CW/etc/mtools.conf is the system-wide configuration file, and CW~/.mtoolsrc is the user's private configuration file.

On some systems, the system-wide configuration file is called CW/etc/defaults/mtools.conf instead.

The configuration files is made up of sections. Each section starts with a keyword identifying the section followed by a colon. Then follow variable assignments and flags. Variable assignments take the following form:

name=value

Lines starting with a hash (CW#) are comments. Newline characters are equivalent to whitespace (except where ending a comment). The configuration file is case insensitive, except for item enclosed in quotes (such as filenames).

For most platforms, mtools contains reasonable compiled-in defaults for physical floppy drives. Thus, you usually don't need to bother with the configuration file, if all you want to do with mtools is to access your floppy drives. On the other hand, the configuration file is needed if you also want to use mtools to access your hard disk partitions and dosemu image files.

Global flags may be set to 1 or to 0.

The following global flags are recognized:

CWMTOOLS_SKIP_CHECK

If this is set to 1, mtools skips most of its sanity checks. This is needed to read some Atari disks which have been made with the earlier ROMs, and which would not be recognized otherwise.

CWMTOOLS_FAT_COMPATIBILITY

If this is set to 1, mtools skips the fat size checks. Some disks have a bigger FAT than they really need to. These are rejected if this option is not set.

CWMTOOLS_LOWER_CASE

If this is set to 1, mtools displays all-upper-case short filenames as lowercase. This has been done to allow a behavior which is consistent with older versions of mtools which didn't know about the case bits.

CWMTOOLS_NO_VFAT

If this is set to 1, mtools won't generate VFAT entries for filenames which are mixed-case, but otherwise legal dos filenames. This is useful when working with DOS versions which can't grok VFAT longnames, such as FreeDos.

CWMTOOLS_DOTTED_DIR

In a wide directory, prints the short name with a dot instead of spaces separating the basename and the extension.

CWMTOOLS_NAME_NUMERIC_TAIL

If this is set to one (default), generate numeric tails for all long names (~1). If set to zero, only generate numeric tails if otherwise a clash would have happened.

CWMTOOLS_TWENTY_FOUR_HOUR_CLOCK

If 1, uses the European notation for times (twenty four hour clock), else uses the UK/US notation (am/pm)

Example: Inserting the following line into your configuration file instructs mtools to skip the sanity checks:

  MTOOLS_SKIP_CHECK=1

Global variables may also be set via the environment:

  export MTOOLS_SKIP_CHECK=1

Global string variables may be set to any value:

CWMTOOLS_DATE_STRING

The format used for printing dates of files. By default, is dd-mm-yyyy.

Per drive flags and values may be described in a drive section. A drive section starts with CWdrive "driveletter" :

Then follow variable-value pairs and flags.

This is a sample drive description:

  drive a:
    file="/dev/fd0" use_xdf=1

Geometry information describes the physical characteristics about the disk. Its has three purposes:

formatting

The geometry information is written into the boot sector of the newly made disk. However, you may also describe the geometry information on the command line. See section mformat, for details.

filtering

On some Unices there are device nodes which only support one physical geometry. For instance, you might need a different node to access a disk as high density or as low density. The geometry is compared to the actual geometry stored on the boot sector to make sure that this device node is able to correctly read the disk. If the geometry doesn't match, this drive entry fails, and the next drive entry bearing the same drive letter is tried. See section multiple descriptions, for more details on supplying several descriptions for one drive letter.

If no geometry information is supplied in the configuration file, all disks are accepted. On Linux (and on Sparc) there exist device nodes with configurable geometry (CW/dev/fd0, CW/dev/fd1 etc), and thus filtering is not needed (and ignored) for disk drives. (Mtools still does do filtering on plain files (disk images) in Linux: this is mainly intended for test purposes, as I don't have access to a Unix which would actually need filtering).

If you do not need filtering, but want still a default geometry for mformatting, you may switch off filtering using the CWmformat_only flag.

If you want filtering, you should supply the CWfilter flag. If you supply a geometry, you must supply one of both flags.

initial geometry

On devices that support it (usually floppy devices), the geometry information is also used to set the initial geometry. This initial geometry is applied while reading the boot sector, which contains the real geometry. If no geometry information is supplied in the configuration file, or if the CWmformat_only flag is supplied, no initial configuration is done.

