From mboxrd@z Thu Jan 1 00:00:00 1970 From: broonie@opensource.wolfsonmicro.com (Mark Brown) Date: Tue, 2 Oct 2012 13:58:20 +0100 Subject: [PATCHv3 1/4] mfd: smsc: Add support for smsc gpio io/keypad driver In-Reply-To: <20121002124321.GD30762@arwen.pp.htv.fi> References: <1349089282-22105-1-git-send-email-sourav.poddar@ti.com> <20121001114454.GA9170@sirena.org.uk> <20121001120907.GK4360@opensource.wolfsonmicro.com> <20121002123856.GX4360@opensource.wolfsonmicro.com> <20121002124321.GD30762@arwen.pp.htv.fi> Message-ID: <20121002125819.GY4360@opensource.wolfsonmicro.com> To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org On Tue, Oct 02, 2012 at 03:43:22PM +0300, Felipe Balbi wrote: > BTW, OTOH writing all children into the DT actually describes the HW, > no ? And depending on the device I feel it'd be better to write that Well, it depends on the hardware. Some hardware has a bunch of nice, neat IPs which can usefully be reproduced and which map sensibly onto OS abstractions but a lot of it doesn't and frequently the abstractions which Linux wants to use don't bear a huge resemblance to the hardware (and Linux's ideas can change over time, as with the clock API being factored out for example). > data to DT. Think of twlxxxx (TI's PMICs), we might have completely > unrelated drivers using one of TWL's GPIO lines as an interrupt source. > If that particular children isn't listed in DT, it can't be used as an > interrupt-parent, right ? You can have the interrupt controller there without having to list every IP in the device, just make the parent device the interrupt controller to DT.