Re: Enabling pmbus power control

From: Mark Brown <broonie@kernel.org>
To: Zev Weiss <zev@bewilderbeest.net>
Cc: linux-hwmon@vger.kernel.org, Jean Delvare <jdelvare@suse.com>,
	Andrew Jeffery <andrew@aj.id.au>,
	openbmc@lists.ozlabs.org, Liam Girdwood <lgirdwood@gmail.com>,
	linux-kernel@vger.kernel.org, Guenter Roeck <linux@roeck-us.net>
Subject: Re: Enabling pmbus power control
Date: Tue, 20 Apr 2021 18:15:40 +0100	[thread overview]
Message-ID: <20210420171540.GG6073@sirena.org.uk> (raw)
In-Reply-To: <YH8D+LWxWPqnFd2h@hatter.bewilderbeest.net>

[-- Attachment #1: Type: text/plain, Size: 2013 bytes --]

On Tue, Apr 20, 2021 at 11:40:24AM -0500, Zev Weiss wrote:
> On Tue, Apr 20, 2021 at 11:13:18AM CDT, Mark Brown wrote:

> > I already suggested writing a driver or drivers that represent the
> > hardware you have, that advice remains.  It's hard to follow what you
> > were trying to say with your long mail earlier today but it seems like

> That email was an attempt to explain why writing a driver for the specific
> hardware devices we're powering seems like a poor fit to me.  To summarize:

>  - There's a wide variety of different devices potentially behind an
> LM25066.

This is true for lots of hardware, we still integrate things into
frameworks.

>  - A hypothetical driver for any one of them would be completely
> non-specific to that device and functionally identical to a driver    for
> any other, because the only hardware it would actually be    touching is the
> LM25066, and in ways that are again completely    non-specific to anything
> but the LM25066 itself.

I don't see why that would be the case at all.  Even within the indended
application as a power controller for a hotpluggable bus there's plenty
of potential for integration into a wider representation of the socket
things get inserted into - for example I've worked with buses that had
support for operator signalling of hotplug (buttons to press to initiate
hot removal, with lights to signal when a clean shutdown of the card had
been completed), you might also want to have additional environment
monitoring and of course the labelling that I mentioned in an earlier
post.  I can imagine you probably have some other connection of some
kind to the host too (eg, network ports) to join up and perhaps sync
hotplug for.

> I'm not at all opposed to using a abstractions or frameworks (I'd very much
> like to do so in fact).  The problem is that I've thus far been unable to
> determine exactly what the appropriate one is.

Perhaps you need to write some kind of generic system for hotplugging
modules if you can't find one.

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