From: Jan Kara Subject: Re: Ordered mode rewrite patch Date: Wed, 9 Apr 2008 16:48:31 +0200 Message-ID: <20080409144831.GB7789@duck.suse.cz> References: <20080408121149.GB6901@duck.suse.cz> <20080408121937.GA26650@infradead.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Nathan Grennan , Marcelo Tosatti , linux-ext4@vger.kernel.org To: Christoph Hellwig Return-path: Received: from styx.suse.cz ([82.119.242.94]:56991 "EHLO mail.suse.cz" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP id S1750732AbYDIOsc (ORCPT ); Wed, 9 Apr 2008 10:48:32 -0400 Content-Disposition: inline In-Reply-To: <20080408121937.GA26650@infradead.org> Sender: linux-ext4-owner@vger.kernel.org List-ID: On Tue 08-04-08 08:19:37, Christoph Hellwig wrote: > On Tue, Apr 08, 2008 at 02:11:49PM +0200, Jan Kara wrote: > > - } else if (!buffer_mapped(bh) && buffer_dirty(bh)) { > > + } else if (!buffer_mapped(bh) && buffer_dirty(bh) > > + && !wbc->skip_unmapped) { > > As mentioned last time unmapped buffer shouldn't happen anymore with > filesystem updated to have a ->page_mkwrite operations. As you've > included the ext3 patch in yours the above condition shoujldn't git > anymore. True. I've used wbc->skip_unmapped to detect inside ext3_ordered_writepage() that we are committing a transaction and therefore we should not add inode into the running transaction's list. But later I've realized that we never need to add inode into transaction's list from ext3_ordered_writepage() once we have ext3_page_mkwrite(). So now I don't need wbc->skip_unmapped at all and I've compeletely removed it. Thanks for pointing this out. Honza -- Jan Kara SUSE Labs, CR