Re: [uCsimm] ucLinux Compact Flash Adapter

From: Tom Walsh (
Date: Fri Oct 06 2000 - 14:05:19 EDT

Sebastian Andersson wrote:
> On Fri, Oct 06, 2000 at 11:08:19AM -0400, Tom Walsh wrote:
> > Glenn,
> >
> > Seems like we both need the same thing, a working CF drive system. I
> > am curious as to why you are going with the True IDE mode as opposed to
> > the Memory Mapped mode of the CF? Have you looked into the IDE drive
> > code under uClinux, how big a job will it be to rewrite the ide.c stuff?
> Is Memory Mapped Mode supported by real CF harddrives like IBMs
> small 1GB drive?

Well, I cannot answer for IBM, but according to the Compact Flash
Association "CF+ and Compact Flash Specification, Version 1.4", it says
that the CF can operate in one of three modes: I/O Mapped, Memory
Mapped, or True IDE.

> We choose True IDE Mode in a custom hardware device of ours so we
> could reuse the design/code if we wanted to add a real HD via an IDE
> cable.

Oh, I agree totally, but on the performance of the DragonBall processor
is so low that adding a "real" IDE drive to the system may make it drop
significantly if you add in the access delays of the drive mechanicals.
I don't see the addition of a "real" hard drive being an actual benefit,
on an ARM system, yes, the system performance is such that you can take
advantage of large data storage arrays.

One drawback to using the Compact Flash in the True IDE mode is that
mode of operation is NOT supported / implemented in the CF+ cards
(according to the spec it is optional). My main reason for selecting CF
memory is that it is available at any photo store. What bothers me is
the Lexar Media device that I purchased locally says it is "CompactFlash
(TM)" but the back of the device it has the CF+ symbol, I find that
disturbing, it would suggest that we would have to be carefull about
what devices / manufacturers we can use (in True IDE mode). I suggest
that the "lowest common denominator" is the better approch (I/O or
Memory Mapping modes), unless there are compelling reasons???

My idea is that the "sectors" are still accessable in Memory Map mode, I
am thinking of running the device as such and hacking the ide.c to
interface to the device in that mode. Another problem with True IDE
mode is that you have to add additional circuitry to 'power cycle' the
CF into the IDE mode.



Tom Walsh - WN3L - Embedded Systems Consultant
'', ''
"Windows? No thanks, I have work to do..."
This message resent by the list server

This archive was generated by hypermail 2b30 : Sun Apr 07 2002 - 00:01:38 EST