A brief description of the VFAT File System
		===========================================

Long filename support for, uh, like,  Windows 95 compatibility, or something
============================================================================

This document constitutes a rough description of the VFAT File system;
No specifications were available, and most of this work is based on the
Linux VFAT driver code. The mapping of long file names to short ones (8+3),
is not always intuitive and there are possibly several special cases that
are incorrectly treated by mount_dos.

The extended FAT file system is almost identical to the FAT file system.
The significant change has been the addition of long file names. These long
file names support upto 255 characters and might include spaces and lower
case characters (as opposed to the traditional DOS 8+3 file names).

In the VFAT file system, Microsoft has used extra directory entries for any
files that have long names. (Any file name that legally does fit into the
8+3 scheme does not have extra directory entries. These "extra" directory
entries are called slots. The size of a slot is the same as the size of a 
directory entry, but the format is different. Each slot holds 13 characters
of a file's extended name. The structure of a slot is as follows:

typedef struct slot {
        u_char sl_id;                   /* sequence number for slot  */
        u_char sl_name0_4[10];          /* first 5 characters in name*/
        u_char sl_attr;                 /* attribute byte            */
        u_char sl_reserved;             /* always 0                  */
        u_char sl_chksum;               /* checksum for 8.3 alias    */
        u_char sl_name5_10[12];         /* 6 more characters in name */
        u_char sl_start[2];             /* starting cluster number   */
        u_char sl_name11_12[4];         /* last 2 characters in name */
} slot_t;

Slots must be distinguishable from normal directory entries. To this end
a number of distinctions are made. They are:

        1) The attribute byte for a slot directory entry is always set
           to 0x0f.  This corresponds to an old directory entry with
           attributes of "hidden", "system", "read-only", and "volume
           label".  Most old software will ignore any directory
           entries with the "volume label" bit set.  Real volume label
           entries don't have the other three bits set.
        2) The starting cluster is always set to 0, an impossible
           value for a DOS file.

Since the VFAT file system is backward compatible, it is conceivable that
a DOS user might modify directory entries (he might merely manipulate files
with their short names, without being aware of the long names and also the
slots associated with this file entry. In order to establish that a slot
does indeed belong to a 8+3 directory entry, the following are adhered to:

	(1) Position: Slots immediately preceed their corresponding 
	    8+3 directory entry. Also, each slot has an id which marks
	    its order in the long file name. A typical layout of the
	    directory entries/slots that constitute a long file name
	    would look something like:

           <proceeding files...>
           <slot num: #3, id = 0x43,  "h is long">
           <slot num: #2, id = 0x02,  "xtension whic">
           <slot num: #1, id = 0x01,  "My Big File.E">
           <directory entry, short-name = "MYBIGFIL.EXT">

           Slots are stored from last to first.  Slots are numbered from 
	   1 to N.  The Nth slot is or'ed with 0x40 to mark it as the very
	   last slot.

        (2) Checksum:  The shortname (8+3, including spaces) is scanned and
	    its checksum is computed. Now, in each slot that constitutes its
	    long name the same checksum is also stored. This allows us to 
	    perform consistency checks later on to verify if a sequence of
	    slots does indeed map to a particular short name.
	    The following is the checksum algorithm used on the 8+3 name:

            for (sum = i = 0; i < 11; i++)
               sum = (((sum&1)<<7)|((sum&0xfe)>>1)) + name[i]

        (3) If there is in the final slot, a Unicode NULL (0x0000) is stored
            after the final character.  After that, all unused characters in
            the final slot are set to Unicode 0xFFFF.

NOTE: All the characters that are stored in slots are stored as unicodes.
Each unicode consists of two bytes and there are a set of ugly tables in
"dos_util.c" that are responsible for conversion between ascii and unicode.

A note on the implementation:
=============================

The crux of the vfat file system support resides in the file "dos_util.c".
This has become a dumping ground of sorts, with a preponderance of routines
that allow long file names to be supported.