Meet the Gang   1   2   3   4   5   6   7   8   9

CRC errors on floppy disks

From "David"

Answered By Dan Wilder, Heather Stern, John Karns

Thankyou for taking the time to read this. I am marked down as one of the university geeks, and it is coming around to the time to hand in essays/dissertations etc. Many people have put misplaced faith in the reliability of floppy disks. Is there any tool for linux, which will allow me to ask the floppy driver in the kernel to keep trying, so I can put together the statistically "best" hex dumps,or extract raw text from Word docs etc.

[Dan] 'Fraid I can't offer any way to recover the data once it's lost to bad sectors, tho no doubt some way is possible.

Instead I'd like to offer the format script I use for floppies before I'll trust 'em with my data. It culls out a lot of floppies, and I find that floppies which once passed, after sitting a few years on the shelf, will often no longer do so.

User who runs this must have write permission to /dev/fd0h1440. Requires mtools. Watch out, some distributions may obsolete fdformat in favor of superformat.

See attached goodfloppy.sh.txt

Today I resuced a 6 page essay 10hrs before it was due,

[Heather] The easiest way to do this sort of thing (I used to do it too) is

1. use dd to extract a raw image of floppy disks. This is sufficiently reliable that it's now the normal way I mount floppies on my laptop.

dd if=/dev/fd0 of=/usr/local/floppy/student-name-ddmmyy.144 bs=18k

In case you're curious 18k is the standard blocksize of a single track, so this is the "native" amount of bits the floppy read heads want to read anyway. It should be fastest and have the best chance of getting all the rescuable bits.

I already take a copy, as I am not the only geek with enough time, and I don't want to write anything back to a damaged floppy. Thanks for the advice with bs=18k I will use that next time.

[Heather] Yep, it's about 8 times faster with that set.

2. the rescue

1. See if Linux will mount the raw image - as it may be more reliable than a floppy whose cookie got slow (in other words a straight read is better than lots of little seeks)
2. if not, use 'strings' on the resulting file, noting that some of the ASCII bits will just be raw strings of letters from headers and things (e.g. "GIF89A")

Ugly but better than nothing during Finals Week.

Most disks that I have come across mount OK, but if all else fails, I have all night, a copy of Peter Nortons PC Programmers Bible (plenty on disk structure, FAT layout etc), and the promise of as much beer as I can drink from the hapless author.

[Heather] I got a full case of Henry Weinhardt's Root Beer once. The gal knows I don't normally drink alcohol. No wait, that wasn't in college, that was in the corporate world. I got some cool coffeecups... uh no, that was in the corporate world too. Nevermind

...but the 200 words held on the bad sectors were well and truly gone.

[Heather] Sorry. Once there's scratch marks there, they lose.

This is not so good, what I was really after is someway to attempt to make best guesses for what is in those bad-sectors. I know that it sounds difficult, but I am looking for something between Scandisk for windows, and professionals in a clean-room!

[Heather] We have dosfsck, of completely unknown reliability, but you're welcome to try it too. What the heck.

Short of trying the floppy again in a floppy bay with different alignment, basically you either got the bits, or you didn't. And indeed, you could try that (though it submits the poor dented floppy to N more visits by drive gear) and see if the md5sums of any of the dd images are different. If so, scour the tweaked ones for anymore bits you can rescue also. Don't hold your breath, but you'll have earned the extra brewskies if it works.

Clean-room efforts are basically able to micro-position the heads in the hopes of seeing the last two to five major writes. Ontrack does such recovery, but I dunno if they do floppies at all. I suspect not.

Is there anything better than dd conv=noerror, that will allow me to access these dodgey disks from linux. I am kind of looking for something like Spinrite by Steve Gibson, but I would like to have it for linux, and for free.

[Heather] spinrite worked on IDE drives, not floppies.

If you want a hex editor in the spirit of Norton Disk Editor, try lde, but frankly, I'd use 'hexedit' (a curses-terminal program) on a dd image of the floppy before I'd consider that ... to reduce the amount of usage tried on an already poor-condition floppy.

Oh yeah, and much as ZIP and LS-120 bays are less popular, their cartridges are much more reliable than standard floppies, and CD-RW even more so (but sadly CD-RW cannot be written to directly, just burned as a complete image -- and DVD-RAM drives are still not in the normal-consumer price range, plus they're rare even in the Silicon Valley).

Forget DVD-RAM / CD-RW. Most people use floppies to move work between computer labs with internet, and our rooms without. Nobody in college has administritave access to the networked PC's, and the amount of people who simply use a floppy to avoid copying files around is scary. We have just got new ZIP-250's, so things should look up for next term!

[Heather] Well, definitely looking up there!

Oh yeah, remind people not to leave their floppies (or cartridges) lying around in the noonday sun on the car shelf, or other places where they would get toasty and you wouldn't put VHS tapes. It's basically the same type of media with the same type of susceptibility to heat damage.

[John] I usually keep floppies in the box they came in or in a small plastic bag. If you examine one, you'll see that they have an open slot at the top edge where the sliding metal access door is. When left out in the open air, floating particles enter and adhere to the surface of the plastic disk inside.

Zip disks don't seem to have such a comparatively large opening in the housing as floppies, which may a reason they fare better. But one is very well advised to keep all magnetic media, including tapes, in a protective case or other dust barrier. I've noticed that plastic materials seem to act like magnets for carbon dust, which when smudged leaves a grease spot on the plastic material.

[Heather] I am reminded of a time that we posted somebody's ruined floppy to the student-center wall... pried loose of its plastic envelope so the round disc was revealed, dent and all. Caption: "Don't do this to your floppy". I forget exactly what did it. The crease wasn't much - maybe someone sat on it, I dunno. But it was definitely dead.

I don't know if your policies are about guest-level access only or you create actual accounts for people there. If the latter then some sort of share server (samba or perhaps one of those file-appliance boxes that have gotten so popular) which looks like the same drive letter on all networked boxes, would get people to use that instead. I like H: for "home" but your tastes may vary. Combined with the permissions-by-login behavior samba can provide it's not too bad. You might not want to make it automatic or perhaps people will abandon logged in machines and mess up each other's space. And not everyone would use it, perhaps they want to take some work back to the dorm with them. They might even get them out of sync. (I think the MSwin buzzword for doing it the right way is "My Briefcase" but don't quote me, I could be wrong these days. Here in linux we use rsync, and have to be careful about the commands we give it, too.) In other words it has its own headaches... but it's not as unreliable at the bits level as floppies are.

I think that's about as much as Linux can really do for throwing software against what is basically a hardware and a clue-ware problem, but there ya go. Best of luck in the efforts!

Thanks in advance

David

P.S. If you publish my e-mail can you obscure it somehow. My uni acct is so far spam-free and I would like to keep it that way.

[Heather] Yes, we can anonymize you. If this stays short enough for Tips we'll put "anonymous" instead of your email, and if enough of the Gang chime in it may be a TAG thread, where emails are not published at all unless directly related to the solution (e.g. project maintainers).

This page edited and maintained by the Editors of Linux Gazette Copyright © 2002
Published in issue 77 of Linux Gazette April 2002

HTML script maintained by Heather Stern of Starshine Technical Services, http://www.starshine.org/

Meet the Gang   1   2   3   4   5   6   7   8   9