On Linux, initial geometry is not really needed, as the configurable devices are able to auto-detect the disk type accurately enough (for most common formats) to read the boot sector.

Wrong geometry information may lead to very bizarre errors. That's why I strongly recommend that you add the CWmformat_only flag to your drive description, unless you really need filtering or initial geometry.

The following geometry related variables are available:

CWcylinders

CWtracks The number of cylinders. (CWcylinders is the preferred form, CWtracks is considered obsolete)

CWheads

The number of heads (sides).

CWsectors

The number of sectors per track.

Example: the following drive section describes a 1.44M drive:

  drive a:
      file="/dev/fd0H1440"
      fat_bits=12
      cylinders=80 heads=2 sectors=18
      mformat_only

The following shorthand geometry descriptions are available:

CW1.44m

high density 3 1/2 disk. Equivalent to: CWfat_bits=12 cylinders=80 heads=2 sectors=18

CW1.2m

high density 5 1/4 disk. Equivalent to: CWfat_bits=12 cylinders=80 heads=2 sectors=15

CW720k

double density 3 1/2 disk. Equivalent to: CWfat_bits=12 cylinders=80 heads=2 sectors=9

CW360k

double density 5 1/4 disk. Equivalent to: CWfat_bits=12 cylinders=40 heads=2 sectors=9

The shorthand format descriptions may be amended. For example, CW360k sectors=8 describes a 320k disk and is equivalent to: CWfat_bits=12 cylinders=40 heads=2 sectors=8

Moreover, the following flags are available:

CWsync

All i/o operations are done synchronously

CWnodelay

The device or file is opened with the O_NDELAY flag. This is needed on some non-Linux architectures.

CWexclusive

The device or file is opened with the O_EXCL flag. On Linux, this ensures exclusive access to the floppy drive. On most other architectures, and for plain files it has no effect at all.

The following general purpose drive variables are available. Depending to their type, these variables can be set to a string (file, precmd) or an integer (all others)

CWfile

The name of the file or device holding the disk image. This is mandatory. The file name should be enclosed in quotes.

CWpartition

Tells mtools to treat the drive as a partitioned device, and to use the given partition. Only primary partitions are accessible using this method, and they are numbered from 1 to 4. For logical partitions, use the more general CWoffset variable. The CWpartition variable is intended for removable media such as Syquests, ZIP drives, and magneto-optical disks. Although traditional DOS sees Syquests and magneto-optical disks as CWgiant floppy disks which are unpartitioned, OS/2 and Windows NT treat them like hard disks, i.e. partioned devices. The CWpartition flag is also useful DOSEMU hdimages. It is not recommended for hard disks for which direct access to partitions is available through mounting.

CWoffset

Describes where in the file the MS-DOS filesystem starts. This is useful for logical partitions in DOSEMU hdimages, and for ATARI ram disks. By default, this is zero, meaning that the filesystem starts right at the beginning of the device or file.

CWfat_bits

The number of FAT bits. This may be 12 or 16. This is very rarely needed, as it can almost always be deduced from information in the boot sector. On the contrary, describing the number of fat bits may actually be harmful if you get it wrong. You should only use it if mtools gets the autodetected number of fat bits wrong, or if you want to mformat a disk with a weird number of fat bits.

CWprecmd

On some variants of Solaris, it is necessary to call 'volcheck -v' before opening a floppy device, in order for the system to notice that there is indeed a disk in the drive. CWprecmd="volcheck -v" in the drive clause establishes the desired behavior.

CWblocksize

This parameter represents a default block size to be always used on this device. All I/O is done with multiples of this block size, independantly of the sector size registered in the filesystem's boot sector. This is useful for character devices whose sector size is not 512, such as for example CD Rom drives on Solaris.

Only the CWfile variable is mandatory. The other parameters may be left out. In that case a default value or an autodetected value is used.

A flag can either be set to 1 (enabled) or 0 (disabled). If the value is ommitted, it is enabled. For example, CWscsi is equivalent to CWscsi=1

CWnolock

Instruct mtools to not use locking on this drive. This is needed on systems with buggy locking semantics. However, enabling this makes operation less safe in cases where several users may access the same drive at the same time.

