0) {
5190		goto out;
5191	}
5192
5193	if (!filetype_issendable(fp->f_type)) {
5194		err = EINVAL;
5195		goto out;
5196	}
5197
5198	if (FP_ISGUARDED(fp, GUARD_FILEPORT)) {
5199		proc_fdlock(p);
5200		err = fp_guard_exception(p, fd, fp, kGUARD_EXC_FILEPORT);
5201		proc_fdunlock(p);
5202		goto out;
5203	}
5204
5205	/* Dropped when port is deallocated */
5206	fg = fp->f_fglob;
5207	fg_ref(fp);
5208
5209	/* Allocate and initialize a port */
5210	fileport = fileport_alloc(fg);
5211	if (fileport == IPC_PORT_NULL) {
5212		err = EAGAIN;
5213		fg_drop(fp);
5214		goto out;
5215	}
5216
5217	/* Add an entry.  Deallocates port on failure. */
5218	name = ipc_port_copyout_send(fileport, get_task_ipcspace(p->task));
5219	if (!MACH_PORT_VALID(name)) {
5220		err = EINVAL;
5221		goto out;
5222	}
5223
5224	err = copyout(&name, user_portaddr, sizeof(mach_port_name_t));
5225	if (err != 0) {
5226		goto out;
5227	}
5228
5229	/* Tag the fileglob for debugging purposes */
5230	lck_mtx_lock_spin(&fg->fg_lock);
5231	fg->fg_lflags |= FG_PORTMADE;
5232	lck_mtx_unlock(&fg->fg_lock);
5233
5234	fp_drop(p, fd, fp, 0);
5235
5236	return 0;
5237
5238out:
5239	if (MACH_PORT_VALID(name)) {
5240		/* Don't care if another thread races us to deallocate the entry */
5241		(void) mach_port_deallocate(get_task_ipcspace(p->task), name);
5242	}
5243
5244	if (fp != FILEPROC_NULL) {
5245		fp_drop(p, fd, fp, 0);
5246	}
5247
5248	return err;
5249}
5250
5251void
5252fileport_releasefg(struct fileglob *fg)
5253{
5254	(void)closef_locked(NULL, fg, PROC_NULL);
5255
5256	return;
5257}
5258
5259
5260/*
5261 * fileport_makefd
5262 *
5263 * Description: Obtain the file descriptor for a given Mach send right.
5264 *
5265 * Parameters:	p		Process calling fileport
5266 * 		uap->port	Name of send right to file port.
5267 *
5268 * Returns:	0		Success
5269 *		EINVAL		Invalid Mach port name, or port is not for a file.
5270 *	fdalloc:EMFILE
5271 *	fdalloc:ENOMEM		Unable to allocate fileproc or extend file table.
5272 *
5273 * Implicit returns:
5274 *		*retval (modified)		The new descriptor
5275 */
5276int
5277fileport_makefd(proc_t p, struct fileport_makefd_args *uap, int32_t *retval)
5278{
5279	struct fileglob *fg;
5280 	struct fileproc *fp = FILEPROC_NULL;
5281	ipc_port_t port = IPC_PORT_NULL;
5282	mach_port_name_t send = uap->port;
5283	kern_return_t res;
5284	int fd;
5285	int err;
5286
5287	res = ipc_object_copyin(get_task_ipcspace(p->task),
5288			send, MACH_MSG_TYPE_COPY_SEND, &port);
5289
5290	if (res != KERN_SUCCESS) {
5291		err = EINVAL;
5292		goto out;
5293	}
5294
5295	fg = fileport_port_to_fileglob(port);
5296	if (fg == NULL) {
5297		err = EINVAL;
5298		goto out;
5299	}
5300
5301	fp = fileproc_alloc_init(NULL);
5302	if (fp == FILEPROC_NULL) {
5303		err = ENOMEM;
5304		goto out;
5305	}
5306
5307	fp->f_fglob = fg;
5308	fg_ref(fp);
5309
5310 	proc_fdlock(p);
5311	err = fdalloc(p, 0, &fd);
5312	if (err != 0) {
5313		proc_fdunlock(p);
5314		goto out;
5315	}
5316	*fdflags(p, fd) |= UF_EXCLOSE;
5317
5318	procfdtbl_releasefd(p, fd, fp);
5319	proc_fdunlock(p);
5320
5321	*retval = fd;
5322	err = 0;
5323out:
5324	if ((fp != NULL) && (0 != err)) {
5325		fileproc_free(fp);
5326	}
5327
5328	if (IPC_PORT_NULL != port) {
5329		ipc_port_release_send(port);
5330	}
5331
5332	return err;
5333}
5334
5335
5336/*
5337 * dupfdopen
5338 *
5339 * Description:	Duplicate the specified descriptor to a free descriptor;
5340 *		this is the second half of fdopen(), above.
