OSDev.org

hi

how can i create an virtual floopy disk in linux and use it like as a normal disk (like vfd in windows)?

Create a image file with requested size using dd:

dd if=/dev/zero of=/tmp/floppy.img bs=1k count=1440

Mount it using loop device;

losetup /dev/loop5 /tmp/floppy.img

Setup partition info using fdisk;

fdisk /dev/loop5

Format drive;

mkfs -t <fstype> /dev/loop5

Mount it to local filesystem;

mount /dev/loop5 /tmp/myfloppymount

If you are using existing image you can skip fdisk and mkfs...

Good luck!

-
Thomas

hi

fdisk /dev/loop5 , dosent work and return this message:

Device contains neither a valid DOS partition table, nor Sun, SGI or OSF disklabel
Building a new DOS disklabel with disk identifier 0x8251a9fd.
Changes will remain in memory only, until you decide to write them.
After that, of course, the previous content won't be recoverable.

You must set cylinders.
You can do this from the extra functions menu.
Warning: invalid flag 0x0000 of partition table 4 will be corrected by w(rite)

That's no miracle, because a floppy disk of course doesn't contain a partition table.

Hi,

The wiki already has information about this (not even including the pages about GRUB):
http://wiki.osdev.org/Disk_Images
http://wiki.osdev.org/Loopback_Device

To get a virtual floppy drive formatted as FAT12 you can simply do something like: Unmounting it can be done using:
Regards,
Stephan J.R. van Schaik.

Be mindful that such mounting of floppy images does _not_ emulate a floppy drive. VFD in Windows _does_ emulate a floppy drive.

I know this thread is a little old, but I've been doing some "Internet clean-up" recently concerning this very subject. Been meaning to comment here on this for a while.

It's important to realise that a plain vanilla Linux kernel does _not_ have the ability to emulate hardware. A plain vanilla Linux kernel _does_ have the ability to take an image of a filesystem and mount it.

You can image any filesystem with dd and mount it in Linux. All this is doing is having the files in that filesystem image displayed in the directory where you mount the image to. It does not emulate hardware. So you cannot "create an virtual floopy disk in linux and use it like as a normal disk (like vfd in windows)". There would be nothing "floppy" about it. It would just be a filesystem mounted at a directory. It is not a "virtual floppy disk".

Pull up a Linux terminal and do the following: "sudo lshw | more". Or even better, "sudo lshw > $HOME/Documents/my_hardware.txt", then view my_hardware.txt.

Run lshw before and after mounting a floppy image with the mount command. The mount command will have absolutely no effect on the hardware list. You are not emulating a floppy drive, or else it would show up in the hardware list.

Now get an iso image of a CD or DVD. Check lshw before and after mounting it with the mount command. Again, no emulation of hardware.

Now install CDEmu and mount the iso with CDEmu. Now check lshw. Ah. Now it's in the hardware list. Now you're emulating hardware.

You can do something parallel to this in Windows. Instead of lshw, you can open device manager. Try mounting an iso with PowerIso (viewing device manager before and after) and also try mounting an iso with anything else (viewing device manager before and after). PowerIso does not truly emulate an optical drive (doesn't produce anything in device manager). This is the only iso mounting software for Windows that I know of that doesn't produce an optical drive in device manager. BTW, here are more reasons why PowerIso is really weird: mounting an iso with PowerIso makes iTunes think a blank disc has been inserted, and mounting a floppy image with PowerIso shows up in the File Browser with an optical disc icon, if I remember right.

VFD in Windows _does_ add a floppy drive in device manager.

I think the aforementioned wikis concerning loopback devices and image mounting need a little tweaking, partially because this lack of emulation needs to be brought to light.

The loopback device entry starts out like this:

A Loopback Device is a mechanism used to interpret files as real devices. The main advantage of this method is that all tools used on real disks can be used with a loopback device.

"Real devices" is a little misleading here. If it just said "device" and not "real device", then I could see the wiki meaning "device" in the *nix sense of the term (like loopback "device", block "device", character "device", etc. You know, abstraction files.). But "real device" definitely would make the average person think "hardware". Again, a loopback device does _not_ emulate hardware.

The main advantage of this method is that all tools used on real disks can be used with a loopback device.

