NetBSD/sys/msdosfs/msdosfs_fat.c

1087 lines
27 KiB
C
Raw Normal View History

/* $NetBSD: msdosfs_fat.c,v 1.32 2000/03/27 17:40:26 jdolecek Exp $ */
/*-
* Copyright (C) 1994, 1995, 1997 Wolfgang Solfrank.
* Copyright (C) 1994, 1995, 1997 TooLs GmbH.
* All rights reserved.
* Original code by Paul Popelka (paulp@uts.amdahl.com) (see below).
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by TooLs GmbH.
* 4. The name of TooLs GmbH may not be used to endorse or promote products
* derived from this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY TOOLS GMBH ``AS IS'' AND ANY EXPRESS OR
* IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
* OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
* IN NO EVENT SHALL TOOLS GMBH BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
* PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS;
* OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
* WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
* OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF
* ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
/*
* Written by Paul Popelka (paulp@uts.amdahl.com)
*
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
* You can do anything you want with this software, just don't say you wrote
* it, and don't remove this notice.
*
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
* This software is provided "as is".
*
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
* The author supplies this software to be publicly redistributed on the
* understanding that the author is not responsible for the correct
* functioning of this software in any circumstances and is not liable for
* any damages caused by this software.
*
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
* October 1992
*/
/*
* kernel include files.
*/
1993-12-18 03:50:51 +03:00
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/buf.h>
#include <sys/file.h>
#include <sys/namei.h>
#include <sys/mount.h> /* to define statfs structure */
#include <sys/vnode.h> /* to define vattr structure */
#include <sys/errno.h>
1995-10-15 18:34:19 +03:00
#include <sys/dirent.h>
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
/*
* msdosfs include files.
*/
1993-12-18 03:50:51 +03:00
#include <msdosfs/bpb.h>
#include <msdosfs/msdosfsmount.h>
#include <msdosfs/direntry.h>
#include <msdosfs/denode.h>
#include <msdosfs/fat.h>
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
/*
* Fat cache stats.
*/
int fc_fileextends; /* # of file extends */
int fc_lfcempty; /* # of time last file cluster cache entry
* was empty */
int fc_bmapcalls; /* # of times pcbmap was called */
#define LMMAX 20
int fc_lmdistance[LMMAX]; /* counters for how far off the last
* cluster mapped entry was. */
int fc_largedistance; /* off by more than LMMAX */
1996-02-09 22:13:39 +03:00
static void fatblock __P((struct msdosfsmount *, u_long, u_long *, u_long *,
u_long *));
void updatefats __P((struct msdosfsmount *, struct buf *, u_long));
static __inline void usemap_free __P((struct msdosfsmount *, u_long));
static __inline void usemap_alloc __P((struct msdosfsmount *, u_long));
static int fatchain __P((struct msdosfsmount *, u_long, u_long, u_long));
int chainlength __P((struct msdosfsmount *, u_long, u_long));
int chainalloc __P((struct msdosfsmount *, u_long, u_long, u_long, u_long *,
u_long *));
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
static void
fatblock(pmp, ofs, bnp, sizep, bop)
struct msdosfsmount *pmp;
u_long ofs;
u_long *bnp;
u_long *sizep;
u_long *bop;
{
u_long bn, size;
bn = ofs / pmp->pm_fatblocksize * pmp->pm_fatblocksec;
size = min(pmp->pm_fatblocksec, pmp->pm_FATsecs - bn)
* pmp->pm_BytesPerSec;
bn += pmp->pm_fatblk + pmp->pm_curfat * pmp->pm_FATsecs;
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
if (bnp)
*bnp = bn;
if (sizep)
*sizep = size;
if (bop)
*bop = ofs % pmp->pm_fatblocksize;
}
/*
* Map the logical cluster number of a file into a physical disk sector
1994-07-17 01:32:06 +04:00
* that is filesystem relative.
*
* dep - address of denode representing the file of interest
* findcn - file relative cluster whose filesystem relative cluster number
* and/or block number are/is to be found
* bnp - address of where to place the file system relative block number.
* If this pointer is null then don't return this quantity.
* cnp - address of where to place the file system relative cluster number.
* If this pointer is null then don't return this quantity.
*
* NOTE: Either bnp or cnp must be non-null.
* This function has one side effect. If the requested file relative cluster
* is beyond the end of file, then the actual number of clusters in the file
* is returned in *cnp. This is useful for determining how long a directory is.
* If cnp is null, nothing is returned.
