Re: [RFC PATCH 1/2] Regulator: Core: Add clock-enable to fixed-regulator

From: Philippe Schenker <philippe.schenker@toradex.com>
To: "broonie@kernel.org" <broonie@kernel.org>
Cc: "lgirdwood@gmail.com" <lgirdwood@gmail.com>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>
Subject: Re: [RFC PATCH 1/2] Regulator: Core: Add clock-enable to fixed-regulator
Date: Wed, 7 Aug 2019 06:46:50 +0000	[thread overview]
Message-ID: <9b3f6c47803bbd672896b682469109bb7201a04b.camel@toradex.com> (raw)
In-Reply-To: <20190806182606.GG4527@sirena.org.uk>

On Tue, 2019-08-06 at 19:26 +0100, Mark Brown wrote:
> On Tue, Aug 06, 2019 at 12:57:32PM +0000, Philippe Schenker wrote:
> > On Mon, 2019-08-05 at 17:37 +0100, Mark Brown wrote:
> > > So the capacitor on the input of the p-FET is keeping the switch
> > > on?
> > > When I say it's not switching with the clock I mean it's not
> > > constantly
> > > bouncing on and off at whatever rate the clock is going at.
> > Ah, that's what you mean. Yes, the capacitor gets slowly charged
> > with
> > the
> > resistor but nearly instantly discharged with the n-FET. So this
> > capacitor
> > is used as a Low-Pass filter to get the p-FET to be constantly
> > switched.
> > It is not bouncing on and off with the clock but rather it is
> > switched
> > constantly.
> 
> Good, I guess this might be part of why it's got this poor ramp time.

Yes, I think so too.

> 
> > > I think you are going to end up with a hack no matter what.
> > That's exactly what I'm trying to prevent. To introduce a fixed
> > regulator that can have a clock is not a hack for me.
> > That the hardware solution is a hack is debatable yes, but why
> > should I
> > not try to solve it properly in software?
> 
> A lot of this discussion is around the definition of terms like "hack"
> and "proper".
> 
> > In the end I just want to represent our hardware in software. Would
> > you
> > agree to create a new clock-regulator.c driver?
> > Or would it make more sense to extend fixed.c to support clocks-
> > enable
> > without touching core?
> 
> At least a separate compatible makes sense, I'd have to see the code
> to
> be clear if a completely separate driver makes sense but it'll need
> separate ops at least.  There'd definitely be a lot of overlap though
> so
> it's worth looking at.

Okay, thanks for discussion! I will try to make something that will fit
in mainline kernel and I will learn more about the regulator subsystem
in general so I can make a solution that fits.
But I'll need some time to do that. I will for sure link to that
discussion when I send the patch.

Philippe