Frequently asked question about the Linux floppy driver
Most of the floppy utilities mentioned in this FAQ can be found in the fdutils package. For others, consult the lsm or archie. Alpha releases of fdutils are named fdutils-version-YearMonthDay..diff.gz
- How do I access floppy drives in Linux?
- How should I report bugs?
- I can't fdformat the new formats.
- Mtools says "12 bit FAT on a: sure?" when trying to access an ED disk (or a similar message) when formatting a disk)
- Mtools says "fat_read: Wrong FAT encoding 12 16" (or similar) when accessing a disk formatted with the old mtools.
- How should I describe the new 2m formats in /etc/mtools.conf?
- How should I describe the Xdf formats in /etc/mtools.conf?
- Which minor device number should I use for 2m formats?
- I have an IBM Thinkpad :-(
- What are the minor device numbers for the floppy device nodes?
- What are the recommended names for the floppy device nodes?
- My drive doesn't detect disk changes / When I type mdir a:, I get the directory listing of the previous disk.
- I get "No such device or address" errors for various floppy related commands.
- I have a HP Omnibook 6000 :(
- I can't get my floppy drive going, is there another method to install?
- Is it possible to use two floppy disk controllers under Linux?
- Is it possible to boot off these special format disks?
- I get "Unable to allocate DMA memory" messages when trying to use the floppy.
- I have a Micron Millenia Transport.
How do I access floppy drives in Linux?
Either use mtools or mount the disk. Mtools can be found here.
It is a collection of utilities to access DOS disks. These utilities behave mostly like their DOS counterparts, i.e. mdir is like dir, mcopy is like copy, etc. Specifics are in the manpages.
To mount a disk on /mnt, use the following command lines:
mount -t msdos /dev/fd0 /mnt
The directory /mnt must already exist:
mkdir /mnt
How should I report bugs?
If you have a question or a bug report about the floppy driver, mail me at alain @ linux.lu . If you post to the news, use preferably one of the groups comp.os.linux.help (for questions) or comp.os.linux.hardware (for bug reports). As the volume in these groups is rather high, be sure to include the word "floppy" (or "FLOPPY") in the subject line.
- In case of a kernel panic, or OOPS, please note the EIP, and the values on the stack (some of these are the calling functions). Then look up those values in your zSystem.map (or System.map)
- In case of a lock-up (operation never completes), try to figure out what operations are needed to reproduce it. (If you don't succeed in reproducing it, please mention the floppy operations you remember.)
- If you can't access a disk, include the output of
into your bug report.floppycontrol -pP --printfdcstate
- If something weird happens during boot (lock-up, no floppy drives accessible, trouble with unrelated hardware), include a listing of the messages at boot. (You can obtain this with dmesg. If the boot doesn't complete only note the most important messages down [few lines before the crash, and floppy related lines]).
In case of I/O errors, first switch the error reporting threshold to 0. If that doesn't yield any more error messages, switch on full debugging with
floppycontrol --debug
andfloppycontrol --reporting 0
. Note thatfloppycontrol --debug
generates a huge output. Thus, try first with--reporting
. Please note also that both commands are drive specific, and that the default drive is /dev/fd0. So, if you experience a problem on /dev/fd1, usefloppycontrol --debug -d/dev/fd1
.Then include the kernel messages as well as the application messages with your bug report. If you're using X windows, you won't see the kernel messages appear in your xterm window. Use dmesg to get them. dmesg includes all messages since boot (or as many as fit in the buffer, if there are too many messages). Use dmesg also before executing the command, in order to find out which messages were generated by the command, and which ones were generated by earlyer commands. Alternatively, you may get the user-printk module which allows you to insert "comments" into the stream of console messages.
