From mboxrd@z Thu Jan 1 00:00:00 1970 From: Arjan van de Ven Subject: Re: [PATCH] async: Don't call async_synchronize_full_special() while holding sb_lock Date: Mon, 12 Jan 2009 03:54:41 +0000 Message-ID: <496ABF01.50001@linux.intel.com> References: <1231425472.21528.13.camel@norville.austin.ibm.com> <20090108072111.1ebadebd@infradead.org> <1231429591.27353.14.camel@norville.austin.ibm.com> <20090108225050.GL9448@disturbed> <49668388.708@linux.intel.com> <20090109014054.GN9448@disturbed> <4966D652.4070105@linux.intel.com> <20090112023138.GD6428@shareable.org> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Cc: Dave Kleikamp , Linus Torvalds , Grissiom , linux-kernel@vger.kernel.org, linux-fsdevel To: Jamie Lokier Return-path: Received: from mga10.intel.com ([192.55.52.92]:52867 "EHLO fmsmga102.fm.intel.com" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP id S1752151AbZALDyx (ORCPT ); Sun, 11 Jan 2009 22:54:53 -0500 In-Reply-To: <20090112023138.GD6428@shareable.org> Sender: linux-fsdevel-owner@vger.kernel.org List-ID: Jamie Lokier wrote: > Arjan van de Ven wrote: >>> - removing a million files and queuing all of the >>> deletes in the async queues.... >> the async code throttles at 32k outstanding. >> Yes 32K is arbitrary, but if you delete a million files fast, all but the >> first few thousand are >> synchronous. > > Hmm. > > If I call unlink() a thousand times and then call fsync() on the > parent directories covering files I've unlinked... I expect the > deletes to be committed to disk when the last fsync() has returned. I > require that a crash and restart will not see the files. Several > kinds of transactional software and even some shell scripts expect this. > > Will these asynchronous deletes break the guaranteed > commit-of-the-delete provided by fsync() on the parent directory? 3 things: 1) removing the name from the directory and removing the data from disk are independent things. The former happens from unlink(), the later happens when the refcount hits 0 (eg no more openers nor any directory on disk referencing it). fsync() on a parent dir obviously only covers the first part, while only the 2nd part was made asynchronous. 2) with the right synchronization point in fsync, it will still work out 3) this code will be redone for 2.6.30; for 2.6.29 it is removed.