CWscsi

When set to 1, this option tells mtools to use raw SCSI I/O instead of the standard read/write calls to access the device. Currently, this is supported on HP/UX, Solaris and SunOs. This is needed because on some architectures, such as SunOs or Solaris, PC media can't be accessed using the CWread and CWwrite syscalls, because the OS expects them to contain a Sun specific "disk label".

As raw Scsi access always uses the whole device, you need to specify the "partition" flag in addition

On some architectures, such as Solaris, mtools needs root privileges to be able to use the CWscsi option. Thus mtools should be installed set uid root on Solaris if you want to access Zip/Jaz drives. Thus, if the CWscsi flag is given, CWprivileged is automatically implied, unless explicitly disabled by CWprivileged=0

Mtools uses its root privileges to open the device, and to issue the actual SCSI I/O calls. Moreover, root privileges are only used for drives described in a system-wide configuration file such as CW/etc/mtools.conf, and not for those described in CW~/.mtoolsrc or CW$MTOOLSRC.

CWprivileged

When set to 1, this instructs mtools to use its set-uid and set-gid privileges for opening the given drive. This option is only valid for drives described in the system-wide configuration files (such as CW/etc/mtools.conf, not CW~/.mtoolsrc or CW$MTOOLSRC). Obviously, this option is also a no op if mtools is not installed setuid or setgid. This option is implied by 'scsi=1', but again only for drives defined in system-wide configuration files. Privileged may also be set explicitely to 0, in order to tell mtools not to use its privileges for a given drive even if CWscsi=1 is set.

Mtools only needs to be installed setuid if you use the CWprivileged or CWscsi drive variables. If you do not use these options, mtools works perfectly well even when not installed setuid root.

CWvold

Instructs mtools to interpret the device name as a vold identifier rather than as a filename. The vold identifier is translated into a real filename using the CWmedia_findname() and CWmedia_oldaliases() functions of the CWvolmgt library. This flag is only available if you configured mtools with the CW--enable-new-vold option before compilation.

CWswap

Consider the media as a word-swapped Atari disk.

CWuse_xdf

If this is set to a non-zero value, mtools also tries to access this disk as an XDF disk. XDF is a high capacity format used by OS/2. This is off by default. See section XDF, for more details.

CWmformat_only

Tells mtools to use the geometry for this drive only for mformatting and not for filtering.

CWfilter

Tells mtools to use the geometry for this drive both for mformatting and filtering.

CWremote

Tells mtools to connect to floppyd (see section floppyd).

It is possible to supply multiple descriptions for a drive. In that case, the descriptions are tried in order until one is found that fits. Descriptions may fail for several reasons:

1.

because the geometry is not appropriate,

2.

because there is no disk in the drive,

3.

or because of other problems.

Multiple definitions are useful when using physical devices which are only able to support one single disk geometry. Example:

  drive a: file="/dev/fd0H1440" 1.44m
  drive a: file="/dev/fd0H720" 720k

This instructs mtools to use /dev/fd0H1440 for 1.44m (high density) disks and /dev/fd0H720 for 720k (double density) disks. On Linux, this feature is not really needed, as the /dev/fd0 device is able to handle any geometry.

You may also use multiple drive descriptions to access both of your physical drives through one drive letter:

  drive z: file="/dev/fd0"
  drive z: file="/dev/fd1"

With this description, CWmdir z: accesses your first physical drive if it contains a disk. If the first drive doesn't contain a disk, mtools checks the second drive.

When using multiple configuration files, drive descriptions in the files parsed last override descriptions for the same drive in earlier files. In order to avoid this, use the CWdrive+ or CW+drive keywords instead of CWdrive. The first adds a description to the end of the list (i.e. it will be tried last), and the first adds it to the start of the list.

If you live in the USA, in Western Europe or in Australia, you may skip this section.

