From mboxrd@z Thu Jan 1 00:00:00 1970 From: Arnd Bergmann Subject: Re: [PATCH 1/4] Create generic early_ioremap() support Date: Wed, 18 Dec 2013 11:40:00 +0100 Message-ID: <201312181140.01134.arnd@arndb.de> References: <1385606679-30446-1-git-send-email-msalter@redhat.com> <1387305938.1979.166.camel@deneb.redhat.com> <20131218091348.GB15594@dhcp-16-126.nay.redhat.com> Mime-Version: 1.0 Content-Type: Text/Plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <20131218091348.GB15594@dhcp-16-126.nay.redhat.com> Sender: linux-kernel-owner@vger.kernel.org To: Dave Young Cc: Mark Salter , linux-kernel@vger.kernel.org, Leif Lindholm , Ingo Molnar , linux-arch@vger.kernel.org List-Id: linux-arch.vger.kernel.org On Wednesday 18 December 2013, Dave Young wrote: > > > > I hadn't really thought about it. Putting these things in mm doesn't > > sound unreasonable, but lib seems to have become the place for a lot > > generic code which could also belong in mm, net, etc. > > > > IMHO, move subsystem lib code to subsystem core dir looks more reasonable. > lib should be the place for general kernel stuff. > > I'm not specific for and object this patch, just give my opinion about the lib/.. Traditionally, lib/ is for things that may or may not be used by other optional code, typically drivers, which I think is a reasonable concept. There has been some shift over the years from lib-$(CONFIG_FOO) towards obj-$(CONFIG_FOO), so it's not always used as a actual library code in the linker sense though. I would agree that moving mm related code such as early_ioremap into mm/ is a good idea. For things like the compression helpers lib/ is probably still the best place. Arnd