5341 *
5342 * Parameters:	fdp				filedesc pointer to fill in
5343 *		indx				fd to dup to
5344 *		dfd				fd to dup from
5345 *		mode				mode to set on new fd
5346 *		error				command code
5347 *
5348 * Returns:	0				Success
5349 *		EBADF				Source fd is bad
5350 *		EACCES				Requested mode not allowed
5351 *		!0				'error', if not ENODEV or
5352 *						ENXIO
5353 *
5354 * Notes:	XXX This is not thread safe; see fdopen() above
5355 */
5356int
5357dupfdopen(struct filedesc *fdp, int indx, int dfd, int flags, int error)
5358{
5359	struct fileproc *wfp;
5360	struct fileproc *fp;
5361#if CONFIG_MACF
5362	int myerror;
5363#endif
5364	proc_t p = current_proc();
5365
5366	/*
5367	 * If the to-be-dup'd fd number is greater than the allowed number
5368	 * of file descriptors, or the fd to be dup'd has already been
5369	 * closed, reject.  Note, check for new == old is necessary as
5370	 * falloc could allocate an already closed to-be-dup'd descriptor
5371	 * as the new descriptor.
5372	 */
5373	proc_fdlock(p);
5374
5375	fp = fdp->fd_ofiles[indx];
5376	if (dfd < 0 || dfd >= fdp->fd_nfiles ||
5377			(wfp = fdp->fd_ofiles[dfd]) == NULL || wfp == fp ||
5378	                (fdp->fd_ofileflags[dfd] & UF_RESERVED)) {
5379
5380	        proc_fdunlock(p);
5381		return (EBADF);
5382	}
5383#if CONFIG_MACF
5384	myerror = mac_file_check_dup(proc_ucred(p), wfp->f_fglob, dfd);
5385	if (myerror) {
5386		proc_fdunlock(p);
5387		return (myerror);
5388	}
5389#endif
5390	/*
5391	 * There are two cases of interest here.
5392	 *
5393	 * For ENODEV simply dup (dfd) to file descriptor
5394	 * (indx) and return.
5395	 *
5396	 * For ENXIO steal away the file structure from (dfd) and
5397	 * store it in (indx).  (dfd) is effectively closed by
5398	 * this operation.
5399	 *
5400	 * Any other error code is just returned.
5401	 */
5402	switch (error) {
5403	case ENODEV:
5404		if (FP_ISGUARDED(wfp, GUARD_DUP)) {
5405			int err = fp_guard_exception(p,
5406			    dfd, wfp, kGUARD_EXC_DUP);
5407			proc_fdunlock(p);
5408			return (err);
5409		}
5410
5411		/*
5412		 * Check that the mode the file is being opened for is a
5413		 * subset of the mode of the existing descriptor.