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
*/
int
pcbmap(dep, findcn, bnp, cnp, sp)
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
struct denode *dep;
u_long findcn; /* file relative cluster to get */
daddr_t *bnp; /* returned filesys relative blk number */
u_long *cnp; /* returned cluster number */
int *sp; /* returned block size */
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
{
int error;
u_long i;
u_long cn;
1996-02-09 22:13:39 +03:00
u_long prevcn = 0; /* XXX: prevcn could be used unititialized */
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
u_long byteoffset;
u_long bn;
u_long bo;
struct buf *bp = NULL;
u_long bp_bn = -1;
struct msdosfsmount *pmp = dep->de_pmp;
u_long bsize;
fc_bmapcalls++;
/*
* If they don't give us someplace to return a value then don't
* bother doing anything.
*/
if (bnp == NULL && cnp == NULL && sp == NULL)
return (0);
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
cn = dep->de_StartCluster;
/*
* The "file" that makes up the root directory is contiguous,
* permanently allocated, of fixed size, and is not made up of
* clusters. If the cluster number is beyond the end of the root
* directory, then return the number of clusters in the file.
*/
if (cn == MSDOSFSROOT) {
if (dep->de_Attributes & ATTR_DIRECTORY) {
1995-10-15 18:34:19 +03:00
if (de_cn2off(pmp, findcn) >= dep->de_FileSize) {
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
if (cnp)
1995-10-15 18:34:19 +03:00
*cnp = de_bn2cn(pmp, pmp->pm_rootdirsize);
return (E2BIG);
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
}
if (bnp)
1995-10-15 18:34:19 +03:00
*bnp = pmp->pm_rootdirblk + de_cn2bn(pmp, findcn);
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
if (cnp)
*cnp = MSDOSFSROOT;
if (sp)
*sp = min(pmp->pm_bpcluster,
1995-10-15 18:34:19 +03:00
dep->de_FileSize - de_cn2off(pmp, findcn));
return (0);
} else { /* just an empty file */
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
if (cnp)
*cnp = 0;
return (E2BIG);
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
}
}
/*
* All other files do I/O in cluster sized blocks
*/
if (sp)
*sp = pmp->pm_bpcluster;
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
/*
* Rummage around in the fat cache, maybe we can avoid tromping
* thru every fat entry for the file. And, keep track of how far
* off the cache was from where we wanted to be.
*/
i = 0;
fc_lookup(dep, findcn, &i, &cn);
if ((bn = findcn - i) >= LMMAX)
fc_largedistance++;
else
fc_lmdistance[bn]++;
/*
* Handle all other files or directories the normal way.
*/
for (; i < findcn; i++) {
/*
* Stop with all reserved clusters, not just with EOF.
*/
if (cn >= (CLUST_RSRVD & pmp->pm_fatmask))
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
goto hiteof;
byteoffset = FATOFS(pmp, cn);
fatblock(pmp, byteoffset, &bn, &bsize, &bo);
if (bn != bp_bn) {
if (bp)
brelse(bp);
1996-02-09 22:13:39 +03:00
error = bread(pmp->pm_devvp, bn, bsize, NOCRED, &bp);
if (error) {
1995-10-15 18:34:19 +03:00
brelse(bp);
return (error);
1995-10-15 18:34:19 +03:00
}
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
bp_bn = bn;
}
prevcn = cn;
if (FAT32(pmp))
cn = getulong(&bp->b_data[bo]);
else
cn = getushort(&bp->b_data[bo]);
if (FAT12(pmp) && (prevcn & 1))
cn >>= 4;
cn &= pmp->pm_fatmask;
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
}
if (!MSDOSFSEOF(pmp, cn)) {
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
if (bp)
brelse(bp);
if (bnp)
*bnp = cntobn(pmp, cn);
if (cnp)
*cnp = cn;
fc_setcache(dep, FC_LASTMAP, i, cn);
return (0);
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
}
hiteof:;
if (cnp)
*cnp = i;
if (bp)
brelse(bp);
/* update last file cluster entry in the fat cache */
fc_setcache(dep, FC_LASTFC, i - 1, prevcn);
return (E2BIG);
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
}
/*
* Find the closest entry in the fat cache to the cluster we are looking
* for.
*/
void
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
fc_lookup(dep, findcn, frcnp, fsrcnp)
struct denode *dep;
u_long findcn;
u_long *frcnp;
u_long *fsrcnp;
{
int i;
u_long cn;
struct fatcache *closest = 0;
for (i = 0; i < FC_SIZE; i++) {
cn = dep->de_fc[i].fc_frcn;
if (cn != FCE_EMPTY && cn <= findcn) {
if (closest == 0 || cn > closest->fc_frcn)
closest = &dep->de_fc[i];
}
}
if (closest) {
*frcnp = closest->fc_frcn;
*fsrcnp = closest->fc_fsrcn;
}
}
/*
* Purge the fat cache in denode dep of all entries relating to file
* relative cluster frcn and beyond.
