I don’t know how that got on there

Someone handed me a couple of floppy disks the other day and wanted me to get some important files from them, these disks had been sitting in the back of a cabinet since 1996. I had three USB floppy drives and a couple of hardware floppy drives handy so I thought I’d give it a go and tried both disks in each drive. All I got was erroring on almost every sector with messages like:

sd 5:0:0:0: [sdc] CDB: Read(10): 28 00 00 00 01 1e 00 00 01 00 00 00 end_request: critical target error, dev sdc, sector 286
sd 5:0:0:0: [sdc] Unhandled sense code
sd 5:0:0:0: [sdc]  Result: hostbyte=invalid driverbyte=DRIVER_SENSE
sd 5:0:0:0: [sdc]  Sense Key : Medium Error [current] Info fld=0x11f
sd 5:0:0:0: [sdc]  Add. Sense: Recorded entity not found

It’s most likely that these errors were caused by age but floppy disk drives also have differing alignment between the drive the data was written with and the drive used for reading the data.

There was only one thing I could do really and that was give ddrescue a try. There are two utilities called ddrescue, to be exact it was GNU ddrescue version 1.11. What I decided to do was try running ddrescue using three different read methods, on each disk in every drive (different alignments). My hope being that by writing to the same image file and using the logfile feature of ddrescue that I could amalgamate the good parts of each read into one complete image file.

Note: You’ll probably need to change the options below for each floppy device, file location, etc, if the drives are in different computers it’s probably easiest to save the files to some central location.

First we try to copy as much data as possible, without splitting sectors or retrying on failures:

sudo ddrescue --verbose --no-split --cluster-size=1 /dev/sdc floppy1.img floppy1.log

Now we retry any previous errors twice, using uncached reads:

sudo ddrescue --verbose --direct --max-retries=2 --cluster-size=1 /dev/sdc floppy1.img floppy1.log

Finally we try again but with the retrim option set, so that ddrescue will try to reread full sectors:

sudo ddrescue --verbose --direct --retrim --max-retries=2 --cluster-size=1 /dev/sdc floppy1.img floppy1.log

At this point you hopefully have a complete floppy disk image (I only had to use three of the available floppy drives) and then depending on the situation and damage you can either run the image through fsck, mount it, write it to a new disk or run it through a file carver.

In my situation I was able to use the mtools suite to copy the required files directly from the floppy disk image files.