From mboxrd@z Thu Jan 1 00:00:00 1970 From: Dmitry Torokhov Subject: Re: [PATCH] arm64: dts: rockchip: enable touchpad button for rk3399-gru-kevin Date: Tue, 10 Oct 2017 10:25:16 -0700 Message-ID: <20171010172516.GA1681@dtor-ws> References: <20171009184028.11829-1-kernel@esmil.dk> <4174027.Z5MyPRUDcj@phil> <20171010170355.GA136735@google.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Content-Disposition: inline In-Reply-To: <20171010170355.GA136735@google.com> Sender: linux-kernel-owner@vger.kernel.org To: Brian Norris Cc: Emil Renner Berthing , Heiko Stuebner , Doug Anderson , Rob Herring , Mark Rutland , devicetree@vger.kernel.org, linux-arm-kernel@lists.infradead.org, linux-rockchip@lists.infradead.org, linux-kernel@vger.kernel.org List-Id: devicetree@vger.kernel.org On Tue, Oct 10, 2017 at 10:03:57AM -0700, Brian Norris wrote: > + Dmitry, who can probably answer better than I can > > On Tue, Oct 10, 2017 at 01:04:20PM +0200, Emil Renner Berthing wrote: > > On 10 October 2017 at 11:33, Heiko Stuebner wrote: > > > Am Montag, 9. Oktober 2017, 20:40:28 CEST schrieb Emil Renner Berthing: > > >> Adding the linux,gpio-keymap entry also has > > >> the side-effect of making the driver register > > >> the touchpad as a touchpad rather than another > > >> touchscreen. > > >> > > >> The index for BTN_LEFT was found by trial and error. > > >> > > >> Signed-off-by: Emil Renner Berthing > > > > > > looks good to me. I'll just give the ChromeOS people a chance to > > > respond before applying [maybe they can even tell what the other > > > keys are :-) ] They are not connected. > > > > Good plan. As far as I can tell their kernel doesn't send anything but > > BTN_LEFT either. It's just that the controller supports multiple > > buttons, but not all are necessarily connected to any button. > > Their version of the driver has logic to detect the proper keymap > > whereas the mainline driver gets it from this devicetree entry. > > I believe Emil is mostly correct. We only are using a BTN_LEFT in our > copy of the driver, and I'm not sure the others are connected to > anything. As Dmitry has explained to me (when I also wondered about > "detect[ing] the proper keymap"), we don't exactly "auto"discover > things; we just heuristically determine the configurations we know are > used on Chrome devices. Right, we had a "legacy" GPIO/button mapping on original Pixels that used T9 multitouch reports, then we switched to a bit different mapping for newer devices that use T100 multitouch reports. We should really have explicit map, like we have in upstream driver. > > So from the sound of it, this keymap property is reasonable. I haven't > actually checked what our driver "auto" detects on gru-kevin, but if > this works then I'll bet it's correct :) > > IOW, I don't know enough to ACK this patch, but I'm certainly not going > to NAK it. I think I can ACK it: Acked-by: Dmitry Torokhov Thanks. -- Dmitry