From mboxrd@z Thu Jan 1 00:00:00 1970 From: Albert ARIBAUD Date: Wed, 01 Dec 2010 13:47:18 +0100 Subject: [U-Boot] [TEST] arm:board_init_f(): mirror BSS and check after each init_fnc() In-Reply-To: <20101201123530.0CD7C134FEF@gemini.denx.de> References: <1291191050-46774-1-git-send-email-andreas.devel@googlemail.com> <4CF6175D.4020702@free.fr> <4CF61DDD.3060304@gmail.com> <4CF63345.9020006@free.fr> <20101201123530.0CD7C134FEF@gemini.denx.de> Message-ID: <4CF643D6.5030507@free.fr> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: u-boot@lists.denx.de Le 01/12/2010 13:35, Wolfgang Denk a ?crit : > Dear Albert ARIBAUD, > > In message<4CF63345.9020006@free.fr> you wrote: >>> I can not recommend a proper place for all boards cause this parameter >>> is highly is board specific. >> >> Understood -- each board maintainer must select an adapted location >> for the mirror. > > I'm not conviced that ech board maintainer should do that, and I > somewhat doubt the approach taken. > > The main purpose of the patch (as I see it) is to find out which > functions might be candidates to write to bss too early. There is not > that much board specific code involved there. > > If we can generate a slist of functions that are called before > relocation, we can use nm to check if these refernce any sumbols in > bss. We could even generate black / white lists. Just to clarify: I see Andreas' proposal as a debug tool, notintended to ever be applied to a branch, but to be kept and applied to local developer repositories during debug. It seems you, Wolfgang, are considering this more like a tool that would end up in the u-boot source tree and be e.g. run systematically after a build, even if the builder had no intent of debugging. Am I right? If so, then yes, a tool that would analyze the ELF binary would be a good approach. > Best regards, > > Wolfgang Denk Amicalement, -- Albert.