- Be aware that after a crash, no user program runs, not even syslog. If you direct your kernel messages only to syslog, you may see nothing at all in the log file after rebooting. Thus, I recommend directing the messages also to the console, and noting them down. If there are more than a few lines, note only the last few lines, and the addresses of EIP and the stack trace, if applicable. Obviously, this only applies when investigating bugs which lead to a crash. If no crash happens (like for instance read errors), syslog is actually an useful and beneficial feature, as it is able to keep far more messages than fit into dmesg's buffer.
- If you have other patches in the kernel, report them (especially ftape). If it is easy enough to undo these other patches, please try whether the problem happens without them.
- Be sure to include complete command lines of the floppy commands which led to the problem, as well as a history of the disk changes.
- Is the problem repeatable?
- If you notice that the problem is very time dependent, try to find out if there is any correlation between the occurrence of the problem, and the state of the drive LED.
- If the problem is that the drive becomes inaccessible, but
everything else just works fine, try to find out the following:
- Are all floppy drives inaccessible?
- Does the drive LED stay on?
- Does the drive make any noises?
- Do all floppy related commands fail, or only those that try to
actually read or write to the drive. In particular, what does
floppycontrol -P
do?
- Give some description of your computer! Is it a Laptop? A particularly slow computer (386 SX)? PCI based?
- If you can't supply all these items, go ahead anyways, I want just to make sure you aren't forgetting anything.
I can't fdformat the new formats.
Fdformat is considered obsolete. Use superformat instead.
Mtools says "12 bit FAT on a: sure?" when trying to access an ED disk (or a similar message) when formatting a disk
Some ED (extra density) disks use 16 bit FATS. The FAT type (12 or 16 bit) is described in the configuration file (/etc/mtools or ~/.mtoolsrc). It is the first number. Example:A /dev/fd0 12 0 0 0
^^ 12 bit FAT
12 | means | 12 bit FAT |
16 | means | 16 bit FAT |
0 | means | "use whatever is appropriate" |
-12 | means | 12 bit FAT, even if it looks fishy |
-16 | means | 16 bit FAT, even if it looks fishy |
CAUTION: If you have an /etc/mtools file AND an ~/.mtoolsrc file, the ~/.mtoolsrc file overrides /etc/mtools. The home directory used for ~/.mtoolsrc is $HOME, if that is undefined, it is derived from $LOGNAME, and if that's undefined too, it is derived from getlogin, and finally from getuid.
There are also compiled-in defaults, which are used if neither /etc/mtools, nor ~/.mtoolsrc are found.
Mtools says "fat_read: Wrong FAT encoding 12 16?" (or similar) when accessing a disk formatted with the old mtools.
mtools does some sanity checks on the size of the FAT. Unfortunately, these are too strict for certain disks formatted with the old mtools, and for hard disk partitions formatted by DOS 6, and probably for other conditions as well. Define the environment variable MTOOLS_FAT_COMPATIBILITY to override this check.
(to do this, type setenv MTOOLS_FAT_COMPATIBILITY 1
in csh or tcsh, and
export MTOOLS_FAT_COMPATIBILITY=1
in sh/bash/zsh/...)
How should I describe the new 2m formats in /etc/mtools?
Support for both disk drives a: and b: for all formats (except Xdf) is already built into the new version of mtools (3.9.10). Definitions for new drive letters no longer override these default definitions.
Just for the record, here are definitions to mimic the standard behavior:
drive a: file="/dev/fd0"
drive b: file="/dev/fd1"
How should I describe the Xdf formats in /etc/mtools?
The following /etc/mtools lines allow to access all disks, including Xdf:
drive a: file="/dev/fd0" use_xdf=1
drive b: file="/dev/fd1" use_xdf=1
However, the use_xdf flag slightly slows down initial access to non Xdf disks.
Which minor device number should I use for 2m formats?
2m formats don't have a minor number associated with them, they can only be accessed using the "generic format" devices (i.e. /dev/fd0, /dev/fd1).
