Re: [PATCH] direct I/O fallback sync simplification

[Christoph Hellwig <hch@lst.de>]

On Wed, Sep 23, 2009 at 03:04:30PM +0100, Jamie Lokier wrote:
> Christoph Hellwig wrote:
> > In the case of direct I/O falling back to buffered I/O we sync data
> > twice currently: once at the end of generic_file_buffered_write using
> > filemap_write_and_wait_range and once a little later in
> > __generic_file_aio_write using do_sync_mapping_range with all flags set.
> > 
> > The wait before write of the do_sync_mapping_range call does not make
> > any sense, so just keep the filemap_write_and_wait_range call and move
> > it to the right spot.
> 
> Are you sure this is an expectation of O_DIRECT?
> 
> A few notes from the net, including some documentation from IBM,
> advise using O_DIRECT|O_DSYNC if you need sync when direct I/O falls
> back to buffered on some other OSes.

Not sure if we're on the same page here.  Before the patch we do the
following in case of a fallback buffered write when opened with
O_DIRECT:

 - a filemap_write_and_wait_range in generic_file_buffered_write for
   (with a too long length, btw)
 - a do_sync_mapping_range(wait,write,wait) from
   __generic_file_aio_write

which expands to

 (1) __filemap_fdatawrite_range(SYNC_ALL)
 (2) wait_on_page_writeback_range
 (3) wait_on_page_writeback_range
 (4) __filemap_fdatawrite_range(SYNC_ALL)
 (5) wait_on_page_writeback_range

which clearly doesn't make any sense.  The patch reduces it to
 
 (1) __filemap_fdatawrite_range(SYNC_ALL)
 (2) wait_on_page_writeback_range

which does exactly the same, that is assuring the _data_ is on disk.
That's still not O_DSNC or O_SYNC semantics, but that's an entirely
different discussion.