On Thu, Nov 22, 2012 at 7:35 PM, Jean-Christophe PLAGNIOL-VILLARD
[off-list ref] wrote:

On 15:31 Thu 22 Nov     , Grant Likely wrote:

quoted

On Thu, 22 Nov 2012 06:41:38 +0100, Jean-Christophe PLAGNIOL-VILLARD [off-list ref] wrote:

quoted

On 18:03 Wed 21 Nov     , Grant Likely wrote:

quoted

On Wed, Nov 21, 2012 at 3:57 PM, Jean-Christophe PLAGNIOL-VILLARD
[off-list ref] wrote:

quoted

From: Sascha Hauer <s.hauer@pengutronix.de>

devicetrees may have a linux,stdout-path or stdout-path property
in the chosen node describing the console device. This adds a helper
function to match a device against this property and retrieve the options
so a driver can call add_preferred_console for a matching device.

NIce. Looks like the right behaviour, but is a little inefficient.
Since the stdout property doesn't change this code could be called
once right after the tree is unflattened and then cache the pointer to
the node anarguments. Can you add the code to of_alias_scan()?

do we assume that we can have only one stdout-path?

I think we could have more serial, framebuffer console

Nope. stdout-path as currently defined is only one device. Someone would
need to propose an extension for multiple stdout's before I'd bother
trying to support that.

ok for this version I do not put it but I need it

I want to have 2 console uart and framebuffer

Personnaly I'd just allow multiple instance of stdout-path

Propose a binding extension.

g.