On Wed, Sep 09 2009, Wu Fengguang wrote: > On Wed, Sep 09, 2009 at 08:28:06PM +0800, Christoph Hellwig wrote: > > On Wed, Sep 09, 2009 at 05:29:01PM +0800, Wu Fengguang wrote: > > > It seems that balance_dirty_pages() is not coupled with MAX_WRITEBACK_PAGES. > > > Instead it uses the much smaller (ratelimit_pages + ratelimit_pages / 2). > > > > With Jen's writeback patches applied balance_dirty_pages does not start > > writeback itself anymore but calls bdi_start_writeback to let the > > flusher thread do it. > > > > it would be good if we do any writeback tuning ontop of these patches.. > > Ah OK. I'm using latest linux-next and expected his patches to be there.. They are there, have been for months! But I think Christoph is a little confused, we'll still do writeback inline from balance_dirty_pages(). It does writeback_inodes_wbc(), which does not schedule async writeout. So if your patches are based and tested off -next, you should be good. -- Jens Axboe -- To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
