From mboxrd@z Thu Jan 1 00:00:00 1970 From: Dave Hansen Subject: Re: [RFC][PATCH 4/5] must hold lock_super() to set initial mount writer Date: Thu, 01 May 2008 09:26:44 -0700 Message-ID: <1209659204.4461.65.camel@nimitz.home.sr71.net> References: <20080429185943.3BA6A050@kernel> <20080429185948.F5D2E512@kernel> Mime-Version: 1.0 Content-Type: text/plain Content-Transfer-Encoding: 7bit Cc: linux-kernel@vger.kernel.org, linux-fsdevel@vger.kernel.org, viro@zeniv.linux.org.uk, hch@infradead.org, trond.myklebust@fys.uio.no To: Miklos Szeredi Return-path: Received: from e31.co.us.ibm.com ([32.97.110.149]:37526 "EHLO e31.co.us.ibm.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753109AbYEAQ0t (ORCPT ); Thu, 1 May 2008 12:26:49 -0400 In-Reply-To: Sender: linux-fsdevel-owner@vger.kernel.org List-ID: On Wed, 2008-04-30 at 11:25 +0200, Miklos Szeredi wrote: > > We need lock_mnt_writers() during a remount in order > > to keep mnt->__mnt_writers from changing so that we > > get a consistent look at if a sb currently has anyone > > writing to it. > > > > But, we need to lock writers out for an extended > > period, even during the ->remount_fs() operation. > > That's because we do conclusively make the fs > > r/o until *after* the ->remount_fs(). > > So? Why don't we mark the fs r/o _before_ calling ->remount_fs() and > if that fails, just mark it r/w again. > > OK, we'll deny writes in that interval, but I don't see that as a big > problem. And it would simplify the implementation considerably. Personally, I think that's a bit messy. People might start getting -EROFS when they never, ever *HAD* a r/o FS. They may have made a request for one, but they never actually hard one. I also understand what you're saying. If we were able to loosen up some of the requirements it would certainly make the patches simpler. How does everyone else feel about this? -- Dave