From: Eric Sandeen Subject: Re: [PATCH] Add flag to files with blocks intentionally past EOF Date: Fri, 22 Jan 2010 11:40:56 -0600 Message-ID: <4B59E328.8020008@redhat.com> References: <4B5627EF.3080804@redhat.com> <87tyuhdu6q.fsf@linux.vnet.ibm.com> <22655F73-6EB0-40AC-9D51-794F6764CF67@sun.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Cc: "Aneesh Kumar K. V" , ext4 development To: Andreas Dilger Return-path: Received: from mx1.redhat.com ([209.132.183.28]:62002 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755857Ab0AVRlL (ORCPT ); Fri, 22 Jan 2010 12:41:11 -0500 In-Reply-To: <22655F73-6EB0-40AC-9D51-794F6764CF67@sun.com> Sender: linux-ext4-owner@vger.kernel.org List-ID: Andreas Dilger wrote: > On 2010-01-20, at 02:03, Aneesh Kumar K. V wrote: ... >> If we remove ext4_ioctl support i guess that patch can become much >> simpler. > > > Sure it will be simpler, but less useful. The point of exposing this > flag via lsattr is to allow userspace to determine which files are > holding blocks beyond EOF, so that if the filesystem is getting too full > it is possible to run e.g. "lsattr -R" and find files with this EOF > attribute and truncate them. Without keeping at least the EOFBLOCKS > flag in USER_VISIBLE this is impossible. Well, I submitted an updated patch without it. We can add another when we sort out what we really want with the flags, but fixing the corruption e2fsck finds seems paramount. I think the case of stray files w/ blocks past EOF is probably pretty rare - but I guess I agree, at least seeing the flags would be nice. But let's handle that separately... -Eric