From mboxrd@z Thu Jan 1 00:00:00 1970 From: arnd@arndb.de (Arnd Bergmann) Date: Tue, 18 Jun 2013 16:13:16 +0200 Subject: [PATCH v4 4/4] arm: add basic support for Rockchip RK3066a boards In-Reply-To: <201306180026.25973.heiko@sntech.de> References: <201306131658.36584.heiko@sntech.de> <20130617213515.GK28497@quad.lixom.net> <201306180026.25973.heiko@sntech.de> Message-ID: <201306181613.16911.arnd@arndb.de> To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org On Tuesday 18 June 2013, Heiko St?bner wrote: > > Side comment: I think it would be nice if the generic code did this > > init if a l2x0 device node was in the device tree, since the only > > reason to override init_machine is to do this call in addition to > > of_platform_populate(). > > Arnd said similar things in the initial version :-). > > Currently I'm wondering if it wouldn't be enough to call always l2x0_of_init > somewhere, because it does the checking for the dt nodes itself already. > > The only obstacle would be platforms having the need to use special aux-values > or which are currently calling the function from some other parts of the boot > process (tegra inits the cache in its tegra_init_early function for example). I think we can handle this by ensuring the function only gets called once, and all platforms with special requirements call it before the common code does. I tried to understand what the requirement for non-zero argument is however and couldn't figure it out. Shouldn't we just be able to specify all the bits as DT properties all the time? Arnd From mboxrd@z Thu Jan 1 00:00:00 1970 From: Arnd Bergmann Subject: Re: [PATCH v4 4/4] arm: add basic support for Rockchip RK3066a boards Date: Tue, 18 Jun 2013 16:13:16 +0200 Message-ID: <201306181613.16911.arnd@arndb.de> References: <201306131658.36584.heiko@sntech.de> <20130617213515.GK28497@quad.lixom.net> <201306180026.25973.heiko@sntech.de> Mime-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Return-path: In-Reply-To: <201306180026.25973.heiko@sntech.de> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: "linux-arm-kernel" Errors-To: linux-arm-kernel-bounces+linux-arm-kernel=m.gmane.org@lists.infradead.org To: Heiko =?iso-8859-1?q?St=FCbner?= Cc: Mike Turquette , devicetree-discuss@lists.ozlabs.org, Rob Herring , Grant Likely , Olof Johansson , "linux-arm-kernel@lists.infradead.org" List-Id: devicetree@vger.kernel.org On Tuesday 18 June 2013, Heiko St=FCbner wrote: > > Side comment: I think it would be nice if the generic code did this > > init if a l2x0 device node was in the device tree, since the only > > reason to override init_machine is to do this call in addition to > > of_platform_populate(). > = > Arnd said similar things in the initial version :-). > = > Currently I'm wondering if it wouldn't be enough to call always l2x0_of_i= nit = > somewhere, because it does the checking for the dt nodes itself already. > = > The only obstacle would be platforms having the need to use special aux-v= alues = > or which are currently calling the function from some other parts of the = boot = > process (tegra inits the cache in its tegra_init_early function for examp= le). I think we can handle this by ensuring the function only gets called once, and all platforms with special requirements call it before the common code does. I tried to understand what the requirement for non-zero argument is however and couldn't figure it out. Shouldn't we just be able to specify all the bits as DT properties all the time? Arnd