Re: [PATCH 2/2] arm: omap: remove *.auto* from device names given in usb_bind_phy

[Kishon Vijay Abraham I <kishon@ti.com>]

Hi,

On Tuesday 30 July 2013 11:58 AM, Felipe Balbi wrote:
> Hi,
> 
> On Tue, Jul 30, 2013 at 11:55:04AM +0530, Kishon Vijay Abraham I wrote:
>>> On Tue, Jul 30, 2013 at 11:41:23AM +0530, Kishon Vijay Abraham I wrote:
>>>>>>>>>> diff --git a/arch/arm/mach-omap2/board-2430sdp.c b/arch/arm/mach-omap2/board-2430sdp.c
>>>>>>>>>> index 244d8a5..17bb076 100644
>>>>>>>>>> --- a/arch/arm/mach-omap2/board-2430sdp.c
>>>>>>>>>> +++ b/arch/arm/mach-omap2/board-2430sdp.c
>>>>>>>>>> @@ -233,7 +233,7 @@ static void __init omap_2430sdp_init(void)
>>>>>>>>>>  	omap_hsmmc_init(mmc);
>>>>>>>>>>  
>>>>>>>>>>  	omap_mux_init_signal("usb0hs_stp", OMAP_PULL_ENA | OMAP_PULL_UP);
>>>>>>>>>> -	usb_bind_phy("musb-hdrc.0.auto", 0, "twl4030_usb");
>>>>>>>>>> +	usb_bind_phy("musb-hdrc.0", 0, "twl4030_usb");
>>>>>>>>>
>>>>>>>>> how about moving usb_bind_phy() calls to omap2430.c ?
>>>>>>>>>
>>>>>>>>> diff --git a/drivers/usb/musb/omap2430.c b/drivers/usb/musb/omap2430.c
>>>>>>>>> index f44e8b5..b6abc1a 100644
>>>>>>>>> --- a/drivers/usb/musb/omap2430.c
>>>>>>>>> +++ b/drivers/usb/musb/omap2430.c
>>>>>>>>> @@ -544,6 +544,9 @@ static int omap2430_probe(struct platform_device *pdev)
>>>>>>>>>  
>>>>>>>>>  		pdata->board_data	= data;
>>>>>>>>>  		pdata->config		= config;
>>>>>>>>> +	} else {
>>>>>>>>> +		/* bind the PHY */
>>>>>>>>> +		usb_bind_phy(dev_name(&musb->dev), 0, "twl4030_usb");
>>>>>>>>
>>>>>>>> This looks like a hack IMHO to workaround the usb phy library. otherwise it is
>>>>>>>> similar to get_phy_by_name.
>>>>>>>
>>>>>>> actually, this is a workaround to the fact that we're not creating all
>>>>>>> platform_devices in arch/arm/mach-omap2/ :-)
>>>>>>>
>>>>>>> If we had the musb allocation there, we could easily handle
>>>>>>> usb_bind_phy()
>>>>>>>
>>>>>>>>> so that's temporary. It might be better than to reintroduce the IDR in
>>>>>>>>> musb_core.c.
>>>>>>>>
>>>>>>>> that’s needed for generic phy framework anyway :-s
>>>>>>>
>>>>>>> right, but generic phy framework can handle everything just fine, the
>>>>>>> only problem is that names are changing.
>>>>>>
>>>>>> right. But if the names change, PHY framework wouldn't be able to return the
>>>>>> reference to the PHY.
>>>>>
>>>>> with my suggestion they can change whenever they want since we're using
>>>>> dev_name() of the just-created musb platform_device. Right ?
>>>>
>>>> right. But the PHY device can be created in a different place from where the
>>>> musb devices are created. And in the PHY framework, the PHY device should have
>>>
>>> this shouldn't be a problem. As long as the phy is created, all should
>>> be good.
>>>
>>>> the list of controller device (names) it can support (PHY framework does not
>>>> maintain a separate list for binding like how we had in USB PHY library). e.g.
>>>> http://www.mail-archive.com/linux-omap@vger.kernel.org/msg92817.html. In such
>>>
>>> this has nothing to do with $subject though. We talk about generic PHY
>>> framework once all these PHY drivers are moved there :-)
>>>
>>>> cases how do we pass the device names. Also will the MUSB core device be
>>>> created before twl4030-usb PHY device?
>>>
>>> and why would that be a problem ? We're telling the framework that the
>>> musb device will use a phy with a name of 'twl4030'. If musb calls
>>> usb_get_phy_dev() and doesn't find a phy, it'll return -EPROBE_DEFER and
>>> try again later.
>>
>> I think we are talking about different problems here ;-) I'm trying to tell
>> using idr in MUSB core is needed for Generic PHY Framework. So in a way, the
>> Generic PHY Framework series depends on this patch series or else MUSB in OMAP3
>> platforms wont work after Generic PHY framework gets merged.
> 
> then you just found a limitation in your framework, right ? :-) I mean,
> imagine if now we have to add an IDR to every single user of your
> framework because they could end up in systems with multiple instances
> of the same IP ?

I raised a similar concern in the PHY framework discussion [1] :-) And since
it's used everywhere else regulators, clkdev, etc.. it's agreed to be used in
PHY as well. Btw if PLATFORM_DEVID_AUTO is used even regulator, clk_get should
fail IMO.

[1] -> http://lkml.indiana.edu/hypermail/linux/kernel/1307.2/03573.html

Thanks
Kishon