On Thu 04-08-11 06:42:10, Christoph Hellwig wrote:
> On Thu, Aug 04, 2011 at 12:36:16PM +0200, Jan Kara wrote:
> > > The first one actually is a synchronous writeout, just implemented in
> > > a rather odd way by doing the xfs_ioend_wait right after it, so your
> > > change is actively harmful for it.
> > Oh, right. BTW cannot be truncate livelocked on a busy file because of
> > that xfs_ioend_wait()?
> Not really. We requite the iolock for new writes to start, and truncate
> holds it exclusively. But I'm working on a series for 3.2 to remove
> xfs_ioend_wait and just rely on inode_dio_wait for direct I/O, so it
> will be gone soon. At this point I'll also have to switch to
> filemap_write_and_wait_range for this caller.
> > > The third one is opportunistic writeout if a file got truncated down on
> > > final release. filemap_flush probably is fine here, but there's no need
> > > for a range version. If you replace it with filemap_flush please also
> > > kill the useless wrapper while you're at it.
> > Do you mean xfs_flush_pages()? OK, I can do that.
> Yes, xfs_flush_pages should go - at least he async version and its
> abuse of the buffer flags.
Hmm, BTW, shouldn't the call to xfs_flush_pages() in
xfs_file_buffered_aio_write() be converted to an asynchronous one? I don't
quite see a point in waiting for io completion... Generally, flushing of
the inode there seems of limited usefulness to me since that inode could be
just a tiny victim not holding much delayallocated blocks.
Jan Kara <jack@xxxxxxx>
SUSE Labs, CR