Unifying dosfsck and mkdosfs documents in common doc directory.

Signed-off-by: Daniel Baumann <mail@daniel-baumann.ch>

doc/ANNOUNCE.mkdosfs

@@ -0,0 +1,41 @@
+Announcing the release of mkdosfs version 0.3b (Yggdrasil)
+
+It seems I didn't get the bug completely fixed in 0.3a.  Some
+borderline cases would still allocate too many sectors for the FAT.
+Again, nothing to worry about, just a nitpick -- this one would only
+in certain cases add one sector per FAT.
+
+Announcing the release of mkdosfs version 0.3a (Yggdrasil)
+
+Fixed a bug which would cause too many sectors to be reserved for the
+FAT (filesystem will still work fine, but have slightly less space
+available).
+
+Announcing the release of mkdosfs version 0.3 (Yggdrasil)
+
+This version correctly handles even very large filesystems, and
+properly supports the modern (3.3+) DOS bootsector format, including a
+message printed on boot attempts.
+
+Peter Anvin
+Yggdrasil Computing, Inc.
+hpa@yggdrasil.com
+
+			    --------------
+
+Announcing the release of mkdosfs version 0.2
+
+
+I've just uploaded mkdosfs to sunsite.unc.edu.  It works in a similar way
+to Remy Card's mke2fs, but creates an MS-DOS file system.
+
+The filename is mkdosfs-0.2.tar.gz.
+
+This second release should fix a small bug that could lead to FAT sizes that
+Linux's dosfs would accept but MS-DOS wouldn't.
+
+The archive contains a manual page, binary and source versions.
+
+
+Dave Hudson
+dave@humbug.demon.co.uk

doc/ChangeLog.dosfsck

@@ -0,0 +1,10 @@
+Changes from version 0 to 1
+===========================
+
+  - fixed an off-by-two error in check.c:check_file
+  - fixed marking clusters bad in fat.c:set_fat
+  - fat.c:reclaim_free was also reclaiming bad clusters.
+  - fixed many incorrect byte sex conversions in check.c and fat.c
+  - -t and -w now require -a or -r
+  - added option -d to drop files.
+  - added option -u to try to "undelete" non-directory files.

