From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Date: Tue, 1 Jul 2008 18:44:24 +0200 From: Jean Delvare To: "Jon Smirl" Subject: Re: [PATCH 2/2] Add the of_find_i2c_device_by_node function, V4 Message-ID: <20080701184424.40641cd0@hyperion.delvare> In-Reply-To: <9e4733910807010938w67bab979tfba888641debe72f@mail.gmail.com> References: <20080630230126.12482.87927.stgit@terra> <20080630230128.12482.79570.stgit@terra> <20080701170544.72051418@hyperion.delvare> <9e4733910807010812h30e49144v5b3012e6c156ecca@mail.gmail.com> <20080701182949.12119d6e@hyperion.delvare> <9e4733910807010938w67bab979tfba888641debe72f@mail.gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Cc: Linuxppc-dev@ozlabs.org, Paul Mackerras , i2c@lm-sensors.org List-Id: Linux on PowerPC Developers Mail List List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , On Tue, 1 Jul 2008 12:38:05 -0400, Jon Smirl wrote: > On 7/1/08, Jean Delvare wrote: > > On Tue, 1 Jul 2008 11:12:58 -0400, Jon Smirl wrote: > > > On 7/1/08, Jean Delvare wrote: > > > > I'm fine with this patch. In particular, exporting i2c_bus_type is OK. > > > > It was un-exported only because it had no user left, but it can be > > > > exported again if needed. > > > > > > Another solution would be to move drivers/of/of_i2c into the i2c > > > directory and make it part of i2c core on powerpc builds. > > > > I don't think this is a good idea. Merging arch-specific code (or > > half-arch-specific code in this case) into arch-neutral drivers ends up > > being a pain to maintain. People will keep sending me patches for stuff > > I don't know anything about and can't help with. Having of-specific > > stuff in just one directory as is the case now sounds much better to > > me. All it's missing is a MAINTAINERS entry. > > A side effect of this is that the small pieces of code in drivers/of > have to be compiled into stand alone modules and they may need access > to internal symbols from the subsystem. If they were directly linked > into the subsystems you wouldn't need to make the internal symbols > visible. Do you think you'll need something else than i2c_bus_type? That I don't consider an i2c-core internal. As I said, the lack of export was simply due to a lack of user. > Now the subsystems have to be careful about breaking the > in-kernel, external users of the symbols Same holds if the code is merged into i2c-core. > and we've made it possible for out of tree drivers to get to internal > structures. Hopefully the namespaced exports which some kernel developers are working on, will become a reality soon, to mitigate that kind of issue. That being said... As an upstream kernel developer / maintainer, I don't care about that. If external modules make use of internals they shouldn't touch, and later changes cause them to break, that's none of my business. I leave this to distribution maintainers. -- Jean Delvare