*/
void
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
fc_purge(dep, frcn)
struct denode *dep;
u_int frcn;
{
int i;
struct fatcache *fcp;
fcp = dep->de_fc;
for (i = 0; i < FC_SIZE; i++, fcp++) {
1994-07-17 01:32:06 +04:00
if (fcp->fc_frcn >= frcn)
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
fcp->fc_frcn = FCE_EMPTY;
}
}
/*
* Update the fat.
* If mirroring the fat, update all copies, with the first copy as last.
* Else update only the current fat (ignoring the others).
1994-07-17 01:32:06 +04:00
*
* pmp - msdosfsmount structure for filesystem to update
* bp - addr of modified fat block
* fatbn - block number relative to begin of filesystem of the modified fat block.
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
*/
void
updatefats(pmp, bp, fatbn)
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
struct msdosfsmount *pmp;
struct buf *bp;
u_long fatbn;
{
int i;
struct buf *bpn;
#ifdef MSDOSFS_DEBUG
printf("updatefats(pmp %p, bp %p, fatbn %lu)\n",
1996-10-11 02:46:11 +04:00
pmp, bp, fatbn);
#endif
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
/*
* If we have an FSInfo block, update it.
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
*/
if (pmp->pm_fsinfo) {
u_long cn = pmp->pm_nxtfree;
if (pmp->pm_freeclustercount
&& (pmp->pm_inusemap[cn / N_INUSEBITS]
& (1 << (cn % N_INUSEBITS)))) {
/*
* The cluster indicated in FSInfo isn't free
* any longer. Got get a new free one.
*/
for (cn = 0; cn < pmp->pm_maxcluster; cn++)
if (pmp->pm_inusemap[cn / N_INUSEBITS] != (u_int)-1)
break;
pmp->pm_nxtfree = cn
+ ffs(pmp->pm_inusemap[cn / N_INUSEBITS]
^ (u_int)-1) - 1;
}
if (bread(pmp->pm_devvp, pmp->pm_fsinfo, 1024, NOCRED, &bpn) != 0) {
/*
* Ignore the error, but turn off FSInfo update for the future.
*/
pmp->pm_fsinfo = 0;
brelse(bpn);
} else {
struct fsinfo *fp = (struct fsinfo *)bpn->b_data;
putulong(fp->fsinfree, pmp->pm_freeclustercount);
putulong(fp->fsinxtfree, pmp->pm_nxtfree);
if (pmp->pm_flags & MSDOSFSMNT_WAITONFAT)
bwrite(bpn);
else
bdwrite(bpn);
}
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
}
if (pmp->pm_flags & MSDOSFS_FATMIRROR) {
/*
* Now copy the block(s) of the modified fat to the other copies of
* the fat and write them out. This is faster than reading in the
* other fats and then writing them back out. This could tie up
* the fat for quite a while. Preventing others from accessing it.
* To prevent us from going after the fat quite so much we use
* delayed writes, unless they specfied "synchronous" when the
* filesystem was mounted. If synch is asked for then use
* bwrite()'s and really slow things down.
*/
for (i = 1; i < pmp->pm_FATs; i++) {
fatbn += pmp->pm_FATsecs;
/* getblk() never fails */
bpn = getblk(pmp->pm_devvp, fatbn, bp->b_bcount, 0, 0);
memcpy(bpn->b_data, bp->b_data, bp->b_bcount);
if (pmp->pm_flags & MSDOSFSMNT_WAITONFAT)
bwrite(bpn);
else
bdwrite(bpn);
}
}
/*
* Write out the first (or current) fat last.
*/
1995-10-15 18:34:19 +03:00
if (pmp->pm_flags & MSDOSFSMNT_WAITONFAT)
bwrite(bp);
else
bdwrite(bp);
/*
* Maybe update fsinfo sector here?
*/
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
}
/*
* Updating entries in 12 bit fats is a pain in the butt.
*
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
* The following picture shows where nibbles go when moving from a 12 bit
* cluster number into the appropriate bytes in the FAT.
*
1994-07-17 01:32:06 +04:00
* byte m byte m+1 byte m+2
* +----+----+ +----+----+ +----+----+
* | 0 1 | | 2 3 | | 4 5 | FAT bytes
* +----+----+ +----+----+ +----+----+
*
1994-07-17 01:32:06 +04:00
* +----+----+----+ +----+----+----+
* | 3 0 1 | | 4 5 2 |
* +----+----+----+ +----+----+----+
* cluster n cluster n+1
*
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
* Where n is even. m = n + (n >> 2)
*
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
*/
1995-01-04 09:32:19 +03:00
static __inline void
1994-06-30 11:25:23 +04:00
usemap_alloc(pmp, cn)
struct msdosfsmount *pmp;
u_long cn;
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
{
pmp->pm_inusemap[cn / N_INUSEBITS] |= 1 << (cn % N_INUSEBITS);
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
pmp->pm_freeclustercount--;
}
1995-01-04 09:32:19 +03:00
static __inline void
1994-06-30 11:25:23 +04:00
usemap_free(pmp, cn)
struct msdosfsmount *pmp;
u_long cn;
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
{
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
pmp->pm_freeclustercount++;
pmp->pm_inusemap[cn / N_INUSEBITS] &= ~(1 << (cn % N_INUSEBITS));
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
}
int
clusterfree(pmp, cluster, oldcnp)
struct msdosfsmount *pmp;
u_long cluster;
u_long *oldcnp;
{
int error;
u_long oldcn;
usemap_free(pmp, cluster);
1996-02-09 22:13:39 +03:00
error = fatentry(FAT_GET_AND_SET, pmp, cluster, &oldcn, MSDOSFSFREE);
if (error) {
usemap_alloc(pmp, cluster);
return (error);
}
/*
* If the cluster was successfully marked free, then update
* the count of free clusters, and turn off the "allocated"
* bit in the "in use" cluster bit map.
*/
if (oldcnp)
*oldcnp = oldcn;
return (0);
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
}
/*
1994-07-17 01:32:06 +04:00
* Get or Set or 'Get and Set' the cluster'th entry in the fat.
*
* function - whether to get or set a fat entry
* pmp - address of the msdosfsmount structure for the filesystem
* whose fat is to be manipulated.
* cn - which cluster is of interest
* oldcontents - address of a word that is to receive the contents of the
* cluster'th entry if this is a get function
* newcontents - the new value to be written into the cluster'th element of
* the fat if this is a set function.
*
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
* This function can also be used to free a cluster by setting the fat entry
* for a cluster to 0.
*
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
* All copies of the fat are updated if this is a set function. NOTE: If
* fatentry() marks a cluster as free it does not update the inusemap in
* the msdosfsmount structure. This is left to the caller.
*/
int
fatentry(function, pmp, cn, oldcontents, newcontents)
int function;
struct msdosfsmount *pmp;
u_long cn;
u_long *oldcontents;
u_long newcontents;
{
int error;
u_long readcn;
u_long bn, bo, bsize, byteoffset;
struct buf *bp;
#ifdef MSDOSFS_DEBUG
printf("fatentry(func %d, pmp %p, clust %lu, oldcon %p, newcon %lx)\n",
function, pmp, cn, oldcontents, newcontents);
1996-10-11 02:46:11 +04:00
#endif
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
#ifdef DIAGNOSTIC
/*
* Be sure they asked us to do something.
*/
if ((function & (FAT_SET | FAT_GET)) == 0) {
1996-10-13 08:10:34 +04:00
printf("fatentry(): function code doesn't specify get or set\n");
return (EINVAL);
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
}
/*
* If they asked us to return a cluster number but didn't tell us
* where to put it, give them an error.
*/
if ((function & FAT_GET) && oldcontents == NULL) {
1996-10-13 08:10:34 +04:00
printf("fatentry(): get function with no place to put result\n");
return (EINVAL);
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
}
#endif
/*
* Be sure the requested cluster is in the filesystem.
