Re: [PATCH 0/7] Add USB remote wakeup driver

From: Chunfeng Yun <chunfeng.yun@mediatek.com>
To: Rob Herring <robh@kernel.org>
Cc: Felipe Balbi <felipe.balbi@linux.intel.com>,
	Matthias Brugger <matthias.bgg@gmail.com>,
	Mathias Nyman <mathias.nyman@intel.com>,
	"Mark Rutland" <mark.rutland@arm.com>,
	Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
	Catalin Marinas <catalin.marinas@arm.com>,
	Will Deacon <will.deacon@arm.com>,
	Jean Delvare <jdelvare@suse.de>,
	"Sean Wang" <sean.wang@mediatek.com>,
	<devicetree@vger.kernel.org>, <linux-kernel@vger.kernel.org>,
	<linux-usb@vger.kernel.org>,
	<linux-arm-kernel@lists.infradead.org>,
	<linux-mediatek@lists.infradead.org>
Subject: Re: [PATCH 0/7] Add USB remote wakeup driver
Date: Thu, 21 Dec 2017 14:48:24 +0800	[thread overview]
Message-ID: <1513838904.17567.177.camel@mhfsdcap03> (raw)
In-Reply-To: <20171215205504.r6ol7fbbeyghb73w@rob-hp-laptop>

On Fri, 2017-12-15 at 14:55 -0600, Rob Herring wrote:
> On Sat, Dec 09, 2017 at 04:45:29PM +0800, Chunfeng Yun wrote:
> >     These patches introduce the SSUSB and SPM glue layer driver which is
> > used to support usb remote wakeup. Usually the glue layer is put into
> > a system controller, such as PERICFG module.
> >     The old way to support usb wakeup is put into SSUSB controller drivers,
> > including xhci-mtk driver and mtu3 driver, but there are some problems:
> >     1. can't disdinguish the relation between glue layer and SSUSB IP
> >        when SoCs supports multi SSUSB IPs;
> >     2. duplicated code for wakeup are put into both xhci-mtk and mtu3
> >        drivers;
> >     3. the glue layer may vary on different SoCs with SSUSB IP, and will
> >        make SSUSB controller drivers complicated;
> >     In order to resolve these problems, it's useful to make the glue layer
> > transparent by extracting a seperated driver, meanwhile to reduce the
> > duplicated code and simplify SSUSB controller drivers.
> 
> Both the driver and binding look overly complicated to me when it looks 
> like you just have 2 versions of enable/disable functions which modify 
> a single register. The complexity may be justified if this was a common 
> binding and driver, but it is not.
> 
> You already have a phandle to the system controller. Can't you add cells 
> to it to handle any differences between instances? That and SoC specific 
> compatible strings should be enough to handle differences.
Yes, adding cells will also work well, I'll try it, thanks a lot
> 
> Rob