doc/ChangeLog.dosfstools-2.x

@@ -0,0 +1,161 @@
+version 2.11
+============
+
+ - all: don't use own llseek() anymore, glibc lseek() does everything we need
+ - dosfsck: lfn.c: avoid segfault
+ - dosfsck: check.c, lfn.c: check for orphaned LFN slots
+ - dosfsck: check.c alloc_rootdir_entry(): set owner of newly alloced clusters
+ - dosfsck: dosfsck.h: better use <byteswap.h> for byte swapping
+ - dosfsck: io.c: added code for real DOS
+ - mkdosfs: raised FAT12_THRESHOLD from 4078 to 4085, introduced MIN_CLUST_32
+ - mkdosfs: fix loop device size
+ - mkdosfs: by default, use FAT32 on devices >= 512MB
+ - mkdosfs: fix a memory leak (blank_sector)
+ - mkdosfs: fix parsing of number of blocks on command line, so that numbers
+   >2G can be used
+ - mkdosfs: add 'b' to getopt() string so this option can be used :)
+ - mkdosfs: fix parsing of -i arg (should be unsigned)
+ - mkdosfs: change default permissions of created images (-C) to 0666 & ~umask
+ - mkdosfs: relax geometry check: if HDIO_GETGEO fails, print a warning and
+   default to H=255,S=63
+ - dosfsck: new option -n (no-op): just check non-interactively, but
+   don't write anything to filesystem
+ - A few #include changes to support compilation with linux 2.6
+   headers (thanks to Jim Gifford <jim@jg555.com>)
+ - dosfsck: remove directory entries pointing to start cluster 0, if they're
+   not "." or ".." entries that should actually point to the root dir
+   (pointed out by Thomas Winkler <twinkler@sysgo.de>)
+ - mkdosfs: new option -h to set number of hidden sectors
+   (thanks to Godwin Stewart <gstewart@spamcop.net>)
+ - all: updated my mail address everywhere...
+
+version 2.10
+============
+
+ - dosfsck: various 64-bit fixes and removed some warnings by Michal
+   Cihar <mcihar@suse.cz>
+ - mkdosfs: better error message if called without parameters (also
+   suggested by Michal)
+
+version 2.9
+===========
+
+ - dosfsck: if EOF from stdin, exit with error code
+ - dosfsck: Fix potential for "Internal error: next_cluster on bad cluster".
+ - dosfsck: When clearing long file names, don't overwrite the dir
+   entries with all zeros, but put 0xe5 into the first byte.
+   Otherwise, some OSes stop reading the directory at that point...
+ - dosfsck: in statistics printed by -v, fix 32bit overflow in number
+   of data bytes.
+ - dosfsck: fix an potential overflow in "too many clusters" check
+ - dosfsck: fix 64bit problem in fat.c (Debian bug #152769)
+ - dosfsck: allow FAT size > 32MB.
+ - dosfsck: allow for only one FAT
+ - dosfsck: with -v, also check that last sector of the filesystem can
+   be read (in case a partition is smaller than the fs thinks)
+ - mkdosfs: add note in manpage that creating bootable filesystems is
+   not supported.
+ - mkdosfs: better error message with pointer to -I if target is a
+   full-disk device.
+
+version 2.8
+===========
+
+ - dosfsck: Fixed endless loop whenever a volume label was present.
+
+version 2.7
+===========
+
+ - dosfsck: Don't check volume label for bad characters, everything
+   seems to be allowed there... Also ignore duplicate names where one of
+   them is a volume label.
+
+version 2.6
+===========
+
+ - mkdosfs: Added correct heads definition for 2.88M floppies if not
+   created via loopback.
+ - dosfsck: If boot sector and its backup are different (FAT32), offer
+   to write the backup to sector 0. (tnx to Pavel Roskin for this)
+ - For 64 bit alpha, struct bootsector in dosfsck.h must be defined
+   with __attribute__((packed)).
+ - mkdosfs now actually accepts -R option. (tnx to David Kerrawn)
+ - Fixed typo in dosfsck boot.c (recognition of boot signature in FSINFO)
+ - Various compilation fixes for 2.4 kernel headers and for ia64.
+
+version 2.5
+===========
+
+ - The llseek() implementation for alpha didn't really work; fixed it.
+
+version 2.4
+===========
+
+ - Fix compiling problem on alpha (made a silly typo...)
+
+version 2.3
+===========
+
+ - mkdosfs: Fixed usage message (printed only "bad address").
+ - both: made man pages and usage statements more consistent.
+ - both: fix llseek function for alpha.
+ - dosfsck: fix reading of unaligned fields in boot sector for alpha.
+ - dosfsck: fixed renaming of files (extension wasn't really written).
+
+version 2.2
+===========
+
+ - Added dosfsck/COPYING, putting dosfsck officially under GPL (Werner
+   and I agree that it should be GPL).
+ - mkdosfs: Allow creation of a 16 bit FAT on filesystems that are too
+   small for it if the user explicitly selected FAT16 (but a warning
+   is printed). Formerly, you got the misleading error message "make
+   the fs a bit smaller".
+ - dosfsck: new option -y as synonym for -y; for compability with
+   other fs checkers, which also accept this option.
+ - dosfsck: Now prints messages similar to e2fsck: at start version
+   and feature list; at end number of files (and directories) and
+   number of used/total clusters. This makes the printouts of *fsck at
+   boot time nicer.
+ - dosfsck: -a (auto repair) now turns on -f (salvage files), too. -a
+   should act as non-destructive as possible, so lost clusters should
+   be assigned to files. Otherwise the data in them might be
+   overwritten later.
+ - dosfsck: Don't drop a directory with lots of bad entries in
+   auto-repair mode for the same reason as above.
+ - dosfsck: avoid deleting the whole FAT32 root dir if something is
+   wrong with it (bad start cluster or the like).
+ - general: also create symlinks {mkfs,fsck}.vfat.8 to the respective
+   real man pages.
+
+version 2.1
+===========
+
+ - Fix some forgotten loff_t's for filesystems > 4GB. (Thanks to
+   <ki@kretz.co.at>).
+ - Fix typo in mkdosfs manpage.
+ - Removed inclusion of <linux/loop.h> from mkdosfs.c; it's unnecessary and
+   caused problems in some environments.
+ - Fix condition when to expect . and .. entries in a directory. (Was
+   wrong for non-FAT32 if first entry in root dir was a directory also.)
+ - Also create mkfs.vfat and fsck.vfat symlinks, so that also
+   filesystems listed with type "vfat" in /etc/fstab can be
+   automatically checked.
+
+version 2.0
+===========
+
+ - merge of mkdosfs and dosfstools in one package
+ - new maintainer: Roman Hodek <Roman.Hodek@informatik.uni-erlangen.de>
+ - FAT32 support in both mkdosfs and dosfsck
+ - VFAT (long filename) support in dosfsck
+ - Support for Atari variant of MS-DOS filesystem in both tools
+ - Working support for big-endian systems in both tools
+ - Better support for loop devices in mkdosfs: usual floppy sizes are
+   detected and media byte etc. set accordingly; if loop fs has no
+   standard floppy size, use hd params
+   (mainly by Giuliano Procida <gpp10@cus.cam.ac.uk>)
+ - Removed lots of gcc warnings
+ - Fixed some minor calculation bugs in mkdosfs.
+
+For change logs previous to 2.0, see the CHANGES files in the subdirectories.