*/
if (cn < CLUST_FIRST || cn > pmp->pm_maxcluster)
return (EINVAL);
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
byteoffset = FATOFS(pmp, cn);
fatblock(pmp, byteoffset, &bn, &bsize, &bo);
1996-02-09 22:13:39 +03:00
if ((error = bread(pmp->pm_devvp, bn, bsize, NOCRED, &bp)) != 0) {
1995-10-15 18:34:19 +03:00
brelse(bp);
return (error);
1995-10-15 18:34:19 +03:00
}
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
if (function & FAT_GET) {
if (FAT32(pmp))
readcn = getulong(&bp->b_data[bo]);
else
readcn = getushort(&bp->b_data[bo]);
if (FAT12(pmp) & (cn & 1))
readcn >>= 4;
readcn &= pmp->pm_fatmask;
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
*oldcontents = readcn;
}
if (function & FAT_SET) {
switch (pmp->pm_fatmask) {
case FAT12_MASK:
1994-07-17 01:32:06 +04:00
readcn = getushort(&bp->b_data[bo]);
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
if (cn & 1) {
readcn &= 0x000f;
1994-07-17 01:32:06 +04:00
readcn |= newcontents << 4;
} else {
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
readcn &= 0xf000;
1994-07-17 01:32:06 +04:00
readcn |= newcontents & 0xfff;
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
}
1994-07-17 01:32:06 +04:00
putushort(&bp->b_data[bo], readcn);
break;
case FAT16_MASK:
1994-07-17 01:32:06 +04:00
putushort(&bp->b_data[bo], newcontents);
break;
case FAT32_MASK:
/*
* According to spec we have to retain the
* high order bits of the fat entry.
*/
readcn = getulong(&bp->b_data[bo]);
readcn &= ~FAT32_MASK;
readcn |= newcontents & FAT32_MASK;
putulong(&bp->b_data[bo], readcn);
break;
}
updatefats(pmp, bp, bn);
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
bp = NULL;
pmp->pm_fmod = 1;
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
}
if (bp)
brelse(bp);
return (0);
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
}
/*
* Update a contiguous cluster chain
1994-07-17 01:32:06 +04:00
*
* pmp - mount point
* start - first cluster of chain
* count - number of clusters in chain
* fillwith - what to write into fat entry of last cluster
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
*/
static int
fatchain(pmp, start, count, fillwith)
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
struct msdosfsmount *pmp;
u_long start;
u_long count;
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
u_long fillwith;
{
int error;
u_long bn, bo, bsize, byteoffset, readcn, newc;
struct buf *bp;
#ifdef MSDOSFS_DEBUG
printf("fatchain(pmp %p, start %lu, count %lu, fillwith %lx)\n",
1996-10-11 02:46:11 +04:00
pmp, start, count, fillwith);
#endif
/*
* Be sure the clusters are in the filesystem.
*/
if (start < CLUST_FIRST || start + count - 1 > pmp->pm_maxcluster)
return (EINVAL);
while (count > 0) {
byteoffset = FATOFS(pmp, start);
fatblock(pmp, byteoffset, &bn, &bsize, &bo);
1996-02-09 22:13:39 +03:00
error = bread(pmp->pm_devvp, bn, bsize, NOCRED, &bp);
if (error) {
1995-10-15 18:34:19 +03:00
brelse(bp);
return (error);
1995-10-15 18:34:19 +03:00
}
while (count > 0) {
start++;
newc = --count > 0 ? start : fillwith;
switch (pmp->pm_fatmask) {
case FAT12_MASK:
readcn = getushort(&bp->b_data[bo]);
if (start & 1) {
readcn &= 0xf000;
readcn |= newc & 0xfff;
} else {
readcn &= 0x000f;
readcn |= newc << 4;
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
}
putushort(&bp->b_data[bo], readcn);
bo++;
if (!(start & 1))
bo++;
break;
case FAT16_MASK:
putushort(&bp->b_data[bo], newc);
bo += 2;
break;
case FAT32_MASK:
readcn = getulong(&bp->b_data[bo]);
readcn &= ~pmp->pm_fatmask;
readcn |= newc & pmp->pm_fatmask;
putulong(&bp->b_data[bo], readcn);
bo += 4;
break;
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
}
if (bo >= bsize)
break;
}
updatefats(pmp, bp, bn);
}
pmp->pm_fmod = 1;
return (0);
}
/*
* Check the length of a free cluster chain starting at start.
*
* pmp - mount point
* start - start of chain
* count - maximum interesting length
*/
int
chainlength(pmp, start, count)
struct msdosfsmount *pmp;
u_long start;
u_long count;
{
u_long idx, max_idx;
u_int map;
u_long len;
max_idx = pmp->pm_maxcluster / N_INUSEBITS;
idx = start / N_INUSEBITS;
start %= N_INUSEBITS;
map = pmp->pm_inusemap[idx];
map &= ~((1 << start) - 1);
if (map) {
len = ffs(map) - 1 - start;
return (len > count ? count : len);
}
len = N_INUSEBITS - start;
if (len >= count)
return (count);
while (++idx <= max_idx) {
if (len >= count)
break;
1997-05-18 00:31:15 +04:00
if ((map = pmp->pm_inusemap[idx]) != 0) {
len += ffs(map) - 1;
break;
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
}
len += N_INUSEBITS;
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
}
return (len > count ? count : len);
}
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
/*
* Allocate contigous free clusters.