5414		 */
5415	        if (((flags & (FREAD|FWRITE)) | wfp->f_flag) != wfp->f_flag) {
5416		        proc_fdunlock(p);
5417			return (EACCES);
5418		}
5419		if (indx > fdp->fd_lastfile)
5420			fdp->fd_lastfile = indx;
5421		(void)fg_ref(wfp);
5422
5423		if (fp->f_fglob)
5424		        fg_free(fp->f_fglob);
5425		fp->f_fglob = wfp->f_fglob;
5426
5427		fdp->fd_ofileflags[indx] = fdp->fd_ofileflags[dfd] |
5428			(flags & O_CLOEXEC) ? UF_EXCLOSE : 0;
5429
5430	        proc_fdunlock(p);
5431		return (0);
5432
5433	default:
5434	        proc_fdunlock(p);
5435		return (error);
5436	}
5437	/* NOTREACHED */
5438}
5439
5440
5441/*
5442 * fg_ref
5443 *
5444 * Description:	Add a reference to a fileglob by fileproc
5445 *
5446 * Parameters:	fp				fileproc containing fileglob
5447 *						pointer
5448 *
5449 * Returns:	void
5450 *
5451 * Notes:	XXX Should use OSAddAtomic?
5452 */
5453void
5454fg_ref(struct fileproc * fp)
5455{
5456	struct fileglob *fg;
5457
5458	fg = fp->f_fglob;
5459
5460	lck_mtx_lock_spin(&fg->fg_lock);
5461
5462#if DIAGNOSTIC
5463	if ((fp->f_flags & ~((unsigned int)FP_VALID_FLAGS)) != 0)
5464		panic("fg_ref: invalid bits on fp %p", fp);
5465
5466	if (fg->fg_count == 0)
5467		panic("fg_ref: adding fgcount to zeroed fg: fp %p fg %p",
5468		    fp, fg);
5469#endif
5470	fg->fg_count++;
5471	lck_mtx_unlock(&fg->fg_lock);
5472}
5473
5474
5475/*
5476 * fg_drop
5477 *
5478 * Description:	Remove a reference to a fileglob by fileproc
5479 *
5480 * Parameters:	fp				fileproc containing fileglob
5481 *						pointer
5482 *
5483 * Returns:	void
5484 *
5485 * Notes:	XXX Should use OSAddAtomic?
5486 */
5487void
5488fg_drop(struct fileproc * fp)
5489{
5490	struct fileglob *fg;
5491
5492	fg = fp->f_fglob;
5493	lck_mtx_lock_spin(&fg->fg_lock);
5494	fg->fg_count--;
5495	lck_mtx_unlock(&fg->fg_lock);
5496}
5497
5498#if SOCKETS
5499/*
5500 * fg_insertuipc
5501 *
5502 * Description:	Insert fileglob onto message queue
5503 *
5504 * Parameters:	fg				Fileglob pointer to insert
5505 *
5506 * Returns:	void
5507 *
5508 * Locks:	Takes and drops fg_lock, potentially many times
5509 */
5510void
5511fg_insertuipc(struct fileglob * fg)
5512{
5513	int insertque = 0;
5514
5515	lck_mtx_lock_spin(&fg->fg_lock);
5516
5517	while (fg->fg_lflags & FG_RMMSGQ) {
5518		lck_mtx_convert_spin(&fg->fg_lock);
5519
5520		fg->fg_lflags |= FG_WRMMSGQ;
5521		msleep(&fg->fg_lflags, &fg->fg_lock, 0, "fg_insertuipc", NULL);
5522	}
5523
5524	fg->fg_count++;
5525	fg->fg_msgcount++;
5526	if (fg->fg_msgcount == 1) {
5527		fg->fg_lflags |= FG_INSMSGQ;
5528		insertque=1;
5529	}
5530	lck_mtx_unlock(&fg->fg_lock);
5531
5532	if (insertque) {
5533		lck_mtx_lock_spin(uipc_lock);
5534		unp_gc_wait();
5535		LIST_INSERT_HEAD(&fmsghead, fg, f_msglist);
5536		lck_mtx_unlock(uipc_lock);
5537		lck_mtx_lock(&fg->fg_lock);
5538		fg->fg_lflags &= ~FG_INSMSGQ;
5539		if (fg->fg_lflags & FG_WINSMSGQ) {
5540			fg->fg_lflags &= ~FG_WINSMSGQ;
5541			wakeup(&fg->fg_lflags);
5542		}
5543		lck_mtx_unlock(&fg->fg_lock);
5544	}
5545
5546}
5547
5548
5549/*
5550 * fg_removeuipc
