From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1751906Ab2GGGDM (ORCPT ); Sat, 7 Jul 2012 02:03:12 -0400 Received: from mail.parknet.co.jp ([210.171.160.6]:42296 "EHLO mail.parknet.co.jp" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751360Ab2GGGDK (ORCPT ); Sat, 7 Jul 2012 02:03:10 -0400 From: OGAWA Hirofumi To: "Steven J. Magnani" Cc: linux-kernel@vger.kernel.org Subject: Re: [PATCH 2/2] fat (exportfs): reconnect file handles to evicted inodes/dentries References: <1341342576-15394-1-git-send-email-steve@digidescorp.com> <1341342576-15394-3-git-send-email-steve@digidescorp.com> <87pq8bokcp.fsf@devron.myhome.or.jp> <1341606824.2214.14.camel@iscandar.digidescorp.com> <871ukobntl.fsf@devron.myhome.or.jp> <1341623782.2021.9.camel@iscandar> Date: Sat, 07 Jul 2012 15:03:04 +0900 In-Reply-To: <1341623782.2021.9.camel@iscandar> (Steven J. Magnani's message of "Fri, 06 Jul 2012 20:16:22 -0500") Message-ID: <87wr2g9kh3.fsf@devron.myhome.or.jp> User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/24.1.50 (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 "Steven J. Magnani" writes: > On Sat, 2012-07-07 at 06:07 +0900, OGAWA Hirofumi wrote: >> "Steven J. Magnani" writes: >> >> > On Wed, 2012-07-04 at 20:07 +0900, OGAWA Hirofumi wrote: >> >> Please don't add new lock_super() usage if it is not necessary. Almost >> >> all of lock_super() just replaced lock_kernel() usage. It rather should >> >> be removed in future. Probably, this should use inode->i_mutex instead. >> >> >> >> BTW, the above issue is same with all of directory read. >> > >> > I don't think there's really an alternative here. The cases addressed by >> > this patch all involve walking on-disk structures via >> > unofficial/temporary inodes created from information in the NFS handle >> > (i.e., outside the normal inode creation paths). When this process is >> > successful we end up with "official" connected inodes/dentries, but >> > getting there is really a "bottom up" strategy instead of the usual "top >> > down" approach. >> > >> > Because the "bottom up" method is lacking guarantees that "top down" >> > takes for granted - i.e., that a cluster on disk that's supposed to be a >> > directory actually *is* a directory - I am adding some defensive code >> > in the next spin of the patch. >> >> I'm not sure what you meant. Where is the problem? ->get_name()? If so, >> it has parent dentry parameter. What is the wrong if we take >> mutex_lock(parent->d_inode)? > > The temporary/"unofficial" inodes are unhashed and thus not visible > outside of the functions using them. They exist only to support access > to directory contents when we can't gain that access via other means > (because we only have "bottom up" information, about an object and > perhaps its parent, in a form that can't be used to look up hashed > inodes/dentries). Hashing them wouldn't help, because they don't have > the correct key (for instance, in the case of a ".." entry). > > Am I missing something? You mean the unhashed inode is created by ->get_parent()? If so, the root cause sounds like ->get_parent() itself. If not, I'm not understanding the meaning of the temporary/unofficial inode here. -- OGAWA Hirofumi