From mboxrd@z Thu Jan  1 00:00:00 1970
From: Marek Vasut <marek.vasut@gmail.com>
Date: Thu, 16 Dec 2010 02:59:55 +0100
Subject: [U-Boot] [PATCH v2] BLOCK: Add freescale IMX51 PATA driver
In-Reply-To: <4CF60BED.3060301@denx.de>
References: <1291087578-16714-1-git-send-email-marek.vasut@gmail.com>
	<20101201053418.0C29A134FEF@gemini.denx.de>
	<4CF60BED.3060301@denx.de>
Message-ID: <201012160259.56131.marek.vasut@gmail.com>
List-Id: <u-boot.lists.denx.de>
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
To: u-boot@lists.denx.de

On Wednesday 01 December 2010 09:48:45 Stefano Babic wrote:
> On 12/01/2010 06:34 AM, Wolfgang Denk wrote:
> > Dear Marek Vasut,
> > 
> > In message <201012010346.21897.marek.vasut@gmail.com> you wrote:
> >> On efikamx, I'll soon submit it mainline (when I clean it up enough).
> >> Here's the relevant part of configuration file (the #define __io is
> >> actually the important
> > 
> >> thing):
> > Is there a way to do without this?
> > 
> >> #define __io
> > 
> > I mean, these empty defines look pretty bogus to me.
> 
> I have found several boards with the same behavior:
> 
> include/configs/mv-common.h:#define __io
> include/configs/r2dplus.h:#define __io
> include/configs/edminiv2.h:#define __io
> include/configs/nhk8815.h:#define __io(a)((void __iomem *)(PCI_IO_VADDR
> + (a)))
> include/configs/r7780mp.h:#define __io
> 
> __io is defined if CMD_IDE is set for most of them, meaning CMD_IDE
> requires this define, even if it has no particular meaning for the
> processor we are using.
> It seems to me that defining __io is only required for enabling the
> switch inside arch/arm/include/asm/io.h:
> #ifdef __io
> #define outsb(p,d,l)                    __raw_writesb(__io(p),d,l)
> ...
> 

Indeed

> Really __io should accept a parameter to make some casting, as I can see
> in the blackfin architecture. Except the nhk8815 board, no parameter is
> accepted and the define enables only the #ifdef switch.

Yes
> 
> Probably there is a better way to do this ;-). Could be maybe better to
> define __io inside io.h if it is not already defined by a board ?

This should go through a separate patch though. Did anything like that emerged 
while I was away or shall someone work on this ? Besides, what about this patch?

Thanks, cheers!
> 
> Best regards,
> Stefano Babic