*
* pmp - mount point.
* start - start of cluster chain.
* count - number of clusters to allocate.
* fillwith - put this value into the fat entry for the
* last allocated cluster.
* retcluster - put the first allocated cluster's number here.
* got - how many clusters were actually allocated.
*/
int
chainalloc(pmp, start, count, fillwith, retcluster, got)
struct msdosfsmount *pmp;
u_long start;
u_long count;
u_long fillwith;
u_long *retcluster;
u_long *got;
{
int error;
u_long cl, n;
for (cl = start, n = count; n-- > 0;)
usemap_alloc(pmp, cl++);
1996-02-09 22:13:39 +03:00
if ((error = fatchain(pmp, start, count, fillwith)) != 0)
return (error);
#ifdef MSDOSFS_DEBUG
printf("clusteralloc(): allocated cluster chain at %lu (%lu clusters)\n",
1996-10-11 02:46:11 +04:00
start, count);
#endif
if (retcluster)
*retcluster = start;
if (got)
*got = count;
return (0);
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
}
/*
* Allocate contiguous free clusters.
*
* pmp - mount point.
* start - preferred start of cluster chain.
* count - number of clusters requested.
* fillwith - put this value into the fat entry for the
* last allocated cluster.
* retcluster - put the first allocated cluster's number here.
* got - how many clusters were actually allocated.
*/
int
clusteralloc(pmp, start, count, retcluster, got)
struct msdosfsmount *pmp;
u_long start;
u_long count;
u_long *retcluster;
u_long *got;
{
u_long idx;
1996-02-09 22:13:39 +03:00
u_long len, newst, foundl, cn, l;
u_long foundcn = 0; /* XXX: foundcn could be used unititialized */
u_long fillwith = CLUST_EOFE;
u_int map;
#ifdef MSDOSFS_DEBUG
printf("clusteralloc(): find %lu clusters\n",count);
#endif
if (start) {
if ((len = chainlength(pmp, start, count)) >= count)
return (chainalloc(pmp, start, count, fillwith, retcluster, got));
} else {
/*
* This is a new file, initialize start
*/
struct timeval tv;
microtime(&tv);
start = (tv.tv_usec >> 10) | tv.tv_usec;
len = 0;
}
/*
* Start at a (pseudo) random place to maximize cluster runs
* under multiple writers.
*/
newst = (start * 1103515245 + 12345) % (pmp->pm_maxcluster + 1);
foundl = 0;
for (cn = newst; cn <= pmp->pm_maxcluster;) {
idx = cn / N_INUSEBITS;
map = pmp->pm_inusemap[idx];
map |= (1 << (cn % N_INUSEBITS)) - 1;
if (map != (u_int)-1) {
cn = idx * N_INUSEBITS + ffs(map^(u_int)-1) - 1;
if ((l = chainlength(pmp, cn, count)) >= count)
return (chainalloc(pmp, cn, count, fillwith, retcluster, got));
if (l > foundl) {
foundcn = cn;
foundl = l;
}
cn += l + 1;
continue;
}
cn += N_INUSEBITS - cn % N_INUSEBITS;
}
for (cn = 0; cn < newst;) {
idx = cn / N_INUSEBITS;
map = pmp->pm_inusemap[idx];
map |= (1 << (cn % N_INUSEBITS)) - 1;
if (map != (u_int)-1) {
cn = idx * N_INUSEBITS + ffs(map^(u_int)-1) - 1;
if ((l = chainlength(pmp, cn, count)) >= count)
return (chainalloc(pmp, cn, count, fillwith, retcluster, got));
if (l > foundl) {
foundcn = cn;
foundl = l;
}
cn += l + 1;
continue;
}
cn += N_INUSEBITS - cn % N_INUSEBITS;
}
if (!foundl)
return (ENOSPC);
if (len)
return (chainalloc(pmp, start, len, fillwith, retcluster, got));
else
return (chainalloc(pmp, foundcn, foundl, fillwith, retcluster, got));
}
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
/*
1994-07-17 01:32:06 +04:00
* Free a chain of clusters.
*
* pmp - address of the msdosfs mount structure for the filesystem
* containing the cluster chain to be freed.
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
* startcluster - number of the 1st cluster in the chain of clusters to be
1994-07-17 01:32:06 +04:00
* freed.