5551 *
5552 * Description:	Remove fileglob from message queue
5553 *
5554 * Parameters:	fg				Fileglob pointer to remove
5555 *
5556 * Returns:	void
5557 *
5558 * Locks:	Takes and drops fg_lock, potentially many times
5559 */
5560void
5561fg_removeuipc(struct fileglob * fg)
5562{
5563	int removeque = 0;
5564
5565	lck_mtx_lock_spin(&fg->fg_lock);
5566	while (fg->fg_lflags & FG_INSMSGQ) {
5567		lck_mtx_convert_spin(&fg->fg_lock);
5568
5569		fg->fg_lflags |= FG_WINSMSGQ;
5570		msleep(&fg->fg_lflags, &fg->fg_lock, 0, "fg_removeuipc", NULL);
5571	}
5572	fg->fg_msgcount--;
5573	if (fg->fg_msgcount == 0) {
5574		fg->fg_lflags |= FG_RMMSGQ;
5575		removeque=1;
5576	}
5577	lck_mtx_unlock(&fg->fg_lock);
5578
5579	if (removeque) {
5580		lck_mtx_lock_spin(uipc_lock);
5581		unp_gc_wait();
5582		LIST_REMOVE(fg, f_msglist);
5583		lck_mtx_unlock(uipc_lock);
5584		lck_mtx_lock(&fg->fg_lock);
5585		fg->fg_lflags &= ~FG_RMMSGQ;
5586		if (fg->fg_lflags & FG_WRMMSGQ) {
5587			fg->fg_lflags &= ~FG_WRMMSGQ;
5588			wakeup(&fg->fg_lflags);
5589		}
5590		lck_mtx_unlock(&fg->fg_lock);
5591	}
5592}
5593#endif /* SOCKETS */
5594
5595/*
5596 * fo_read
5597 *
5598 * Description:	Generic fileops read indirected through the fileops pointer
5599 *		in the fileproc structure
5600 *
5601 * Parameters:	fp				fileproc structure pointer
5602 *		uio				user I/O structure pointer
5603 *		flags				FOF_ flags
5604 *		ctx				VFS context for operation
5605 *
5606 * Returns:	0				Success
5607 *		!0				Errno from read
5608 */
5609int
5610fo_read(struct fileproc *fp, struct uio *uio, int flags, vfs_context_t ctx)
5611{
5612	return ((*fp->f_ops->fo_read)(fp, uio, flags, ctx));
5613}
5614
5615
5616/*
5617 * fo_write
5618 *
5619 * Description:	Generic fileops write indirected through the fileops pointer
5620 *		in the fileproc structure
5621 *
5622 * Parameters:	fp				fileproc structure pointer
5623 *		uio				user I/O structure pointer
5624 *		flags				FOF_ flags
5625 *		ctx				VFS context for operation
5626 *
5627 * Returns:	0				Success
5628 *		!0				Errno from write
5629 */
5630int
5631fo_write(struct fileproc *fp, struct uio *uio, int flags, vfs_context_t ctx)
5632{
5633	return((*fp->f_ops->fo_write)(fp, uio, flags, ctx));
5634}
5635
5636
5637/*
5638 * fo_ioctl
5639 *
5640 * Description:	Generic fileops ioctl indirected through the fileops pointer
5641 *		in the fileproc structure
5642 *
5643 * Parameters:	fp				fileproc structure pointer
5644 *		com				ioctl command
5645 *		data				pointer to internalized copy
5646 *						of user space ioctl command
5647 *						parameter data in kernel space
5648 *		ctx				VFS context for operation
5649 *
5650 * Returns:	0				Success
5651 *		!0				Errno from ioctl
5652 *
5653 * Locks:	The caller is assumed to have held the proc_fdlock; this
5654 *		function releases and reacquires this lock.  If the caller
5655 *		accesses data protected by this lock prior to calling this
5656 *		function, it will need to revalidate/reacquire any cached
5657 *		protected data obtained prior to the call.
