From mboxrd@z Thu Jan  1 00:00:00 1970
From: Wu Fengguang <fengguang.wu@intel.com>
Subject: Re: [PATCH 1/3] writeback: introduce wbc.for_sync to cover the two
 sync stages
Date: Mon, 2 May 2011 11:23:35 +0800
Message-ID: <20110502032335.GA13191@localhost>
References: <20110430223605.034517922@intel.com>
 <20110430224552.227301247@intel.com>
 <20110501074603.GC13542@dastard>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Cc: Andrew Morton <akpm@linux-foundation.org>, Jan Kara <jack@suse.cz>,
	Christoph Hellwig <hch@lst.de>,
	LKML <linux-kernel@vger.kernel.org>,
	"linux-fsdevel@vger.kernel.org" <linux-fsdevel@vger.kernel.org>
To: Dave Chinner <david@fromorbit.com>
Return-path: <linux-kernel-owner@vger.kernel.org>
Content-Disposition: inline
In-Reply-To: <20110501074603.GC13542@dastard>
Sender: linux-kernel-owner@vger.kernel.org
List-Id: linux-fsdevel.vger.kernel.org

On Sun, May 01, 2011 at 03:46:04PM +0800, Dave Chinner wrote:
> On Sun, May 01, 2011 at 06:36:06AM +0800, Wu Fengguang wrote:
> > sync(2) is performed in two stages: the WB_SYNC_NONE sync and the
> > WB_SYNC_ALL sync. Tag both stages with wbc.for_sync for livelock
> > prevention.
> > 
> > Note that writeback_inodes_sb() is called by not only sync(), they
> > are treated the same because the other callers need also need livelock
> > prevention.
> > 
> > Impacts:
> > 
> > - it changes the order in which pages/inodes are synced to disk. Now in
> >   the WB_SYNC_NONE stage, it won't proceed to write the next inode until
> >   finished with the current inode.
> > 
> > - this adds a new field to the writeback trace events and may possibly
> >   break some scripts.
> .....
> > --- linux-next.orig/mm/page-writeback.c	2011-05-01 06:35:16.000000000 +0800
> > +++ linux-next/mm/page-writeback.c	2011-05-01 06:35:17.000000000 +0800
> > @@ -892,12 +892,12 @@ int write_cache_pages(struct address_spa
> >  			range_whole = 1;
> >  		cycled = 1; /* ignore range_cyclic tests */
> >  	}
> > -	if (wbc->sync_mode == WB_SYNC_ALL)
> > +	if (wbc->for_sync)
> >  		tag = PAGECACHE_TAG_TOWRITE;
> >  	else
> >  		tag = PAGECACHE_TAG_DIRTY;
> >  retry:
> > -	if (wbc->sync_mode == WB_SYNC_ALL)
> > +	if (wbc->for_sync)
> >  		tag_pages_for_writeback(mapping, index, end);
> >  	done_index = index;
> >  	while (!done && (index <= end)) {
> 
> Doesn't that break anything that uses
> filemap_write_and_wait{_range}() or filemap_fdatawrite{_range}()?
> e.g. fsync, sync buffered writes, etc? i.e. everything that
> currently relies on WB_SYNC_ALL for data integrity writeback is now
> b0rken except for sync(1)?

Right, they'll become livelockable.. Good catch, thanks! I'll update
the patches to do

-   if (wbc->sync_mode == WB_SYNC_ALL)
+   if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_sync)

The alternative is to ask the other WB_SYNC_ALL callers to set
wbc.tagged_sync, but that seems more error prone.

Thanks,
Fengguang