Subject: Re: NAND flash support
To: None <>
From: Toru Nishimura <>
List: tech-embed
Date: 03/13/2007 10:07:53
Bucky Katz wrote;

> I've argued before that NOR and NAND are enough different that they
> really should have separate MI layers, and I still believe that to be
> true.

I think it's one of mistakes Linux MTD has ever made.  NOR and NAND
are distinct mammals not to able to feed the same meal.  In addition to
structural differences of their substance, operational difference poses
issues on computer HW design.  NOR erase is slow.  So slow to complete.
It will takes _tens_ of ticks for 100Hz timeslice.  One tech doc prepared
by Toshiba in 2003 emphasises NOR takes 700msec for block erase.
This means it's un-realistic to wait for erase completion in busy-looping.
I remember that some industrial grade single board computers of 1990's
(MVME compatible) provided erase-done notification via a dedicated
interrupt signal line.  HW engineers did understand well about how
cumbersome for SW to cope with.

My opinion about NOR is that it will loose the popularity as general purpose
storage soon (if not ever).  NOR will scale back to the age when it was used
for small capacity "program store."

Toru Nishimura/ALKYL Technology