From mboxrd@z Thu Jan 1 00:00:00 1970 From: Andrew Cooper Subject: Re: [PATCH] Rename credit2 names to csched2_* Date: Fri, 1 Mar 2013 10:38:32 +0000 Message-ID: <51308528.5060201@citrix.com> References: <550a413da1ac4333d4ec.1361967571@nehalem1> <512E0D9A02000078000C17E2@nat28.tlf.novell.com> <5130454A.9090802@ts.fujitsu.com> <513072F602000078000C2476@nat28.tlf.novell.com> <51306B3E.6030805@ts.fujitsu.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <51306B3E.6030805@ts.fujitsu.com> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xen.org Errors-To: xen-devel-bounces@lists.xen.org To: Juergen Gross Cc: Jan Beulich , xen-devel List-Id: xen-devel@lists.xenproject.org On 01/03/13 08:47, Juergen Gross wrote: > On 01.03.2013 09:20, Jan Beulich wrote: >>>>> On 01.03.13 at 07:06, Juergen Gross wrote: >>> On 27.02.2013 13:43, Jan Beulich wrote: >>>>>>> On 27.02.13 at 13:19, Juergen Gross wrote: >>>>> Functions, variables, structures and macros in the credit2 scheduler had >>>>> partially the same names as in the credit scheduler. This makes it hard to >>>>> find the correct functions in backtraces or cscope. >>>>> >>>>> Rename all names in credit2 from csched_*/CSCHED_* to csched2_*/CSCHED2_* >>>> I don't think this is a suitable approach - if anything, we should aim >>>> at having printed static symbols (in backtraces etc) prefixed with >>>> their source file name. I would even question quite a few of the >>>> csched_ prefixes in credit1... >>> Just one other thought: unique names would help for analyzing crash dumps, >>> too. >> Only if the crash dump analyzing tool is dumb enough to also >> not properly qualify non-global names. ELF has all that is needed >> here, just the code consuming ELF symbol tables in the whole >> Linux world seems to be ignoring this capability. > I completely agree with you. > > OTOH this does not help really very much for analyzing dumps, using cscope or > interpreting backtraces of the hypervisor. Unless the tools are changed to > include the file names (including paths, please!). > > Changing the hypervisor accordingly is in our hands. > > Using cscope (or similar tools) is possible with duplicate names, even if this > isn't optimal IMO. > > Changing the dump analyzing tool(s) might be possible, but requires > potentially a longish discussion with the maintainer(s)... > > Up to the point when all of this is done, changing the names to be unique seems > to be a sensible way to make life easier. > > > Juergen > And for my crashdump analyser, all I have to go on is /boot/xen-$VERSION.map, which contains no de-duplication of similarly-named symbols. FWIW, I agree with the original patch, as I have fallen over exactly this issue with the credit scheduler before. ~Andrew