From mboxrd@z Thu Jan 1 00:00:00 1970 From: ebiederm@xmission.com (Eric W. Biederman) Subject: Re: [PATCH review 23/85] ocfs2: Convert uid and gids between in core and on disk inodes Date: Thu, 14 Feb 2013 01:10:01 -0800 Message-ID: <877gmbdzx2.fsf@xmission.com> References: <87621w14vs.fsf@xmission.com> <1360777934-5663-1-git-send-email-ebiederm@xmission.com> <1360777934-5663-23-git-send-email-ebiederm@xmission.com> <20130214083547.GE18408@localhost> Mime-Version: 1.0 Content-Type: text/plain Cc: Linux Containers , linux-kernel@vger.kernel.org, "Serge E. Hallyn" , Mark Fasheh To: linux-fsdevel@vger.kernel.org Return-path: Received: from out01.mta.xmission.com ([166.70.13.231]:34775 "EHLO out01.mta.xmission.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1756496Ab3BNJKK (ORCPT ); Thu, 14 Feb 2013 04:10:10 -0500 In-Reply-To: <20130214083547.GE18408@localhost> (Joel Becker's message of "Thu, 14 Feb 2013 00:35:48 -0800") Sender: linux-fsdevel-owner@vger.kernel.org List-ID: Joel Becker writes: > This has the same problem with two machines in different userspace > namespaces. Should this be punted to the sysadmin or helped out in the > filesystem? Do you define this case as unsupported? First I define two machines in the same user namespace as two machines with synchronized password files. As today connecting two machines without synchronizing the password entries is a problem, won't work, and is punted to the sysadmin to make keep from doing something silly. At the same time I don't want to make it easy to get confused, or cause breakage. With these patches ocfs2 only still only works in the initial user namespace. I don't set .fs_flags = FS_USERNS_MOUNT in struct filesystem. Which means that while users in other user namespaces can store files on ocfs2, they can't mount ocfs2. Eric