Not true. A loopback device is only a mounted filesystem. Thus one loopback device cannot include more than one partition. Anything outside of the one mounted filesystem is not in the loopback device. This means no partitioning scheme, no MBR, no GPT, no no-man's land between the MBR and the first parition, none of that. Things like fdisk, gdisk, and gParted aren't going to help you here. If you want to make an MBR or GPT schemed raw image with GRUB2 installed to the MBR+, you might want to use something like qemu-nbd instead of a loopback device. But I have found qemu-nbd to be unstable, and a hypervisor seems to be the next choice since I don't know of any hard drive emulators for Linux, or even anything other than qemu-nbd that creates an abstraction file for a "whole hard drive image", without the actual emulation of hardware in Linux. And of course, low level formatting tools and any tools that have to do with hard drive firmware or the settings thereof is out of the question with a loopback device.

As you OS Devs know, only the kernel has the privileges necessary to see hardware, so all non-kernel programs only know anything about hardware from the information the kernel feeds them. Thus hardware can be emulated simply by the kernel lying and saying there is this (non-existent) hardware. That is what CDEmu does via a Linux kernel module.

Of course there is more to emulating hardware than just showing up in the hardware list, and CDEmu does quite a few things, such as responding to certain protocols, such as the SCSI command set and whatever the Raw DAO 96 protocol is.
I imagine that to emulate a floppy drive, there would be at least a few protocols that need to be supported (which certainly aren't supported by a loopback device).

I have searched and not found a single floppy drive emulator for Linux. However, mounting a floppy image with the mount command is often sufficient for what people are trying to do.

Cheers,
Jake

SpawnHappyJake wrote:A loopback device is only a mounted filesystem.

Sorry, but this is totally wrong. A loopback device is not a file system, but a block device; more precisely, a file mapped to a block device.

A block device (no matter if it's called loop0 or sda) can contain a file system. It can also contain a partition table. It can contain anything. Using fdisk on a loop device is meaningful and if you want to create a partitioned image, it is what you should do.

Now if you have a partitioned image and you want to actually mount it or create a file system on the image, what you need is block devices for the individual partitions. You can either do that manually by using the --offset option of losetup, or you can use kpartx. I recommend the latter.

qemu-nbd does have its uses, but they are really related to non-raw file formats, say if you want to mount a qcow2 or a VMDK image.

Sorry, but this is totally wrong. A loopback device is not a file system, but a block device; more precisely, a file mapped to a block device.

Ok, I see. Thanks. Sorry about that. My world has been embettered. After looking at that wiki again, seeing the losetup command opened new horizons for me. I shall play around with that.

Cheers,
Jake

Thanks again Kevin! Boy I'm a happy camper! I made a 1 GiB image with dd, and used losetup to connect it to /dev/loop0.

Then I used gdisk (like fdisk, but does gpt. It's really cool. I like that it rounds to 2048-sector boundaries for you. (one 2048-sector boundary = 1 MiB)) to partition it. I could have sicked gdisk directly to the image, but for fun I aimed it at /dev/loop0 (since that is what I was playing with), which works just as well. I made 4 partitions: a storage parition, an EFI System parition, a BIOS boot partition, and a second storage partition.

After scheming the image as GPT and making partitions, the kernel did not auto-create additional abstraction files for each partition, not even after detaching and reattaching the image to /dev/loop0.

But I haven't used all my free loopback abstraction files according to "mount" and "ls /dev". So I hooked the first storage partition to /dev/loop1 and the second storage partition to /dev/loop2. I was careful to use the -o (offset) and --sizelimit options just right to only attach the partitions, nothing more, nothing less, to the abstraction files.

I was then able to use mkfs.ext3 on them to format the partitions to ext3.

Then I could mount them like this (after making /media/play1 and /media/play2 with "sudo mkdir"):
sudo mount -o users /dev/loop1 /media/play1
sudo mount -o users /dev/loop2 /media/play2

The "-o users" allowing all users to mount and unmount them. But I couldn't read or write to them without root privileges. Thanks to Mr. Willimason over here: http://superuser.com/questions/174776/m ... xt4-volume, that was a short-lived problem:

sudo chmod 777 /media/play1
sudo chmod 777 /media/play2

Yeah, works great! I'm don't need qemu-ndb for raw images anymore! That's a relief. I think next I'll try installing grub2 to it and seeing if I can boot the ArchBang installer in QEMU. There's a bug in the version of QEMU I have that makes it get hung-up on the ISOLINUX bootloader, so I can't just boot the iso. I tried VirtualBox, but the VM crashes every time I run rankmirrors. Running in QEMU will tell me if that's a VirtualBox bug or a bug somewhere in ArchBang. Tried compiling the latest released tarball of QEMU as well as QEMU out of the git, but got make errors, even with --disable-werror. Oh well. This losetup stuff is a big help.

Thanks again,
Jake