From: Jan Kara <jack@suse.cz>
Subject: Re: [PATCH 3/5 resend] VFS: Fix s_umount thaw/write deadlock
Date: Thu, 8 Dec 2011 01:05:19 +0100
Message-ID: <20111208000519.GW4622@quack.suse.cz>
References: <1323118489-16326-1-git-send-email-kamal@canonical.com>
 <1323118489-16326-4-git-send-email-kamal@canonical.com>
 <20111206113544.GA21589@infradead.org>
 <20111207231658.GQ4622@quack.suse.cz>
 <20111207234919.GW4387@parisc-linux.org>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Cc: Jan Kara <jack@suse.cz>, Christoph Hellwig <hch@infradead.org>,
	Kamal Mostafa <kamal@canonical.com>,
	Alexander Viro <viro@zeniv.linux.org.uk>,
	Andreas Dilger <adilger.kernel@dilger.ca>,
	Randy Dunlap <rdunlap@xenotime.net>,
	Theodore Tso <tytso@mit.edu>, linux-doc@vger.kernel.org,
	linux-ext4@vger.kernel.org, linux-fsdevel@vger.kernel.org,
	linux-kernel@vger.kernel.org, Surbhi Palande <csurbhi@gmail.com>,
	Valerie Aurora <val@vaaconsulting.com>,
	Christopher Chaltain <christopher.chaltain@canonical.com>,
	"Peter M. Petrakis" <peter.petrakis@canonical.com>,
	Mikulas Patocka <mpatocka@redhat.com>,
	Miao Xie <miaox@cn.fujitsu.com>
To: Matthew Wilcox <matthew@wil.cx>
Return-path: <linux-doc-owner@vger.kernel.org>
Content-Disposition: inline
In-Reply-To: <20111207234919.GW4387@parisc-linux.org>
Sender: linux-doc-owner@vger.kernel.org
List-Id: linux-ext4.vger.kernel.org

On Wed 07-12-11 16:49:20, Matthew Wilcox wrote:
> On Thu, Dec 08, 2011 at 12:16:58AM +0100, Jan Kara wrote:
> > On Tue 06-12-11 06:35:44, Christoph Hellwig wrote:
> > > We make sure to not dirty any new inodes after the first phase of the
> > > freeze, so this should be a BUG_ON/WARN_ON.
> >   This is not really true in presence of mmaped writes. To block mmaped
> > writes on a frozen filesystem, we need some synchronization between
> > page_mkwrite() and freezing code. Currently, to avoid any additional
> > locking overhead, we set page dirty and *then* check for filesystem being
> > frozen. Only this order can make sure either the page is written (and
> > write-protected) or the frozen check triggers and we wait... (see the
> > comment in block_page_mkwrite()). The nasty sideeffect of this is that
> > there can be dirty pages & inodes on a frozen filesystem. We are blocked in
> > the page fault of these pages so user cannot write any data to these pages
> > but still they are marked dirty.
> 
> Can't we make the mmaps read-only during the freezing process?
  But then application will get EFAULT instead of blocking? I don't think
that is acceptable (we'd see lots of applications dying ;).

								Honza
-- 
Jan Kara <jack@suse.cz>
SUSE Labs, CR