OK, anybody have any hints and tips to get an MFM drive working again?

OK, anybody have any hints and tips to get an MFM drive working again?

[Rogier Wolff]

Hi,

We need to recover some data off an old MFM drive. We've got a bunch
of those cards, we've got a bunch of drives to experment with, but 
once we get things working we'll recover some 2x20Mb of data off two
20Mb drives....

We've tried the modern IDE driver, and the 2.4.20 one, and the "old
hd only" driver. 

Going "retro": Compiling 2.0.39 gives me: "bus error" while doing 
make dep. 

I THINK we have a couple of those cards that don't have any 
interrupts. Would Linux be able to work with those?

Yes, we pass "hda=615,4,17" to the IDE driver. 

Any suggestions of things to try?

		Roger. 

-- 
** R.E.Wolff@BitWizard.nl ** http://www.BitWizard.nl/ ** +31-15-2600998 **
*-- BitWizard writes Linux device drivers for any device you may have! --*
**** "Linux is like a wigwam -  no windows, no gates, apache inside!" ****

Re: OK, anybody have any hints and tips to get an MFM drive working again?

[Måns Rullgård]

Rogier Wolff <Rogier@wolff.net> writes:

> Hi,
>
> We need to recover some data off an old MFM drive. We've got a bunch

[...]

> Any suggestions of things to try?
>
> 		Roger. 
>
> -- 
> ** R.E.Wolff@BitWizard.nl ** http://www.BitWizard.nl/ ** +31-15-2600998 **
> *-- BitWizard writes Linux device drivers for any device you may have! --*

Write a driver ;-)

-- 
Måns Rullgård
mru@kth.se

Re: OK, anybody have any hints and tips to get an MFM drive working again?

[alan]

On Thu, 29 Jul 2004, Rogier Wolff wrote:

> We need to recover some data off an old MFM drive. We've got a bunch
> of those cards, we've got a bunch of drives to experment with, but 
> once we get things working we'll recover some 2x20Mb of data off two
> 20Mb drives....
> 
> We've tried the modern IDE driver, and the 2.4.20 one, and the "old
> hd only" driver. 
> 
> Going "retro": Compiling 2.0.39 gives me: "bus error" while doing 
> make dep. 
> 
> I THINK we have a couple of those cards that don't have any 
> interrupts. Would Linux be able to work with those?
> 
> Yes, we pass "hda=615,4,17" to the IDE driver. 
> 
> Any suggestions of things to try?

First of all, what format is the data on the drives? ext2 or fat?

What mfm card are you using? I assume it is an isa card.  (I have never 
heard of a pci mfm card.)

If this is a one time thing, you might try tracking down an older version 
of Linux. (One with XT drive support.)

If you can't find one, contact me off-line.  I have versions going back to 
the Yggdrasil beta.  An old version of Slackware should work.

Re: OK, anybody have any hints and tips to get an MFM drive working again?

[Jeffrey E. Hundstad]

Use a 2.2 kernel.

Rogier Wolff wrote:

>Hi,
>
>We need to recover some data off an old MFM drive. We've got a bunch
>of those cards, we've got a bunch of drives to experment with, but 
>once we get things working we'll recover some 2x20Mb of data off two
>20Mb drives....
>
>We've tried the modern IDE driver, and the 2.4.20 one, and the "old
>hd only" driver. 
>
>Going "retro": Compiling 2.0.39 gives me: "bus error" while doing 
>make dep. 
>
>I THINK we have a couple of those cards that don't have any 
>interrupts. Would Linux be able to work with those?
>
>Yes, we pass "hda=615,4,17" to the IDE driver. 
>
>Any suggestions of things to try?
>
>		Roger. 
>
>  
>

Re: OK, anybody have any hints and tips to get an MFM drive working again?

[Alan Cox]

On Iau, 2004-07-29 at 21:14, Rogier Wolff wrote:
> I THINK we have a couple of those cards that don't have any 
> interrupts. Would Linux be able to work with those?

The old hd driver should support this in 2.4 and 2.6. Its fair to
say that neither hd.c nor xd.c get much testing nowdays. One of the
biggest problems tends to be finding a machine you can use the old MFM
cards in because any motherboard disk controller will clash.

If you aren't in any hurry you can call handler() from the start of the
timer timeout function if you lack IRQ's.

Re: OK, anybody have any hints and tips to get an MFM drive working again?

