From mboxrd@z Thu Jan  1 00:00:00 1970
From: Deepak Saxena <deepak_saxena@mentor.com>
Date: Fri, 10 Dec 2010 10:04:34 -0800
Subject: [U-Boot] [PATCH] Honor /memory/reg node in DTB files
In-Reply-To: <20101208223405.6AC07CF5CA6@gemini.denx.de>
References: <4CFD863A.7070000@mentor.com>	<20101207065234.DC3BB280@gemini.denx.de>	<4CFE775C.6050001@mentor.com>	<20101207190936.EC285D0E429@gemini.denx.de>	<4CFFCEC1.6000103@mentor.com>	<20101208205359.6A1DDCF5CA6@gemini.denx.de>	<4CFFF3C4.20800@mentor.com>	<20101208213811.E218CCF5CA6@gemini.denx.de>	<0DDCBDA1-188F-433D-BDCC-5FDCF709A131@digitaldans.com>
	<20101208223405.6AC07CF5CA6@gemini.denx.de>
Message-ID: <4D026BB2.6020609@mentor.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 12/08/2010 02:34 PM, Wolfgang Denk wrote:

> 
> I guess we can argue that the normal situation is that U-Boot will
> know how to update the DT such as needed to boot the OS. So what we
> are dealing with is a small percentage of cases where we need special
> behaviour, and where it may be acceptable if the solution is only
> semi-perfect ;-)
> 
> My current thinking is to introduce something like
> 
> 	dt_skip=memory,mac-address
> 
> including eventually "dt_skip=ALL".  This should cover most of the
> current use cases.
> 
> If someone gets fancy he can add wildcard support.
> 
> And if we need even more flexibility, we can add some "dt_include"
> with higher priority, so one could do for example
> 
> 	dt_skip=ALL
> 	dt_include=memory

I imagine this being rather ugly to implement and to keep the code clean
and maintained. Who parses these variables? Does each and every piece of
code in U-Boot that now touches a piece of the DT need to check for this
variable? I could see something like this working
if there was a central DT handler that read nodes and then called
platform-specific over-ride function, i.e.:

	for_each_node_in_dt() {
		if (dt_include(node->type))
			platform_of_dt_node_process(node, boot_stage);
	}

	where boot_stage tells us whether we are at early init, about to
        boot an OS image, or in some other step in the process.

This would provide a consistent method of handling that variable.
Without something like this, I think and environment variable is just
going to create confusion for users and developers.

~Deepak