From mboxrd@z Thu Jan 1 00:00:00 1970 From: Eric Sandeen Subject: Re: [PATCH] mke2fs: enable bigalloc if -C is specified Date: Thu, 31 May 2012 15:44:42 -0500 Message-ID: <4FC7D83A.40003@redhat.com> References: <4FC7A1D5.3040805@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit To: ext4 development Return-path: Received: from mx1.redhat.com ([209.132.183.28]:28194 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1756142Ab2EaUoo (ORCPT ); Thu, 31 May 2012 16:44:44 -0400 Received: from int-mx12.intmail.prod.int.phx2.redhat.com (int-mx12.intmail.prod.int.phx2.redhat.com [10.5.11.25]) by mx1.redhat.com (8.14.4/8.14.4) with ESMTP id q4VKihLb022369 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=OK) for ; Thu, 31 May 2012 16:44:43 -0400 Received: from liberator.sandeen.net (ovpn01.gateway.prod.ext.phx2.redhat.com [10.5.9.1]) by int-mx12.intmail.prod.int.phx2.redhat.com (8.14.4/8.14.4) with ESMTP id q4VKigT3016569 (version=TLSv1/SSLv3 cipher=DHE-RSA-CAMELLIA256-SHA bits=256 verify=NO) for ; Thu, 31 May 2012 16:44:43 -0400 In-Reply-To: <4FC7A1D5.3040805@redhat.com> Sender: linux-ext4-owner@vger.kernel.org List-ID: On 5/31/12 11:52 AM, Eric Sandeen wrote: > If -C is specified w/o -O bigalloc it has no effect and generates > no error. If a cluster size is specified, that should imply > bigalloc. Hm, so should -O bigalloc even be supported (or documented), or should this always be done via -C XXXX? It seems better to specify the size rather than have some other -O option which picks an (arbitrary?) default of 16x. I'm also wondering what kind of guidance we should offer for choosing cluster sizes - or if we should default to a cluster size given either fs size, inode count, or combinations thereof. I think the hard cold truth is that ext4 just isn't sufficiently scalable at larger sizes without a larger cluster size, so I'm inclined to start thinking about choosing some increasing cluster sizes as defaults, what do you think? And finally, is this stuff robust enough to start documenting in the manpages yet? I'm not sure the enospc problems have been worked out yet... -Eric > Signed-off-by: Eric Sandeen > --- > > diff --git a/misc/mke2fs.c b/misc/mke2fs.c > index 7ec8cc2..d1944dc 100644 > --- a/misc/mke2fs.c > +++ b/misc/mke2fs.c > @@ -1351,6 +1351,8 @@ profile_error: > optarg); > exit(1); > } > + fs_param.s_feature_ro_compat |= > + EXT4_FEATURE_RO_COMPAT_BIGALLOC; > break; > case 'D': > direct_io = 1; > >