I have an IBM Thinkpad :-(
Don't panic. Add the line 'floppy=thinkpad' to your lilo boot parameters. This can be entered at the lilo prompt after the name of the kernel image. It can also be specified in the lilo.conf file, by adding the following line:
append = "floppy=thinkpad"
With some IBM thinkpads, it's also possible to make their floppy drive work by toggling the "FLOPPY" option in CMOS from "auto" to "on". However, apparently this CMOS option is not available on all models.
On some models, both options may be required, on others, none are required.
More info about IBM Tinkpads on Linux can be found here.
What are the minor device numbers for the floppy device nodes?
The major device number for the floppy drives is 2. The minor device number is calculated using the following formula:
minor_device = format_nr * 4 + 128 * fdc_nr + unit_nr
(fdc_nr identifies the floppy disk controller, and unit_nr identifies which drive on the floppy disk controller to be used) format_nr identifies the format (see the README file in fdutils for more details). If format_nr is 0, the device is an autodetection device. Use the MAKEFLOPPIES script included in the fdutils package to automatically create the needed devices.
What are the recommended names for the floppy device nodes?
Floppy drives are named fd drive_nr type max_capacity.
Drive_nr identifies the drive. It ranges from 0 to 3 for drives connected to the first controller, and from 4 to 7 for drives connected to the second controller.
Type is the type of the drive or media. The exact meaning of this (drive or media) is still being discussed. Traditionally, this letter described the type of the drive (density, 5 1/4 or 3 1/2). Lower case letters described 5 1/4 drives, and upper case letters describe 3 1/2 drives. h or H meant high density drives, d or D meant double density drives.
Initially this letter described the type of the drive and not the type of the media in the drive. For instance, a format meant for accessing 720k double density disks in a high density drive was called h720k and not d720k. The reason for this is that for 5 1/4 drives the parameters for accessing a disk do not only depend on the format of the disk, but also on the drive type. The format of the disk is already implied by the capacity, and so the type letter is used to describe the type of the drive.
However, this dependency on the drive type only exists for 5 1/4 drives. For 3 1/2 drives, this is not needed, and as 3 1/2 drives are more frequent, people started forgetting about this, and understood that the letter described the media format. When support for extra density drives was added to the kernel, people still used the name H1440 for accessing high density disks using their extra density drive. By the way, several distribution use this scheme.
In order to calm down the confusion, we propose to make the type letter again describe the type of the drive. However, all 3 1/2 drives are now described by a unique type letter: "u". This means universal (one letter fits all 3 1/2 formats). Moreover the u resembles the Greek letter mu, as in microfloppy.
So the recommended name for a device node allowing to read 1440K disks in a 3 1/2 drive (no matter its density) is called u1440. Note that this is now lowercase.
However, this new convention is still subject to discussion.
Capacity is the capacity of the media in K bytes. Fortunately, this leads to no ambiguity.
Example:
A device node allowing to use 720K 5 1/4 floppies in a high density drive connected as first unit to the second controller is called /dev/fd4h720
Right now, MAKEFLOPPIES supports all three conventions (media, drive, and "u"), you may chose amongst them using command line switches.
My drive doesn't detect disk changes / When I type mdir a:, I get the directory listing of the previous disk.
This means that there is a problem with the disk change line. This problem may have several reasons:
- The disk change line is near the edge of the cable, and is the first line to suffer if the cable is not inserted straight. Press gently on the connectors of your floppy disk cable, in order to ensure that all wires make contact.
- Because the disk change line is near the edge of the cable, it is also the first line to suffer if the cable is damaged. If necessary, buy a new cable, they cost less than five dollars.
- On some drives, the disk change line may be chosen by jumper. Make sure that your floppy controller board and your drive agree which line is the disk change line. The usual location is line 34.
- Some older drives don't support the disk change line at all. In this case, you have to configure the floppy driver to work around the problem. Use 'floppycontrol --broken_dcl' to do this. Now the floppy driver assumes that the disk is changed whenever the device node is closed and then re-opened. This is less efficient, as it results in many useless cache flushes, so use it only if you really need it. This command has to be issued after each reboot, so I suggest you put it into your /etc/rc files.