5658 */
5659int
5660fo_ioctl(struct fileproc *fp, u_long com, caddr_t data, vfs_context_t ctx)
5661{
5662	int error;
5663
5664	proc_fdunlock(vfs_context_proc(ctx));
5665	error = (*fp->f_ops->fo_ioctl)(fp, com, data, ctx);
5666	proc_fdlock(vfs_context_proc(ctx));
5667	return(error);
5668}
5669
5670
5671/*
5672 * fo_select
5673 *
5674 * Description:	Generic fileops select indirected through the fileops pointer
5675 *		in the fileproc structure
5676 *
5677 * Parameters:	fp				fileproc structure pointer
5678 *		which				select which
5679 *		wql				pointer to wait queue list
5680 *		ctx				VFS context for operation
5681 *
5682 * Returns:	0				Success
5683 *		!0				Errno from select
5684 */
5685int
5686fo_select(struct fileproc *fp, int which, void *wql, vfs_context_t ctx)
5687{
5688	return((*fp->f_ops->fo_select)(fp, which, wql, ctx));
5689}
5690
5691
5692/*
5693 * fo_close
5694 *
5695 * Description:	Generic fileops close indirected through the fileops pointer
5696 *		in the fileproc structure
5697 *
5698 * Parameters:	fp				fileproc structure pointer for
5699 *						file to close
5700 *		ctx				VFS context for operation
5701 *
5702 * Returns:	0				Success
5703 *		!0				Errno from close
5704 */
5705int
5706fo_close(struct fileglob *fg, vfs_context_t ctx)
5707{
5708	return((*fg->fg_ops->fo_close)(fg, ctx));
5709}
5710
5711
5712/*
5713 * fo_kqfilter
5714 *
5715 * Description:	Generic fileops kqueue filter indirected through the fileops
5716 *		pointer in the fileproc structure
5717 *
5718 * Parameters:	fp				fileproc structure pointer
5719 *		kn				pointer to knote to filter on
5720 *		ctx				VFS context for operation
5721 *
5722 * Returns:	0				Success
5723 *		!0				Errno from kqueue filter
5724 */
5725int
5726fo_kqfilter(struct fileproc *fp, struct knote *kn, vfs_context_t ctx)
5727{
5728        return ((*fp->f_ops->fo_kqfilter)(fp, kn, ctx));
5729}
5730
5731/*
5732 * The ability to send a file descriptor to another
5733 * process is opt-in by file type.
5734 */
5735boolean_t
5736filetype_issendable(file_type_t fdtype)
5737{
5738	switch (fdtype) {
5739		case DTYPE_VNODE:
5740		case DTYPE_SOCKET:
5741		case DTYPE_PIPE:
5742		case DTYPE_PSXSHM:
5743			return TRUE;
5744		default:
5745			/* DTYPE_KQUEUE, DTYPE_FSEVENTS, DTYPE_PSXSEM */
5746			return FALSE;
5747	}
5748}
5749
5750
5751struct fileproc *
5752fileproc_alloc_init(__unused void *arg)
5753{
5754	struct fileproc *fp;
5755
5756	MALLOC_ZONE(fp, struct fileproc *, sizeof (*fp), M_FILEPROC, M_WAITOK);
5757	if (fp)
5758		bzero(fp, sizeof (*fp));
5759
5760	return (fp);
5761}
5762
5763void
5764fileproc_free(struct fileproc *fp)
5765{
5766	switch (FILEPROC_TYPE(fp)) {
5767	case FTYPE_SIMPLE:
5768		FREE_ZONE(fp, sizeof (*fp), M_FILEPROC);
5769		break;
5770	case FTYPE_GUARDED:
5771		guarded_fileproc_free(fp);
5772		break;
5773	default:
5774		panic("%s: corrupt fp %p flags %x", __func__, fp, fp->f_flags);
5775	}
5776}
5777