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
*/
int
freeclusterchain(pmp, cluster)
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
struct msdosfsmount *pmp;
u_long cluster;
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
{
int error;
struct buf *bp = NULL;
u_long bn, bo, bsize, byteoffset;
u_long readcn, lbn = -1;
while (cluster >= CLUST_FIRST && cluster <= pmp->pm_maxcluster) {
byteoffset = FATOFS(pmp, cluster);
fatblock(pmp, byteoffset, &bn, &bsize, &bo);
if (lbn != bn) {
if (bp)
updatefats(pmp, bp, lbn);
1996-02-09 22:13:39 +03:00
error = bread(pmp->pm_devvp, bn, bsize, NOCRED, &bp);
if (error) {
1995-10-15 18:34:19 +03:00
brelse(bp);
return (error);
1995-10-15 18:34:19 +03:00
}
lbn = bn;
}
usemap_free(pmp, cluster);
switch (pmp->pm_fatmask) {
case FAT12_MASK:
readcn = getushort(&bp->b_data[bo]);
if (cluster & 1) {
cluster = readcn >> 4;
readcn &= 0x000f;
readcn |= MSDOSFSFREE << 4;
} else {
cluster = readcn;
readcn &= 0xf000;
readcn |= MSDOSFSFREE & 0xfff;
}
putushort(&bp->b_data[bo], readcn);
break;
case FAT16_MASK:
cluster = getushort(&bp->b_data[bo]);
putushort(&bp->b_data[bo], MSDOSFSFREE);
break;
case FAT32_MASK:
cluster = getulong(&bp->b_data[bo]);
putulong(&bp->b_data[bo],
(MSDOSFSFREE & FAT32_MASK) | (cluster & ~FAT32_MASK));
break;
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
}
cluster &= pmp->pm_fatmask;
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
}
if (bp)
updatefats(pmp, bp, bn);
return (0);
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
}
/*
* Read in fat blocks looking for free clusters. For every free cluster
* found turn off its corresponding bit in the pm_inusemap.
*/
int
fillinusemap(pmp)
struct msdosfsmount *pmp;
{
struct buf *bp = NULL;
u_long cn, readcn;
int error;
u_long bn, bo, bsize, byteoffset;
/*
* Mark all clusters in use, we mark the free ones in the fat scan
* loop further down.
*/
for (cn = 0; cn < (pmp->pm_maxcluster + N_INUSEBITS) / N_INUSEBITS; cn++)
pmp->pm_inusemap[cn] = (u_int)-1;
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
/*
* Figure how many free clusters are in the filesystem by ripping
1994-07-17 01:32:06 +04:00
* through the fat counting the number of entries whose content is
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
* zero. These represent free clusters.
*/
pmp->pm_freeclustercount = 0;
for (cn = CLUST_FIRST; cn <= pmp->pm_maxcluster; cn++) {
byteoffset = FATOFS(pmp, cn);
bo = byteoffset % pmp->pm_fatblocksize;
if (!bo || !bp) {
/* Read new FAT block */
if (bp)
brelse(bp);
fatblock(pmp, byteoffset, &bn, &bsize, NULL);
1996-02-09 22:13:39 +03:00
error = bread(pmp->pm_devvp, bn, bsize, NOCRED, &bp);
if (error) {
1995-10-15 18:34:19 +03:00
brelse(bp);
return (error);
1995-10-15 18:34:19 +03:00
}
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
}
if (FAT32(pmp))
readcn = getulong(&bp->b_data[bo]);
else
readcn = getushort(&bp->b_data[bo]);
if (FAT12(pmp) && (cn & 1))
readcn >>= 4;
readcn &= pmp->pm_fatmask;
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
if (readcn == 0)
usemap_free(pmp, cn);
}
brelse(bp);
return (0);
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
}
/*
1994-07-17 01:32:06 +04:00
* Allocate a new cluster and chain it onto the end of the file.
*
* dep - the file to extend
* count - number of clusters to allocate
* bpp - where to return the address of the buf header for the first new
* file block
* ncp - where to put cluster number of the first newly allocated cluster
* If this pointer is 0, do not return the cluster number.
* flags - see fat.h
*
* NOTE: This function is not responsible for turning on the DE_UPDATE bit of
* the de_flag field of the denode and it does not change the de_FileSize
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
* field. This is left for the caller to do.
*/
int
extendfile(dep, count, bpp, ncp, flags)
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
struct denode *dep;
u_long count;
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
struct buf **bpp;
u_long *ncp;
int flags;
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
{
int error;
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
u_long frcn;
u_long cn, got;
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
struct msdosfsmount *pmp = dep->de_pmp;
struct buf *bp;
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
/*
* Don't try to extend the root directory
*/
if (dep->de_StartCluster == MSDOSFSROOT
&& (dep->de_Attributes & ATTR_DIRECTORY)) {
1996-10-13 08:10:34 +04:00
printf("extendfile(): attempt to extend root directory\n");
return (ENOSPC);
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
}
/*
* If the "file's last cluster" cache entry is empty, and the file
* is not empty, then fill the cache entry by calling pcbmap().