I get "No such device or address" errors for various floppy related commands.
This message can mean one of several things:
- No floppy driver is compiled into the kernel, and no floppy module is inserted. Kerneld may fail to automatically insert the floppy module when you upgraded your kernel without upgrading the module and moving it to a place where kerneld finds it.
- The drive doesn't exist (mistyped drive name?)
- No disk is in the drive.
- A problem with the disk change line: The disk change line is not only used to detect disk changes, but also to tell whether there is a disk in the drive at all.) See question 13 for fixes for disk change line problem.
- The floppy geometry is unknown for the getfdprm or fdformat programs. If you format a disk, rather use superformat.
I have a HP Omnibook 6000 :(
These laptops are supplied with a floppy-drive & fdc combo which doesn't support DMA :( Just addfloppy=nodma
to your Lilo
command line.
More info about the Omnibook and linux can be found here.
I can't get my floppy drive going, is there another method to install?
You may also install via hard disk and initrd. You need a big enough hard disk to do this (and it takes a little bit more time than the usual method). The following method presumes that it is possible to access the disk using DOS.
- Leave a small DOS partition (with enough spare space to hold the tar files from a few install disks). The bigger that partition is, the less come and goes from DOS are needed later. However, the bigger it is, the less space is available for Linux itself.
- On a desktop computer, using Linux, make a filesystem on a
smallish spare partition (for instance on your swap partition). This
filesystem should contain the files usually found on the install root
disks. Add a '/linuxrc' file which contains the following lines:
#!/bin/sh . etc/rc /bin/sh
- On the desktop computer, unmount the filesystem, and dd it to a file. Compress the file, and copy it to a floppy disk.
- Copy it to the Laptops hard disks.
- Copy the tar files from the first few install disks to the laptops hard disk.
- Use loadlin to boot Linux on the laptop (use the umsdos fs as root).
- Start installation (tell the install program to look for the tar files on your hard disk).
- If the first few files are installed, move back to DOS to erase the tar files which you already unpacked, and copy the next ones.
- reboot Linux.
- repeat until all is installed.
- After installation, it is wise to leave that DOS partition, in case you need to bring more files later.
Is it possible to use two floppy disk controllers under Linux?
YES. To use both controllers, boot the kernel with the lilo parameter 'floppy=two_fdc'. This parameter can be given on the lilo prompt, just after the name of the kernel. It can also be specified in the /etc/lilo.conf file by adding the following line:
append = "floppy=two_fdc"
It is assumed that your second controller lives at 0x370. If it uses another I/O base-address, change this address in the floppy.c. The drives on the second controller use the minor device numbers 128-131 for the generic device, and 132-255 for the fixed geometry device. The following formula gives the minor number:
minor = fdc * 128 + geometry * 4 + unit
See the README file in fdutils for more details.
Is it possible to boot off these special format disks?
Yes. Make sure you have a version of lilo more recent than v18, and put the following line into your /etc/lilo.defines before compiling it:
-DXL_SECS=44
This enables Lilo to boot from disks with up to 44 sectors (i.e. more than you'll ever see :-) ) It allows to boot from ED disks and from disks with more sectors than usual (up to 21 on a HD disk, 42 on an ED disk). However, 2m and Xdf disks cannot be booted using this method.
I get "Unable to allocate DMA memory" messages when trying to use the floppy.
This happens whenever the memory is too scarce to allocate the floppy
driver's DMA buffer. It is possible to use the floppy driver without
DMA, and in that case, the floppy driver uses vmalloc'ed memory, which
is more readily available. In order to disable DMA, use the
floppy=nodma
boot flag. If you use the floppy driver as a
module, set the environment variable 'floppy' to 'dma'.
I have a Micron Millenia Transport.
Apparently, these laptops are supplied with a floppy controller which doesn't support the fifo.
Switch off the FIFO by adding
floppy=nofifo
to your Lilo command line.