doc/ChangeLog.mkdosfs

@@ -0,0 +1,18 @@
+28th January 1995		H. Peter Anvin (hpa@yggdrasil.com)
+
+	Better algorithm to select cluster sizes on large filesystems.
+	Added bogus boot sector code that on attempts to boot prints a
+	message (which can be chosen at mkdosfs time) and lets the user
+	press any key and try again.  Corrected support for 1.2 Mb
+	floppies.  mkdosfs now generates the extended bootsector
+	(superblock) format of DOS 3.3+, with support for volume ID's and
+	volume labels (volume labels are also written to the root
+	directory, as they should).
+	
+18th February 1994		Dave Hudson (dave@humbug.demon.co.uk)
+
+	Released version 0.2 - clears a bug in the FAT sizing code.
+
+1st September 1993		Dave Hudson (dave@humbug.demon.co.uk)
+
+	Released version 0.1 - ALPHA release of mkdosfs

doc/README.dosfsck

@@ -0,0 +1,60 @@
+dosfsck, version 1
+==================
+
+WARNING: This is ALPHA test software. Use at your own risk.
+
+dosfsck is the Linux equivalent of PC/MS-DOS' CHKDSK. It checks the
+consistency of PC/MS-DOS file systems and optionally tries to repair
+them. The tests dosfsck performs are described in the man page.
+
+dosfsck needs header files from dosfs.9 (or later) to compile.
+
+Before using dosfsck to repair a file system that contains data of any
+value, you should verify that dosfsck is able to correct all reported
+errors. (Except fatal errors and those reported as unfixable, of
+course.) In order to do this, run it with the -V option, e.g.
+
+   dosfsck -V /dev/sda1		(automatic check)
+or dosfsck -V -r /dev/sda1	(interactive check and repair)
+
+dosfsck will perform two passes: in the first pass, inconsistencies are
+detected and a list of changes to correct the problems is generated. In
+the second pass, those changes are applied whenever dosfsck reads data
+from disk. Hence no fixable errors should be reported in the second
+pass if the first pass was successful.
+
+Please notify the author if fixable errors are reported in the second
+pass.
+
+After verifying that dosfsck appears to be able to perform the desired
+operations, either confirm that you want the changes to be performed
+(if dosfsck was started with -r) or re-run dosfsck with the -a option
+(if it was started without -r).
+
+Please send bug reports, comments, flames, etc. to
+almesber@nessie.cs.id.ethz.ch  or  almesber@bernina.ethz.ch
+
+- Werner
+
+FAT32 and LFN support
+=====================
+
+I've finally implemented some of the new features of MS-DOS
+filesystems: FAT32 and long filenames.
+
+FAT32 is automatically detected and of course the different FAT
+structure is handled. (Internally many changes were needed, so 32 bit
+variables for all cluster numbers and 64 bit vars for offsets inside
+the filesystem.) New checks for FAT32 are most notably on the backup
+boot sector and the new info sector. Also the possibility that the
+root directory resides in a cluster chain (instead of in a static
+area) on FAT32 is handled.
+
+dosfscheck also knows about VFAT long filenames now. It parses those
+names and uses them in listings etc. when available. There are also
+some checks on the (cruel) structure of how LFNs are stored and some
+attempts to fix problems.
+
+- Roman <roman@hodek.net>
+
+BTW, version 2 isn't ALPHA anymore :-)

