Re: [PATCH 2/2] eeepc-wmi: Add support for T101MT Home/Express Gate key

From: Dmitry Torokhov <dmitry.torokhov@gmail.com>
To: Chris Bagwell <chris@cnpbagwell.com>
Cc: Corentin Chary <corentin.chary@gmail.com>,
	Matthew Garrett <mjg@redhat.com>,
	acpi4asus-user@lists.sourceforge.net,
	platform-driver-x86@vger.kernel.org,
	linux-kernel@vger.kernel.org, vojtech@suze.cz
Subject: Re: [PATCH 2/2] eeepc-wmi: Add support for T101MT Home/Express Gate key
Date: Sun, 27 Mar 2011 12:11:16 -0700	[thread overview]
Message-ID: <20110327191116.GB31692@core.coreip.homeip.net> (raw)
In-Reply-To: <AANLkTikBnAdXHR4iqzq9Zd6D-OYe4yp-zELPh6iLr6sj@mail.gmail.com>

On Sun, Mar 27, 2011 at 01:32:45PM -0500, Chris Bagwell wrote:
> On Sun, Mar 27, 2011 at 12:13 PM, Dmitry Torokhov
> <dmitry.torokhov@gmail.com> wrote:
> > On Fri, Mar 25, 2011 at 01:58:24PM -0500, Seth Forshee wrote:
> >> On Fri, Mar 25, 2011 at 10:03:33AM -0700, Dmitry Torokhov wrote:
> >> > > So, would the following make sense (KEY_TBD is may abbreviation for "key
> >> > > to be determinded"):
> >> > >
> >> > >  - On press (0xe4), send KEY_TBD, value = 1
> >> > >  - On hold (0xea), send KEY_TBD, value = 2
> >> > >  - On release (0xe5), send KEY_TBD, value = 0
> >> >
> >> > Yes.
> >>
> >> There's a little problem with this. The driver uses sparse-keymap, which
> >> calls input_report_key(), which is doing this:
> >>
> >>       input_event(dev, EV_KEY, code, !!value);
> >
> > Ah, right, it really did not have in mind passing on auto-repeating keys
> > as is when it was written.
> >
> >>
> >> I don't see the hold events in userspace, so they must be getting
> >> dropped as duplicates as you suggested.
> >>
> >> I'm not sure who is at fault here. Should input_report_key() not be
> >> forcing value to 0 or 1? Or should sparse-keymap be calling
> >> input_event() directly? Or is sparse-keymap the wrong tool for handling
> >> keys this way?
> >
> > I do not thik we shoudl burden sparse-keymap with handling repeats. Just
> > mark the device as auto-repeating (by setting EV_REP bit) and let input
> > core autorepeat logic do it for you.
> 
> It would be nice to have this key sent roughly every 2 seconds to
> re-execute a rotate script using generic key-to-script support in most
> X environments.  So to use kernel auto-repeat we'd need to bump up
> REP_DELAY quite a bit or do filtering in userspace.  If userspace is
> filtering then its not much more work to do their own auto-repeat as
> well.

That is true as well.

> 
> If we do set up auto-repeat and increase REP_DELAY, I'm guessing this
> would enable auto-repeat for all other keys defined in driver?  That
> needs to have some thought on if could have negative impact (any other
> keys not using auto-release?).

Right. Right now there are 4 autoreprat options (in general):

- hardware autorepeat (if hardware supports it);
- input core software autorepeat (one delay and rate per input device);
- driver-implemented software autorepeat - in cases when different
  repeat rate is needed;
- userspace autorepeat (like X does nowadays);

Well, 4th option is not mutually exclusive with the other 3...

-- 
Dmitry