Re: [PATCH] usb: chipidea: Grab the (legacy) USB PHY by phandle first

[Paul Kocialkowski]

Hi,

On Thu, 2019-01-17 at 06:44 +0000, Peter Chen wrote:
>  
> > On Wed, 2019-01-16 at 14:44 +0100, Thomas Petazzoni wrote:
> > > Well prior to your code, there was already a possibility for both
> > > ci->phy and ci->usb_phy to be valid. I don't think it's really useful
> > > to avoid the fallback when a generic PHY has already been found, it's
> > > confusing. If really you want to clarify that, it should be:
> > > 
> > > 	/* Let's first try to find a generic PHY */
> > > 	ci->phy = devm_phy_get(dev->parent, "usb-phy");
> > > 	if (IS_ERR(ci->phy)) {
> 
> Please consider -EPROBE_DEFER case
> 
> > > 		/* Fall back to legacy USB PHY */
> > > 		ci->usb_phy = devm_usb_get_phy_by_phandle(dev->parent, "phys",
> > 0);
> 
> Please consider -EPROBE_DEFER case
> 
> > > 		if (IS_ERR(ci->usb_phy))
> > > 			ci->usb_phy = devm_usb_get_phy(dev->parent,
> > USB_PHY_TYPE_USB2);
> > > 	}
> > > 
> 
> Below code 
>                 if (IS_ERR(ci->phy) && IS_ERR(ci->usb_phy)) {
>                         ret = -EPROBE_DEFER;
>                         goto ulpi_exit;
>                 }    
> 
> needs to change as:
> 
>                 if (PTR_ERR(ci->phy) ==-EPROBE_DEFER  ||  PTR_ERR(ci->usb_phy) == -EPROBE_DEFER) {
>                         ret = -EPROBE_DEFER;
>                         goto ulpi_exit;
>                 }    

Well I think this was more of a general outline than proper code to be
included in the driver anyway :)

I'm rather considering implementing what Thomas suggested at first,
which is going for the fallback even if a generic PHY was found, as to
stick more closely to the existing behavior.

Cheers,

Paul

> Peter
> 
> > > With that, you would only have either ci->phy or ci->usb_phy be valid,
> > > and never both. With  your change, you can have ci->phy and
> > > ci->usb_phy both be valid if the legacy USB PHY was found using
> > > devm_usb_get_phy_by_phandle(), but not if we fell back to
> > > devm_usb_get_phy().
> > 
> > Okay that makes sense, your suggestion is indeed more consistent with the existing
> > behavior. I'll go with that in the next revision!
> > 
> 
> 
-- 
Paul Kocialkowski, Bootlin
Embedded Linux and kernel engineering
https://bootlin.com