doc/README.dosfstools-2.x

@@ -0,0 +1,60 @@
+
+Atari format support
+====================
+
+Both mkdosfs and dosfsck now can also handle the Atari variation of
+the MS-DOS filesystem format. The Atari format has some minor
+differences, some caused by the different machine architecture (m68k),
+some being "historic" (Atari didn't change some things that M$
+changed).
+
+Both tools automatically select Atari format if they run on an Atari.
+Additionally the -A switch toggles between Atari and MS-DOS format.
+I.e., on an Atari it selects plain DOS format, on any other machine it
+switches to Atari format.
+
+The differences are in detail:
+
+ - Atari TOS doesn't like cluster sizes != 2, so the usual solution
+   for bigger partitions was to increase the logical sector size. So
+   mkdosfs can handle sector sizes != 512 now, you can also manually
+   select it with the -S option. On filesystems larger than approx. 32
+   MB, the sector size is automatically increased (stead of the
+   cluster size) to make the filesystem fit. mkdosfs will always use 2
+   sectors per cluster (also with the floppy standard configurations),
+   except when directed otherwise on the command line.
+
+ - From the docs, all values between 0xfff8 and 0xffff in the FAT mark
+   an end-of-file. However, DOS usually uses 0xfff8 and Atari 0xffff.
+   This seems to be only an consmetic difference. At least TOS doesn't
+   complain about 0xffff EOF marks. Don't know what DOS thinks of
+   0xfff8 :-) Anyway, both tools use the EOF mark common to the
+   system (DOS/Atari).
+
+ - Something similar of the bad cluster marks: On Atari the FAT values
+   0xfff0 to 0xfff7 are used for this, under DOS only 0xfff7 (the
+   others can be normal cluster numbers, allowing 7 more clusters :-)
+   However, both systems usually mark with 0xfff7. Just dosfsck has to
+   interpret 0xfff0...0xfff7 differently.
+
+ - Some fields in the boot sector are interpreted differently. For
+   example, Atari has a disk serial number (used to aid disk change
+   detection) where DOS stores the system name; the 'hidden' field is
+   32 bit for DOS, but 16 bit for Atari, and there's no 'total_sect'
+   field; the 12/16 bit FAT decision is different: it's not based on
+   the number of clusters, but always FAT12 on floppies and FAT16 on
+   hard disks. mkdosfs nows about these differences and constructs the
+   boot sector accordingly.
+
+ - In dosfsck, the boot sector differences also have to known, to not
+   warn about things that are no error on Atari. In addition, most
+   Atari formatting tools fill the 'tracks' and 'heads' fields with 0
+   for hard disks, because they're meaningless on SCSI disks (Atari
+   has/had no IDE). Due to this, the check that they should be
+   non-zero is switched off.
+
+ - Under Atari TOS, some other characters are illegal in filenames:
+   '<', '>', '|', '"', and ':' are allowed, but all non-ASCII chars
+   (codes >= 128) are forbidden.
+
+- Roman <Roman.Hodek@informatik.uni-erlangen.de>

doc/README.mkdosfs

@@ -0,0 +1,50 @@
+mkdosfs - Make DOS file system utilty.
+
+
+I wrote this, partially to complement the dosfsck utility written by Werner
+Almesberger (who graciously gave me some pointers when I asked for some
+advice about writing this code), and also to avoid me having to boot DOS
+just to create data partitions (I use Linux to back up DOS :-) ).
+
+The code is really derived from Remy Card's mke2fs utility - I used this as a
+framework, although all of the file system specific stuff was removed and the
+DOS stuff inserted.  I believe originally mke2fs was based on Linus' mkfs
+code, hence the acknowledgements in the source code.
+
+Neither Remy nor Linus have had any involvement with mkdosfs, so if there are
+any bugs they're almost certainly "all my own work".
+
+The code has been available for ftp since 1st September 1993, and I have yet
+to receive any bug reports from users.  I don't know of any bugs, but if you
+do find a bug or have any constructive comments, please mail me!
+
+The only bug I found with version 0.1 was an obscure fault that could lead
+to an invalid (for MS-DOS, not Linux's dos fs) number of sectors used in the
+file allocation table(s).
+
+
+Dave Hudson
+dave@humbug.demon.co.uk
+
+
+FAT32 support
+=============
+
+mkdosfs now can also create filesystems in the new FAT32 format. To do
+this, give mkdosfs a "-F 32" option. FAT32 isn't selected
+automatically (yet), even if very large clusters are needed with
+FAT16. With FAT32 you have two additional options, -R to select the
+number of reserved sectors (usually 32), and -b to select the location
+of the backup boot sector (default 6). Of course such a backup is
+created, as well as the new info sector. On FAT32, the root directory
+is always created as a cluster chain. Sorry, there's no switch to
+generate an old static root dir.
+
+One bigger bug fix besides FAT32 was to reject filesystems that need a
+16 bit FAT to fit all possible clusters, but the bigger FAT needs some
+more sectors, so the total number of clusters drop below the border
+where MS-DOS expects a 12 bit FAT. So such filesystems would be FAT16,
+but interpreted as FAT32 by DOS. The fix is to reduce filesystem size
+a bit.
+
+- Roman <roman@hodek.net>

doc/TODO.dosfstools-2.x

@@ -0,0 +1,15 @@
+                                                    -*- mode: indented-text -*-
+
+ - dosfsck: Better checking of file times: ctime <= mtime <= atime
+
+ - mkdosfs: If /etc/bootsect.dos (or similar) exists, use it as a
+   template for generating boot sectors. This way, you can, e.g., make
+   bootable DOS disks.
+
+   Addendum: Don't know if that's so wise... There are really many
+   variants of DOS/Windows bootcode out in the wild, and the code is
+   proprietary, too.
+
+ - dosfsck: read-only sector test (-t without -a or -r); just print
+   out errors.
+