From mboxrd@z Thu Jan  1 00:00:00 1970
From: Ben Warren <biggerbadderben@gmail.com>
Date: Thu, 23 Jul 2009 14:25:34 -0700
Subject: [U-Boot] [PATCH] Using I/O accessor instead of volatile
 pointers in SMC911x driver
In-Reply-To: <20090723195422.8E91F832E416@gemini.denx.de>
References: <1248250473-12694-1-git-send-email-matthias.weisser@graf-syteco.de>
	<20090723195422.8E91F832E416@gemini.denx.de>
Message-ID: <4A68D54E.3060902@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

Wolfgang Denk wrote:
> Dear Matthias Weisser,
>
> In message <1248250473-12694-1-git-send-email-matthias.weisser@graf-syteco.de> you wrote:
>   
>> Volatile pointer usage caused lockup with arm-gcc 4.3.2
>> Using I/O accessor fixed that.
>>     
>
> Hm...
>
>   
>> -	return *(volatile u32*)addr;
>> +	return readl(addr);
>>     
>
> On big-endian systems like PowerPC, readl() is a byte-swapping
> (little-endian) input function, which means that your patch changes
> the byte order of all I/O operations.
>
>   
Yeah, that's why I left these alone when re-working this driver.  I 
don't know what the correct approach is when using memory-mapped devices 
that could in theory go on the local bus of any processor.  Most people 
I know would hook them up according to the native endianness.

regards,
Ben