Compact Flash is a valid canidate for an MTD device.

Compact Flash is a valid canidate for an MTD device.

[Eric W Biederman]

The Compact Flash extensions to the IDE command set, there is information
that they can indeed be treated as a normal mtd character device.

In particular the following commands were added to IDE.
-  CFA REQUEST EXTENDED ERROR CODE
-  CFA WRITE SECTORS WITHOUT ERASE
-  CFA ERASE SECTORS
-  CFA WRITE MULTIPLE WITHOUT ERASE
-  CFA TRANSLATE SECTOR
-  SET FEATURES Enable/Disable 8-bit transfer

They aren't a perfectly normal flash chip but, you do have the ability
to just erase or to just write sectors.  CFA TRANSLATE SECTOR gives
information on how many times an individual sector has been erased, so
it is possible to implement wear-leveling above the CFA device.

It appears to me there would be noticeable benefit to configuring a
Compact Flash device as an MTD device, and running JFFS on top of it.

I haven't had a chance to try it yet, but hopefully this will point
someone in a useful direction.

Eric

Re: Compact Flash is a valid canidate for an MTD device.

[Jasmine Strong]

On Friday, July 26, 2002, at 07:23 AM, Eric W Biederman wrote:
> In particular the following commands were added to IDE.
> -  CFA REQUEST EXTENDED ERROR CODE
> -  CFA WRITE SECTORS WITHOUT ERASE
> -  CFA ERASE SECTORS
> -  CFA WRITE MULTIPLE WITHOUT ERASE
> -  CFA TRANSLATE SECTOR
> -  SET FEATURES Enable/Disable 8-bit transfer
>

If you can bypass the device's internal wearlevelling logic, then
maybe;  but it doesn't look awfully like one can.  Can you, for
example, write your own out-of-band data, or write untranslated sectors?

The problem we had was the internal ASIC getting confused and causing
the whole device to cease to function.  It doesn't look awfully like
these extra commands will help this.

Jas.

Re: Compact Flash is a valid canidate for an MTD device.

[Eric W Biederman]

Jasmine Strong <jasmine@regolith.co.uk> writes:

> On Friday, July 26, 2002, at 07:23 AM, Eric W Biederman wrote:
> > In particular the following commands were added to IDE.
> > -  CFA REQUEST EXTENDED ERROR CODE
> > -  CFA WRITE SECTORS WITHOUT ERASE
> > -  CFA ERASE SECTORS
> > -  CFA WRITE MULTIPLE WITHOUT ERASE
> > -  CFA TRANSLATE SECTOR
> > -  SET FEATURES Enable/Disable 8-bit transfer
> >
> 
> If you can bypass the device's internal wearlevelling logic, then
> maybe;  but it doesn't look awfully like one can.  Can you, for
> example, write your own out-of-band data, or write untranslated sectors?

Good question.  The bit about out-of-band data I don't quite get.  I
haven't seen an ordinary flash chip that supports it, so I'm not
certain what the concept is.   ECC data?  You can read/write the ECC
data if needed.

The CFA TRANSLATE SECTOR looks like it gives attempts to give the
mapping from translated to untranslated sectors, as well as the number
of times a sector has been erased.

> The problem we had was the internal ASIC getting confused and causing
> the whole device to cease to function.  It doesn't look awfully like
> these extra commands will help this.

ASIC confusion is certainly hard to guard against.  At the minimum
a window is provided into the workings of the Compact Flash.  Beyond
that the window might just be large enough to configure the device
as an MTD device.

At the same time given that the normal filesystem is FAT16 on a
Compact Flash device I won't expect miracles.  But it looks to me like
it might be worth it to see if the Compact Flash variant of the IDE
command set, is acually useful.  Instead of saying it looks just like
an IDE block device so ignore it.

Eric