[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20170717164006.GB11246@kroah.com>
Date: Mon, 17 Jul 2017 18:40:06 +0200
From: Greg KH <gregkh@...uxfoundation.org>
To: Jiri Pirko <jiri@...nulli.us>
Cc: linux-kernel@...r.kernel.org, davem@...emloft.net,
mchehab@...nel.org, jikos@...nel.org
Subject: Re: [patch 0/2] add actuators support
On Mon, Jul 17, 2017 at 06:34:58PM +0200, Jiri Pirko wrote:
> Mon, Jul 17, 2017 at 06:28:38PM CEST, gregkh@...uxfoundation.org wrote:
> >On Mon, Jul 17, 2017 at 06:19:12PM +0200, Jiri Pirko wrote:
> >> From: Jiri Pirko <jiri@...lanox.com>
> >>
> >> I am owner of height adjustable desk and naturally, as it has an USB
> >> interface, I need to controll it from my computer. Started to think
> >> about what would be the best way, I realized that I need to introduce
> >> a new driver class in kernel. The reason is a need to have one API
> >> for all possible kinds of actuator devices (USB, I2C, gpio, etc).
> >
> >Why does this have to be a kernel driver at all? Your USB driver should
> >really just be a simple userspace application (use libusb to have it
> >work on all operating systems.)
>
> Yeah, I was thinking about it as well. To have some kind of single
> purpose app like sispmctl (control of surge protector). But that would
> limit you only for this specific device. And for multiple devices you
> would have to have multiple apps.
>
> So it seems to me like a suitable solution to have this as a driver
> class.
For USB devices, especially ones as simple as this, we don't want them
to have kernel drivers. It's just overkill.
And never create a class if you only have one type of device, otherwise
you don't know if you really have it properly defined or not. 3 is a
good number to start with.
sorry,
greg k-h
Powered by blists - more mailing lists