*/
fc_fileextends++;
if (dep->de_fc[FC_LASTFC].fc_frcn == FCE_EMPTY &&
dep->de_StartCluster != 0) {
fc_lfcempty++;
error = pcbmap(dep, CLUST_END, 0, &cn, 0);
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
/* we expect it to return E2BIG */
if (error != E2BIG)
return (error);
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
}
while (count > 0) {
/*
1996-02-09 22:13:39 +03:00
* Allocate a new cluster chain and cat onto the end of the
* file. * If the file is empty we make de_StartCluster point
* to the new block. Note that de_StartCluster being 0 is
* sufficient to be sure the file is empty since we exclude
* attempts to extend the root directory above, and the root
* dir is the only file with a startcluster of 0 that has
* blocks allocated (sort of).
*/
if (dep->de_StartCluster == 0)
cn = 0;
else
cn = dep->de_fc[FC_LASTFC].fc_fsrcn + 1;
error = clusteralloc(pmp, cn, count, &cn, &got);
1996-02-09 22:13:39 +03:00
if (error)
return (error);
count -= got;
/*
* Give them the filesystem relative cluster number if they want
* it.
*/
if (ncp) {
*ncp = cn;
ncp = NULL;
}
if (dep->de_StartCluster == 0) {
dep->de_StartCluster = cn;
frcn = 0;
} else {
1996-02-09 22:13:39 +03:00
error = fatentry(FAT_SET, pmp,
dep->de_fc[FC_LASTFC].fc_fsrcn,
0, cn);
if (error) {
clusterfree(pmp, cn, NULL);
return (error);
}
frcn = dep->de_fc[FC_LASTFC].fc_frcn + 1;
}
/*
* Update the "last cluster of the file" entry in the denode's fat
* cache.
*/
fc_setcache(dep, FC_LASTFC, frcn + got - 1, cn + got - 1);
if (flags & DE_CLEAR) {
while (got-- > 0) {
/*
* Get the buf header for the new block of the file.
*/
if (dep->de_Attributes & ATTR_DIRECTORY)
bp = getblk(pmp->pm_devvp, cntobn(pmp, cn++),
pmp->pm_bpcluster, 0, 0);
else {
bp = getblk(DETOV(dep), de_cn2bn(pmp, frcn++),
pmp->pm_bpcluster, 0, 0);
/*
* Do the bmap now, as in msdosfs_write
*/
if (pcbmap(dep,
de_bn2cn(pmp, bp->b_lblkno),
&bp->b_blkno, 0, 0))
bp->b_blkno = -1;
if (bp->b_blkno == -1)
panic("extendfile: pcbmap");
}
clrbuf(bp);
if (bpp) {
*bpp = bp;
bpp = NULL;
} else
bdwrite(bp);
}
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
}
}
return (0);
brought in fixed/renamed/matching MS-DOS FS code, from Jeff Polk <polk@bsdi.com>. His notes are as follows: ------------------------------------------------------------------------------ July 22, 1993 - Changed name of entire package from PCFS to MSDOSFS - Fixed bugs: root directory size in clusters instead of bytes growing directory didn't update in-core size link, symlink, mknod didn't free locked parent (deadlock) lookup returned real error on create and rename instead of EJUSTRETURN rename changed `.' entry in child instead of name entry in parent rename removed `.' entry in child instead of removing entry in parent when moving a directory from one dir to another createde() left new node locked when write of parent failed (deadlock) removede() decremented refcount even on error (rmdir's which failed due to write errors left in-core cache entries inconsistent) changed validation for filesystem to not check for the boot signature since some disks (e.g., mtools) aren't bootable directories are always show current time as modify time (needed for NFS export since DOS never updates dir mod times -- ctime is true create time). - Added support for cookies changes to the readdir() vnode interface (#ifdef __bsdi__) - Punted on the whole problem of inode generation numbers. This means that there's a chance of using a stale file handle to access a new file, but it doesn't appear to be the common case, and I don't see how to generate reasonable generation numbers without changing something on the disk (which is the way the SVR4 filesystem survival kit guys did it). I don't think it would be very safe to change the on-disk format. Jeff Polk (polk@BSDI.COM) ------------------------------------------------------------------------------
1993-08-13 15:35:13 +04:00
}