Message-ID: <54E53278.70705@broadcom.com>
Date:	Wed, 18 Feb 2015 16:46:48 -0800
From:	Arun Ramamurthy <arun.ramamurthy@...adcom.com>
To:	Arnd Bergmann <arnd@...db.de>,
	<linux-arm-kernel@...ts.infradead.org>
CC:	Mark Rutland <mark.rutland@....com>, <devicetree@...r.kernel.org>,
	"Scott Branden" <sbranden@...adcom.com>,
	Pawel Moll <pawel.moll@....com>,
	"Ian Campbell" <ijc+devicetree@...lion.org.uk>,
	Ray Jui <rjui@...adcom.com>, <linux-kernel@...r.kernel.org>,
	Kishon Vijay Abraham I <kishon@...com>,
	Jonathan Richardson <jonathar@...adcom.com>,
	Rob Herring <robh+dt@...nel.org>,
	<bcm-kernel-feedback-list@...adcom.com>,
	"Dmitry Torokhov" <dtor@...gle.com>,
	Kumar Gala <galak@...eaurora.org>,
	"Anatol Pomazau" <anatol@...gle.com>, <stern@...land.harvard.edu>
Subject: Re: [PATCH 1/2] phy: usbphy: Add dt documentation for Broadcom Cygnus
 USB PHY driver



On 15-02-18 07:15 AM, Arnd Bergmann wrote:
> On Tuesday 17 February 2015 13:05:50 Arun Ramamurthy wrote:
>> On 15-02-17 12:53 PM, Arnd Bergmann wrote:
>>> On Tuesday 17 February 2015 12:00:49 Arun Ramamurthy wrote:
>>>> Arnd, I patched the ehci and ohci driver to accept multiple phys so they
>>>> require different names and cannot both be "usb". That patch was
>>>> accepted by Alen Stern but I did not update the bindings documentation.
>>>> I will send out another patch for that. Could we go with the naming
>>>> scheme of "usb" + "p" + port number or do you have other suggestions?
>>>
>>> I don't have a good idea, but I think it would be best if the first
>>> phy could remain named "usb" for compatibility with the existing binding.
>>>
>> The patch was written in a way that all the existing and new drivers can
>> continue to use "usb" if they are using only one phy so that we remain
>> compatible. The names need to be different only if more than one phy is
>> specified. In such cases i don't think the first phy should be "usb" as
>> it would be confusing to have
>>          phy-names = "usb","usbp1"
>
> I see your patch now, as 7e7a0e67f2c ("usb: ehci-platform: add support for
> multiple phys per controller"), and I'm not too happy about the way you
> did this.

> We already concluded that there should have been a binding change
> to go along with this, and that would have caught the fact that you
> circumvent the API here by reading the phy names manually. That
> part should never have made it into the kernel.
>
> I think we can do this either by defining specific names for the
> phy, or by changing the generic PHY binding to allow anonymous
> phy references (leaving out "phy-names" entirely), and adding a
> proper API for that.
>
Thanks Arnd, I will wait for Alan's comments before proceeding. I am 
happy to patch the ehci-platform driver to use a new api instead of 
devm_phy_get if that is the best option.

>> Should I run this by Alan Stern?
>
> I've added him to Cc here. He clearly didn't know the background about
> the DT binding change, and should not need to, but he may have an opinion
> on what names we should use.
>

>>> What is the reason for having two phys in your case? Are these
>>> identical phy devices connected to a single controller or do they
>>> server different purposes?
>>>
>> Yes, we have three identical phys connected to a single host controller
>> and one of the phys is also connected to the device controller
>
> Ok, no problem with that, let's just make sure we come up with a
> good binding for it.
>
> 	Arnd
>
--
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