DOS uses a different character code mapping than Unix. 7-bit characters still have the same meaning, only characters with the eight bit set are affected. To make matters worse, there are several translation tables available depending on the country where you are. The appearance of the characters is defined using code pages. These code pages aren't the same for all countries. For instance, some code pages don't contain upper case accented characters. On the other hand, some code pages contain characters which don't exist in Unix, such as certain line-drawing characters or accented consonants used by some Eastern European countries. This affects two things, relating to filenames:

upper case characters

In short names, only upper case characters are allowed. This also holds for accented characters. For instance, in a code page which doesn't contain accented uppercase characters, the accented lowercase characters get transformed into their unaccented counterparts.

long file names

Micro$oft has finally come to their senses and uses a more standard mapping for the long file names. They use Unicode, which is basically a 32 bit version of ASCII. Its first 256 characters are identical to Unix ASCII. Thus, the code page also affects the correspondence between the codes used in long names and those used in short names

Mtools considers the filenames entered on the command line as having the Unix mapping, and translates the characters to get short names. By default, code page 850 is used with the Swiss uppercase/lowercase mapping. I chose this code page, because its set of existing characters most closely matches Unix's. Moreover, this code page covers most characters in use in the USA, Australia and Western Europe. However, it is still possible to chose a different mapping. There are two methods: the CWcountry variable and explicit tables.

The CWCOUNTRY variable is recommended for people which also have access to MS-DOS system files and documentation. If you don't have access to these, I'd suggest you'd rather use explicit tables instead.

Syntax:

CWCOUNTRY="countryCW[,[codepageCW], country-fileCW]"

This tells mtools to use a Unix-to-DOS translation table which matches codepage and an lowercase-to-uppercase table for country and to use the country-file file to get the lowercase-to-uppercase table. The country code is most often the telephone prefix of the country. Refer to the DOS help page on "country" for more details. The codepage and the country-file parameters are optional. Please don't type in the square brackets, they are only there to say which parameters are optional. The country-file file is supplied with MS-DOS, and is usually called CWCOUNTRY.SYS, and stored in the CWC:\DOS directory. In most cases you don't need it, as the most common translation tables are compiled into mtools. So, don't worry if you run a Unix-only box which lacks this file.

If codepage is not given, a per country default code page is used. If the country-file parameter isn't given, compiled-in defaults are used for the lowercase-to-uppercase table. This is useful for other Unices than Linux, which may have no CWCOUNTRY.SYS file available online.

The Unix-to-DOS are not contained in the CWCOUNTRY.SYS file, and thus mtools always uses compiled-in defaults for those. Thus, only a limited amount of code pages are supported. If your preferred code page is missing, or if you know the name of the Windows 95 file which contains this mapping, could you please drop me a line at CWalain@linux.lu.

The CWCOUNTRY variable can also be set using the environment.

Translation tables may be described in line in the configuration file. Two tables are needed: first the DOS-to-Unix table, and then the Lowercase-to-Uppercase table. A DOS-to-Unix table starts with the CWtounix keyword, followed by a colon, and 128 hexadecimal numbers. A lower-to-upper table starts with the CWfucase keyword, followed by a colon, and 128 hexadecimal numbers.

The tables only show the translations for characters whose codes is greater than 128, because translation for lower codes is trivial.

Example:

 tounix:
   0xc7 0xfc 0xe9 0xe2 0xe4 0xe0 0xe5 0xe7 
   0xea 0xeb 0xe8 0xef 0xee 0xec 0xc4 0xc5 
   0xc9 0xe6 0xc6 0xf4 0xf6 0xf2 0xfb 0xf9 
   0xff 0xd6 0xdc 0xf8 0xa3 0xd8 0xd7 0x5f 
   0xe1 0xed 0xf3 0xfa 0xf1 0xd1 0xaa 0xba 
   0xbf 0xae 0xac 0xbd 0xbc 0xa1 0xab 0xbb 
   0x5f 0x5f 0x5f 0x5f 0x5f 0xc1 0xc2 0xc0 
   0xa9 0x5f 0x5f 0x5f 0x5f 0xa2 0xa5 0xac 
   0x5f 0x5f 0x5f 0x5f 0x5f 0x5f 0xe3 0xc3 
   0x5f 0x5f 0x5f 0x5f 0x5f 0x5f 0x5f 0xa4 
   0xf0 0xd0 0xc9 0xcb 0xc8 0x69 0xcd 0xce 
   0xcf 0x5f 0x5f 0x5f 0x5f 0x7c 0x49 0x5f 
   0xd3 0xdf 0xd4 0xd2 0xf5 0xd5 0xb5 0xfe 
   0xde 0xda 0xd9 0xfd 0xdd 0xde 0xaf 0xb4 
   0xad 0xb1 0x5f 0xbe 0xb6 0xa7 0xf7 0xb8 
   0xb0 0xa8 0xb7 0xb9 0xb3 0xb2 0x5f 0x5f 

 fucase:
   0x80 0x9a 0x90 0xb6 0x8e 0xb7 0x8f 0x80 
   0xd2 0xd3 0xd4 0xd8 0xd7 0xde 0x8e 0x8f 
   0x90 0x92 0x92 0xe2 0x99 0xe3 0xea 0xeb 
   0x59 0x99 0x9a 0x9d 0x9c 0x9d 0x9e 0x9f 
   0xb5 0xd6 0xe0 0xe9 0xa5 0xa5 0xa6 0xa7 
   0xa8 0xa9 0xaa 0xab 0xac 0xad 0xae 0xaf 
   0xb0 0xb1 0xb2 0xb3 0xb4 0xb5 0xb6 0xb7 
   0xb8 0xb9 0xba 0xbb 0xbc 0xbd 0xbe 0xbf 
   0xc0 0xc1 0xc2 0xc3 0xc4 0xc5 0xc7 0xc7 
   0xc8 0xc9 0xca 0xcb 0xcc 0xcd 0xce 0xcf 
   0xd1 0xd1 0xd2 0xd3 0xd4 0x49 0xd6 0xd7 
   0xd8 0xd9 0xda 0xdb 0xdc 0xdd 0xde 0xdf 
   0xe0 0xe1 0xe2 0xe3 0xe5 0xe5 0xe6 0xe8 
   0xe8 0xe9 0xea 0xeb 0xed 0xed 0xee 0xef 
   0xf0 0xf1 0xf2 0xf3 0xf4 0xf5 0xf6 0xf7 
   0xf8 0xf9 0xfa 0xfb 0xfc 0xfd 0xfe 0xff 

The first table maps DOS character codes to Unix character codes. For example, the DOS character number 129. This is a u with to dots on top of it. To translate it into Unix, we look at the character number 1 in the first table (1 = 129 - 128). This is 0xfc. (Beware, numbering starts at 0). The second table maps lower case DOS characters to upper case DOS characters. The same lower case u with dots maps to character 0x9a, which is an uppercase U with dots in DOS.

If an existing MS-DOS name contains Unicode character greater than 256, these are translated to underscores or to characters which are close in visual appearance. For example, accented consonants are translated into their unaccented counterparts. This translation is used for mdir and for the Unix filenames generated by mcopy. Linux does support Unicode too, but unfortunately too few applications support it yet to bother with it in mtools. Most importantly, xterm can't display Unicode yet. If there is sufficient demand, I might include support for Unicode in the Unix filenames as well.

Caution: When deleting files with mtools, the underscore matches all characters which can't be represented in Unix. Be careful with mdel!

The configuration files are parsed in the following order:

1.

compiled-in defaults

2.

CW/etc/mtools.conf

3.

CW/etc/mtools This is for backwards compatibility only, and is only parsed if CWmtools.conf doesn't exist.

4.

CW~/.mtoolsrc.

5.

CW$MTOOLSRC (file pointed by the CWMTOOLSRC environmental variable)

Options described in the later files override those described in the earlier files. Drives defined in earlier files persist if they are not overridden in the later files. For instance, drives A and B may be defined in CW/etc/mtools.conf and drives C and D may be defined in CW~/.mtoolsrc However, if CW~/.mtoolsrc also defines drive A, this new description would override the description of drive A in CW/etc/mtools.conf instead of adding to it. If you want to add a new description to a drive already described in an earlier file, you need to use either the CW+drive or CWdrive+ keyword.

The syntax described herein is new for version CWmtools-3.0. The old line-oriented syntax is still supported. Each line beginning with a single letter is considered to be a drive description using the old syntax. Old style and new style drive sections may be mixed within the same configuration file, in order to make upgrading easier. Support for the old syntax will be phased out eventually, and in order to discourage its use, I purposefully omit its description here.