Date:	Thu, 15 Nov 2012 17:15:17 +0100
From:	Per Förlin <per.forlin@...ricsson.com>
To:	Grant Likely <grant.likely@...retlab.ca>
Cc:	"rob.herring@...xeda.com" <rob.herring@...xeda.com>,
	Linus Walleij <linus.walleij@...aro.org>,
	"lee.jones@...aro.org" <lee.jones@...aro.org>,
	Fredrik SODERSTEDT <fredrik.soderstedt@...ricsson.com>,
	"devicetree-discuss@...ts.ozlabs.org" 
	<devicetree-discuss@...ts.ozlabs.org>,
	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] dt: platform: Extract device name from device tree blob

On 11/15/2012 04:52 PM, Grant Likely wrote:
> On Mon, 12 Nov 2012 23:54:40 +0100, Per Förlin <per.forlin@...ricsson.com> wrote:
>> On 11/12/2012 04:20 PM, Grant Likely wrote:
>>> On Mon, Nov 12, 2012 at 2:59 PM, Per Forlin <per.forlin@...ricsson.com> wrote:
>>>> Add support to extract device name from device tree blob.
>>>> If the property "dev-name" is set in the DTS this name will
>>>> be used when creating the device.
>>>> The auxdata_lookup has precedence and will override
>>>> the "dev-name" property.
>>>
>>> Using a 'dev-name' property has the same problem that the 'cell-index'
>>> properties have in that it is encoding part of the global namespace
>>> local to the node and it becomes easy to create collisions. Instead of
>>> this check to see if one of the properties in /aliases points to the
>>> node and use that for the name.
>>>
>>> g.
>>>
>> Thanks Grant for your feedback,
>>
>> Extract from exynos5250.dtsi:
>> -----------
>> 	aliases {
>> 		spi0 = &spi_0;
>> 		spi1 = &spi_1;
>> 		spi2 = &spi_2;
>> 	};
>>
>> 	spi_0: spi@...20000 {
>> ...
>> 	};
>>
>> 	spi_1: spi@...30000 {
>> ...
>> 	};
>>
>> 	spi_2: spi@...40000 {
>> ...
>> 	};
>> ---------------
>>
>> Alias refers to the device node. The device node is not aware of the alias.
>>
>> How to get a device name from the aliases.
>> 1. Traverse all aliases for each device node (time consuming if there are many aliases)
>> 2. Make a new function of_alias_get_name(), today there is only of_alias_get_id()
>> 3. The functionality of setting device_name based on alias name needs to be optional because one may want to use aliases without changing the name of the device.
>> All this is feasible but perhaps not optimal.
>>
>> I don't really see how come name space is a big issue in this case. The name space of "dev-name" is local to the device node. A child device node can use the same dev-name as the parent (unless I'm mistaken which happens quite often). Introducing yet another property name pollutes the name space of the device node. Still I think the pros are stronger than the cons.
>>
>> Do you still prefer to use the name of the Alias? Could you please elaborate a bit more how this can be done in practice?
>> I would agree with you if there was a reference from the device node to the alias.
> 
> Oh, I see what you're trying to do. As Lee pointed out you're trying to
> make the Linux internal way of matching up clocks and regulators happy.
> That is very much a Linux-kernel internal thing and should be solved in
> the kernel. Trying to solve it with fixed names in the device tree will
> cause problems down the road.
> 
> I though you were wanting to have logical names for the devices that
> make sense to the user which is how aliases is used now.
> 
> So, no, don't do this.
> 
> g.
> 
Hi,

I came to the same conclusion when I dag into it some more. I replied to my own comment and concluded (https://lkml.org/lkml/2012/11/13/309).
The solution is to move clocks and regulators into the DTS. When this is done there will be no need for setting a specific device-name (all those auxdata_lookup can be removed)

BR
Per

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists