From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1755752Ab2ECNhR (ORCPT ); Thu, 3 May 2012 09:37:17 -0400 Received: from mail-yw0-f46.google.com ([209.85.213.46]:55097 "EHLO mail-yw0-f46.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753154Ab2ECNhO (ORCPT ); Thu, 3 May 2012 09:37:14 -0400 MIME-Version: 1.0 In-Reply-To: <20120503131904.GA11741@polaris.bitmath.org> References: <1335175627-2270-1-git-send-email-rydberg@euromail.se> <20120503122359.GA11552@polaris.bitmath.org> <20120503131904.GA11741@polaris.bitmath.org> Date: Thu, 3 May 2012 15:37:13 +0200 Message-ID: Subject: Re: [PATCH v3 0/6] hid: Introduce device groups From: Benjamin Tissoires To: Henrik Rydberg Cc: Jiri Kosina , Dmitry Torokhov , linux-input@vger.kernel.org, linux-kernel@vger.kernel.org, Stephane Chatty Content-Type: text/plain; charset=ISO-8859-1 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Thu, May 3, 2012 at 3:19 PM, Henrik Rydberg wrote: >> > 1) Add the devices in question back to the have_special_drivers list. >> >> Well... The device presents valid mouse and keyboard interface that >> should be handled by hid-generic. >> The behavior of this particular device is the following: >> - when 1 finger is in use, then it sends events over the mouse interface >> - when 2 fingers are present, it sends events over the multitouch interface >> - when you physically trigger the switch mode button, a keyboard >> appears and it sends key events over the keyboard interface, and >> eventually mouse events if you press the "mouse" key.... ;-) >> >> This crap is all inherited by the fact that Microsoft do not want to >> handle indirect touch, and the device maker found this solution to >> counter this. >> >> To sum up, adding it to the have_special_drivers driver list won't >> work as we need part of the device to be handled by hid-generic. > > So was this particular device never listed in have_special_drivers? No, and that's the way it should (not being part of have_special_driver). > >> > 2) Add the interface type to the group descision, which should >> > probably be done anyway. I have a patch in the pipe that, will send it >> > later today. >> >> A simpler solution consists in adding the macros HID_USB_MT_DEVICE(v, >> p) and HID_BLUETOOTH_MT_DEVICE(v, p) as you had introduced in a >> earlier patch (I don't know why it disappeared). > > No, the specific entries in the hid-multitouch device list matches any > group, so those defines were simplified away in the second version. disagree: a device can present several interface (because it has several "devices") and only those presenting Contact ID can and should be handled by hid-multitouch. Cypress for instance presents one interface for the multitouch layer, and one other for specific controls that are seen as a keyboard. However, in this particular case, I'm not sure we want to show this interface to the end user.... ;-) > >> The problem came out because: >> - hid-multitouch registered the triplet BUS_USB / VID / PID. >> - For each interface, it asks udev (or the kernel) which driver to >> use, and whatever .group was, it was always hid-multitouch that came >> out. >> >> So it's just safer to specify the group for all multitouch devices. > > This is still confusing. I thought the real problem was that the > non-mt interfaces do not match hid-generic. Solution 2) should take > care of that. What I don't understand is how those other interfaces > came to be handled by hid-generic before this patch, unless this > device was never listed in have_special_driver. The think is that they do match hid-generic (they get the group HID_GROUP_GENERIC). However they also match hid-multitouch (as hid-multitouch does not ask for a particular group). So, if hid-multitouch is loaded __before__ hid-generic, it will be given the device whatever the match with hid-generic. And again, yes it was never listed in have_special_driver. > > Are we talking about USB_DEVICE_ID_TOPSEED2_PERIPAD_701 here? yep - For consistency, I'd rather specifying the group for any devices. This because hid-multitouch can not handle other interfaces than multitouch one. Though the catchall is interesting in the sense that it may help us to hide unwanted interfaces. - For backward compatibility, we should adapt each device (currently, I only spotted this particular one) to decide if we need to catch the group or not. Jiri, any thought? Benjamin > > Henrik