[alan]

On Thu, 29 Jul 2004, Alan Cox wrote:

> On Iau, 2004-07-29 at 21:14, Rogier Wolff wrote:
> > I THINK we have a couple of those cards that don't have any 
> > interrupts. Would Linux be able to work with those?
> 
> The old hd driver should support this in 2.4 and 2.6. Its fair to
> say that neither hd.c nor xd.c get much testing nowdays. One of the
> biggest problems tends to be finding a machine you can use the old MFM
> cards in because any motherboard disk controller will clash.

Not to mention that many MFM boards were not very static resistant and 
tended to die if handled without proper grounding.  (I lost one that way.)

Re: OK, anybody have any hints and tips to get an MFM drive working again?

[Rogier Wolff]

On Thu, Jul 29, 2004 at 08:44:29PM -0400, Albert Cahalan wrote:
> > We've tried the modern IDE driver, and the 2.4.20 one,
> > and the "old hd only" driver. 
> >
> > Going "retro": Compiling 2.0.39 gives me: "bus error"
> > while doing make dep. 
> 
> You need the old "xd" driver. It isn't IDE at all.
> You can be pretty sure you need the "xd" driver when:
> 
> a. the drive is 5, 10, or 20 MB

20. Yes, we're sure we have an MFM drive. 
(The KL330 is a 30Mb RLL drive. It requires a different
controller, but would require exactly the same driver
treatment from Linux)

> b. the drive has 3 cables: data, power, and control

Yup. 

> c. the drive is called MFM, RLL, or some other strange thing

Yup. 

> The mkdep bus error is easy to fix. The program is
> relying on an old well-loved Linux behavior that
> turned out to violate the POSIX standard. Linux used
> to give you a free zero page when doing a mmap()
> past the end of a file.
> 
> You might grab the mkdep from a later kernel.

OK. Will do. 

> Anyway, the 2.0.xx kernel is about right. Linux
> didn't get a working "xd" driver until long after
> the hardware was obsolete. The driver wasn't kept
> around for very long, because all the hardware
> was gone... except for your stash, if we assume
> that those drives really work.

Last thing we tried is booting it. It boots! i.e. we've now
verified that the hardware actually works..... But this card
has a jumper where 1-2 means IRQ5, 2-3 means IRQ 7, and
there is no jumper installed. Guess what that means for
the interrupt line it uses.... :-)

> Um, you'll most likely need an i386 to even boot
> with the card installed. Maybe you can boot on a
> somewhat recent machine if you disable IDE from
> the BIOS. Got SCSI? How about a network boot?

We've got two machines up-and-running which seem to be up to the 
task. One is an i386 (WITH coprocessor!!!), only ISA slots, no 
"integrated peripherals". It goes "NFS root" as soon as it's loaded
the kernel off the boot floppy. (We couldn't find any 4Mb SIMMs, 
so it only has 8M ram...)

The other is an Intel Celeron 466, which has one last ISA slot. 
When I disable "integrated IDE controller" (and the floppy for cards
that have a floppy), then I get the same results as on the 386. 
(It loads a boot-prom off the floppy and goes "diskless" one
step earlier: before loading the kernel)

Ah! Shit! During this whole excercise I forgot to lookup the 
source-line-that-generated-the-error! That is usually helpful. 

Look! This is the line that triggers: 

  if ((blockend < block) || (blockend > drive->part[minor&PARTN_MASK].nr_sects)) {

I'll take a guess that if I initialize an MFM drive, nr_sects get
initlialized to "0". Any block I ask for is beyond end-of-disk. 

I'll put a drive->part[0].nr_sects = 63960 in there. 
(That's the number of sectors that the drives would have if they
are RLL formatted and not MFM). Other things will go wrong if I
access a sector beyond the end-of-the-disk. "just don't do that
then"!

And, I've made the beginners mistake of not including the error
messages I got. I apologise.... 

	Roger. 

-- 
** R.E.Wolff@BitWizard.nl ** http://www.BitWizard.nl/ ** +31-15-2600998 **
*-- BitWizard writes Linux device drivers for any device you may have! --*
**** "Linux is like a wigwam -  no windows, no gates, apache inside!" ****

Re: OK, anybody have any hints and tips to get an MFM drive working again?

[Jeff Woods]

At 7/29/2004 10:14 PM +0200, Rogier Wolff wrote:
>We need to recover some data off an old MFM drive. We've got a bunch
>of those cards, we've got a bunch of drives to experment with, but
>once we get things working we'll recover some 2x20Mb of data off two
>20Mb drives....

>Any suggestions of things to try?

What OS was used to write to the MFM drives?  Use it.

--
Jeff Woods <kazrak+kernel@cesmail.net>