Re: [PATCH v5 5/5] hwmon: Add support for Amphenol ChipCap 2

From: Mark Brown <broonie@kernel.org>
To: Javier Carrasco <javier.carrasco.cruz@gmail.com>
Cc: Rob Herring <robh+dt@kernel.org>,
	Krzysztof Kozlowski <krzysztof.kozlowski+dt@linaro.org>,
	Conor Dooley <conor+dt@kernel.org>,
	Jean Delvare <jdelvare@suse.com>,
	Guenter Roeck <linux@roeck-us.net>,
	Jonathan Corbet <corbet@lwn.net>,
	Liam Girdwood <lgirdwood@gmail.com>,
	Rob Herring <robh@kernel.org>,
	devicetree@vger.kernel.org, linux-kernel@vger.kernel.org,
	linux-hwmon@vger.kernel.org, linux-doc@vger.kernel.org
Subject: Re: [PATCH v5 5/5] hwmon: Add support for Amphenol ChipCap 2
Date: Thu, 18 Jan 2024 16:04:20 +0000	[thread overview]
Message-ID: <f5827df7-34fa-4c11-aca9-ecc6c83c512d@sirena.org.uk> (raw)
In-Reply-To: <30b9ab0c-f3cb-4b5a-a726-de9f7c61769b@gmail.com>

[-- Attachment #1: Type: text/plain, Size: 2554 bytes --]

On Thu, Jan 18, 2024 at 04:30:37PM +0100, Javier Carrasco wrote:
> On 18.01.24 14:49, Mark Brown wrote:
> > On Mon, Jan 15, 2024 at 09:02:25PM +0100, Javier Carrasco wrote:

> >> +static int cc2_enable(struct cc2_data *data)
> >> +{
> >> +	int ret;

> >> +	if (regulator_is_enabled(data->regulator))
> >> +		return 0;

> > This is generally a sign that the regulator API usage is not good, the
> > driver should not rely on references to the regulator held by anything
> > else since whatever else is holding the regulator on could turn it off
> > at any time.  If the driver did the enable itself then it should know
> > that it did so and not need to query.

> The driver handles a dedicated regulator, but I wanted to account for
> the cases where the attempts to enable and disable the regulator fail
> and keep parity. If the disabling attempt fails, will the regulator not
> stay enabled? In that case, an additional call to regulator_enable would
> not be required, right?
> That is the only reason I am using regulator_is_enabled(), but maybe
> things don't work like that.

With exclusive use you can get away with this, you should have a comment
for that case though.

> >> +	ret = regulator_enable(data->regulator);
> >> +	if (ret < 0)
> >> +		return ret;
> >> +
> >> +	/*
> >> +	 * TODO: the startup-delay-us property of the regulator might be
> >> +	 * added to the delay (if provided).
> >> +	 * Currently there is no interface to read its value apart from
> >> +	 * a direct access to regulator->rdev->constraints->enable_time,
> >> +	 * which is discouraged like any direct access to the regulator_dev
> >> +	 * structure. This would be relevant in cases where the startup delay
> >> +	 * is in the range of milliseconds.
> >> +	 */
> >> +	usleep_range(CC2_STARTUP_TIME_US, CC2_STARTUP_TIME_US + 125);

> > Note that the regulator startup delay is the time taken for the
> > regulator to power up so if the device needs additional delay then that
> > will always need to be in addition to whatever the regulator is doing.

> What I mean by that is that the device cannot be ready until the
> regulator powers it up (obvious) plus the start up time of the device
> itself once it gets powered up. So if a regulator takes for example 1 ms
> to power up, the sleep function could (and should) wait for 1 ms longer.

No, the sleep function should do nothing of the sort - if any delay is
neeeded for the regulator it will be handled as part of enabling the
regulator.  This is not exposed to client drivers because it is
transparent to them.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]