Re: [PATCH] MTD: fix dataflash 64-bit divisions

[Artem Bityutskiy <dedekind@infradead.org>]

On Wed, 2008-12-17 at 09:56 -0800, David Brownell wrote:
> On Wednesday 17 December 2008, Artem Bityutskiy wrote:
> > From: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>
> > Date: Wed, 17 Dec 2008 19:42:38 +0200
> > Subject: [PATCH] MTD: fix dataflash 64-bit divisions
> > 
> > MTD has recently been upgraded for 64-bit support, see commit
> > number 69423d99fc182a81f3c5db3eb5c140acc6fc64be in the
> > mtd-2.6.git tree (git://git.infradead.org/mtd-2.6.git)
> 
> Hmm, in another thread I was just reading about how Linux
> would only support the first 2 GBytes of a 4 GByte NAND
> chip (Samsung MT29F16G08DAA) ... the updates to support
> large pages were easy, but signed 32-bit offsets prevented
> the full size from being recognized.  Slightly older parts
> integrated four dies with 2K pages, not two with 4KB ones,
> and gave no trouble.

Yeah, we still do not support 4KiB pages, and >2GiB NANDs.

> Would this be part of a set of patches making 4 GByte
> (and eventually, larger) NAND chips behave?

Well, this patch does only part of the job - it changes
in-kernel API. Yes, makes >4GiB NANDs behave, and we tested
it with NAND simulator (nandsim).

However, all the user-space interfaces are still 32-bit.
And the interfaces are not extendible, so someone should
invent completely new MTD interfaces.

And to support 4KiB-page NANDs, which have 128bytes OOB,
one needs to change user-space interfaces (ioctls), because
they support 64-bit OOBs at max. On the other hand, I
personally do not care about OOB support, because it is
in general better to avoid any use of OOB.

-- 
Best regards,
Artem Bityutskiy (Битюцкий Артём)