Re: [PATCH 0/2] rtc: pch-rtc: add Intel Series PCH built-in read-only RTC

From: Alexandre Belloni <alexandre.belloni@bootlin.com>
To: Milton Miller II <miltonm@us.ibm.com>
Cc: Paul Fertser <fercerpav@gmail.com>,
	Ivan Mikhaylov <i.mikhaylov@yadro.com>,
	openbmc@lists.ozlabs.org, linux-kernel@vger.kernel.org,
	Alessandro Zummo <a.zummo@towertech.it>
Subject: Re: [PATCH 0/2] rtc: pch-rtc: add Intel Series PCH built-in read-only RTC
Date: Tue, 17 Aug 2021 22:05:45 +0200	[thread overview]
Message-ID: <YRwWmeQiVC3dGAjH@piout.net> (raw)
In-Reply-To: <OFBF96A764.75CCED3A-ON00258734.0062B30C-00258734.00634222@ibm.com>

On 17/08/2021 18:04:09+0000, Milton Miller II wrote:
> 
> On Aug 16, 2021, Alexandre Belloni wrote:
> >On 15/08/2021 01:42:15+0300, Paul Fertser wrote:
> >> On Tue, Aug 10, 2021 at 06:44:34PM +0300, Ivan Mikhaylov wrote:
> >> > Add RTC driver with dt binding tree document. Also this driver
> >adds one sysfs
> >> > attribute for host power control which I think is odd for RTC
> >driver.
> >> > Need I cut it off and use I2C_SLAVE_FORCE? I2C_SLAVE_FORCE is not
> >good
> >> > way too from my point of view. Is there any better approach?
> >> 
> >> Reading the C620 datasheet I see this interface also allows other
> >> commands (wake up, watchdog feeding, reboot etc.) and reading
> >statuses
> >> (e.g Intruder Detect, POWER_OK_BAD).
> >> 
> >> I think if there's any plan to use anything other but RTC via this
> >> interface then the driver should be registered as an MFD.
> >> 
> >
> >This is not the current thinking, if everything is integrated, then
> >there is no issue registering a watchdog from the RTC driver. I'll
> >let
> >you check with Lee...
> 
> I think the current statement is "if they are truly disjoint 
> hardware controls" then an MFD might suffice, but if they require 
> software cordination the new auxillary bus seems to be desired.
> 

Honestly, the auxiliary bus doesn't provide anything that you can't do
by registering a device in multiple subsystem from a single driver.
(Lee Jones, Mark Brown and I did complain at the time that this was yet
another back channel for misuses).

> >>However, I'm not sure what is the correct interface for
> >poweroff/reboot
> >control.
> 
> While there is a gpio interface to a simple regulator switch,
> the project to date has been asserting direct or indirect 
> gpios etc to control the host.   If these are events to 
> trigger a change in state and not a direct state change
> that some controller trys to follow, maybe a message delivery 
> model?   (this is not to reboot or cycle the bmc).
> 
> milton

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