Date:	Thu, 25 Feb 2010 23:44:22 -0800
From:	Dmitry Torokhov <dmitry.torokhov@...il.com>
To:	Stefan Achatz <stefan_achatz@....de>
Cc:	Jiri Kosina <jkosina@...e.cz>,
	Jussi Kivilinna <jussi.kivilinna@...et.fi>, wylda@...ny.cz,
	Pavel Machek <pavel@...e.cz>,
	Alessandro Guido <ag@...ssandroguido.name>,
	Tomas Hanak <tomas.hanak@...il.com>,
	Jason Noble <nobleja@...ezero.com>, simon.windows@...il.com,
	Sean Hildebrand <silverwraithii@...il.com>,
	Sid Boyce <sboyce@...eyonder.co.uk>,
	Henning Glawe <glaweh@...ian.org>, linux-input@...r.kernel.org,
	linux-kernel@...r.kernel.org
Subject: Re: [PATCH 3/3] Adding documentation to sysfs attributes of roccat
 kone driver

On Tue, Feb 23, 2010 at 09:03:08AM +0100, Stefan Achatz wrote:
> > Thank you for adding documentation for the sysfs attributes. I have one
> > more request though - the documentation adds a lot of anformation about
> > whta the fields mean but unfortunately it is silent on _why_ all these
> > fields are needed.
> 
> > For example, why weight is interesting? Is it just shown just because we can?
> > What woudl yuser do with this information? Why would I need to
> > manipulate several profiles instead of letting udev configure desired
> > state for me every time mouse is plugged into the box? Should not TCU
> > calibration be simply performed during driver binding instead of
> > on-demand from userspace? And so forth.
> 

> I would have done all with libusb in userspace if I had not some
> problems with it:
> - devio wouldn't let me access the mouseprofiles because the
> manufacturer uses a way to access them check_ctrlrecip() forbids.
> - removing usbhid temporarily would render the mouse not responding
> for the time the action takes.
> - readding usbhid is quirky and one needs to replug the mouse if the
> software fails in some way.
> 
> I'm unsure if all this functionality is wanted inside the kernel. This
> device is Hardware for gamers which i find interesting to be supported
> by linux also.  In my attempt to raise acceptance for kernel
> integration i moved all unnecessary functionality like calculating
> real weight and dpi values into my userland tools to keep only the
> basic functions in the module.  This driver is available as externally
> compiled module for some months now, in my eyes its the most elegant
> (and the only workable) solution and these patches are also an attempt
> to find out if this project generally has a chance to be integrated in
> the kernel (but there is also a module for sonys ps3 controller).
> There are more devices from this manufacturer which have similar
> functions like a keyboard with a whole bunch of macro keys where the
> macros are stored in the keyboard which I find useful as (amateur)
> programmer and admin.
> 
> The weight is really just a "because I can" value.
> 
> Switching the profiles is done via software or pressing a button on
> the mouse and is useful if you wan't to play games with high
> sensitivity and use a low sensitivity profile for detail work in a
> graphics application. Also different button mappings (the mouse can
> play macros with can contain keyboard and mouse button events) are
> useful for different applications. As example some applications
> (viewers) do pagescroling with "page up/down" some with arrow keys.
> You are free to change profiles in the mouse or load new profiles as
> you like.  You can use a script to backup profiles and load a whole
> set of specific profiles before starting maybe a game and restore the
> old profiles on exit.
> 
> TCU calibration is only needed if you switch the surface the mouse is
> being used on. Also you can save and restore already aquired
> calibration data with the settings attribute without the need of
> recalibration.
> 
> Maybe this answers some of the questions. I'm new to this kernel
> patching stuff and I might benefit from verbatim questions and
> statements.

OK, I see. In this case you need to separate "because I can" attributes
that are not really useful to the users from the ones that actually
provide useful data and resubmit.

Version, firmware and weight should go, maybe DPI as well. You should
use VID/PID already exported in the kernel to locate the devices you are
interested in.

TCU - you probaly want it write only and block until device is
calibrated.

Sysfs attributes altering state of the device should not be user
writable (so 0644 or 0600).

There should be some locking around sysfs methods - will the device
survive sumiltaneous access from different processes at the same time?

Thanks.

-- 
Dmitry
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists