RE: [PATCH 1/3] mfd: ds1374: Introduce Dallas/Maxim DS1374 MFD core driver

From: "Johnson CH Chen (陳昭勳)" <JohnsonCH.Chen@moxa.com>
To: Alexandre Belloni <alexandre.belloni@bootlin.com>,
	Guenter Roeck <linux@roeck-us.net>
Cc: Lee Jones <lee.jones@linaro.org>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
	"linux-rtc@vger.kernel.org" <linux-rtc@vger.kernel.org>,
	"linux-watchdog@vger.kernel.org" <linux-watchdog@vger.kernel.org>,
	Alessandro Zummo <a.zummo@towertech.it>,
	Wim Van Sebroeck <wim@linux-watchdog.org>
Subject: RE: [PATCH 1/3] mfd: ds1374: Introduce Dallas/Maxim DS1374 MFD core driver
Date: Tue, 23 Jun 2020 07:12:42 +0000	[thread overview]
Message-ID: <HK2PR01MB3281B2BC29C32315759B7ED2FA940@HK2PR01MB3281.apcprd01.prod.exchangelabs.com> (raw)
In-Reply-To: <20200622150126.GJ131826@piout.net>

Hi,

> On 22/06/2020 07:26:56-0700, Guenter Roeck wrote:
> > On Mon, Jun 22, 2020 at 12:14:13PM +0100, Lee Jones wrote:
> > > On Mon, 22 Jun 2020, Johnson CH Chen (陳昭勳) wrote:
> > >
> > > > Dallas/Maxim DS1374 is a counter designed to continuously count
> > > > time in seconds. It provides an I2C interface to the host to
> > > > access RTC clock or Alarm/Watchdog timer.
> > > >
> > > > Add MFD Core driver, supporting the I2C communication to RTC and
> > > > Watchdog devices.
> > > >
> > > > Signed-off-by: Johnson Chen <johnsonch.chen@moxa.com>
> > > > ---
> > > >  drivers/mfd/Kconfig  |  11 +++++
> > > >  drivers/mfd/Makefile |   2 +
> > > >  drivers/mfd/ds1374.c | 101
> > > > +++++++++++++++++++++++++++++++++++++++++++
> > > >  3 files changed, 114 insertions(+)  create mode 100644
> > > > drivers/mfd/ds1374.c
> > >
> > > Not sure I see the point of this driver.
> > >
> >
> > Not entirely sure either. Seems to me the idea is to use the watchdog
> > subsystem for watchdog functionality, but that is just a guess and not
> > really necessary (the conversion could be done in the rtc driver).
> > I don't think the code as written works - the rtc code uses a mutex
> > which the watchdog driver obviously isn't aware of. The mutex would
> > have to be moved into the mfd code, with respective access functions.
> >
> > Overall this adds a lot of complexity, and it seems the
> > interdependencies between rtc and watchdog functionality are not well
> > understood. Plus, other watchdog drivers have recently been added to
> > other rtc clock chips, so this adds some inconsistencies in the rtc
> > subsystem. Are we going to see this change for all those combined
> rtc/watchdog drivers ?
> > If so, it might make sense to communicate that now to ensure consistency.
> >
> 
> I read the datasheet again and I agree the watchdog part can live in the rtc
> driver. As only the RTC alarm and the watchdog are mutually exclusive. I don't
> think an MFD driver is necessary. Converting the current driver to the
> watchdog subsystem seems to be the correct way forward.
> 
Thanks for your good thinking. If we want to add watchdog function such as 
"nowayout" to the driver, it's good to try to upstream this in rtc-ds1374.c in 
rtc subsystem?

It seems like more complexity if we want to separate rtc and watchdog for one 
chip supports. For one chip which supports rtc/alarm and watchdog, can we 
upstream rtc/alarm and watchdog functions to these driver no matter where it's 
in rtc or watchdog subsystem?

> --
> Alexandre Belloni, Bootlin
> Embedded Linux and Kernel engineering
> https://bootlin.com

Best regards,
Johnson