From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754352AbbI3VeI (ORCPT ); Wed, 30 Sep 2015 17:34:08 -0400 Received: from mail-wi0-f178.google.com ([209.85.212.178]:37605 "EHLO mail-wi0-f178.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754125AbbI3VeG (ORCPT ); Wed, 30 Sep 2015 17:34:06 -0400 From: Rasmus Villemoes To: ebiederm@xmission.com (Eric W. Biederman) Cc: Oleg Nesterov , linux-kernel@vger.kernel.org, Linux Containers Subject: Re: [RFC 0/4] reduce pid_namespace footprint when not configured Organization: D03 References: <1443646163-31316-1-git-send-email-linux@rasmusvillemoes.dk> <87zj03ei64.fsf@x220.int.ebiederm.org> X-Hashcash: 1:20:150930:ebiederm@xmission.com::kM67z5XbrlGJt6zI:00000000000000000000000000000000000000000glR X-Hashcash: 1:20:150930:containers@lists.linux-foundation.org::cGrWUeDJKRu3t2j5:0000000000000000000000003J6n X-Hashcash: 1:20:150930:oleg@redhat.com::ZQA2aZbTMCvtzBwl:008wB5 X-Hashcash: 1:20:150930:linux-kernel@vger.kernel.org::HLmjTUgZvSRfznzJ:000000000000000000000000000000000AJqR Date: Wed, 30 Sep 2015 23:34:02 +0200 In-Reply-To: <87zj03ei64.fsf@x220.int.ebiederm.org> (Eric W. Biederman's message of "Wed, 30 Sep 2015 16:00:51 -0500") Message-ID: <87r3lfegmt.fsf@rasmusvillemoes.dk> User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/24.3 (gnu/linux) MIME-Version: 1.0 Content-Type: text/plain Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Wed, Sep 30 2015, ebiederm@xmission.com (Eric W. Biederman) wrote: > Rasmus Villemoes writes: > >> A few suggestions for reducing the footprint of pid namespaces when >> CONFIG_PID_NS=n. The savings are admittedly not huge: >> >> $ scripts/bloat-o-meter /tmp/vmlinux.master vmlinux >> add/remove: 0/7 grow/shrink: 30/25 up/down: 368/-1032 (-664) > > #ifdefs in C files. > More code to maintain. > Very minor savings. > > I don't think it is worth the pain. If anything I would prefer to > remove CONFIG_PID_NS. The CONFIG_PID_NS ifdef really is about keeping > from exposing people to the feature tha don't use it more than it is > about size reduction. > > Or do you have a case where this makes the difference between a usable > kernel and a kernel that is not? No, it was mostly a small tinification effort, and we'd presumably also save a few cycles on thread creation, but not anything measurable. Rasmus