From mboxrd@z Thu Jan 1 00:00:00 1970 From: Theodore Tso Subject: Re: [PTCH] push down lock_super and BKL into ->put_super Date: Wed, 6 May 2009 03:34:29 -0400 Message-ID: <20090506073429.GC6976@mit.edu> References: <20090505134036.GA4127@lst.de> <20090506020916.GN8633@ZenIV.linux.org.uk> <20090506062300.GA12718@lst.de> <20090506064622.GQ8633@ZenIV.linux.org.uk> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Christoph Hellwig , linux-fsdevel@vger.kernel.org To: Al Viro Return-path: Received: from THUNK.ORG ([69.25.196.29]:44302 "EHLO thunker.thunk.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751310AbZEFHeq (ORCPT ); Wed, 6 May 2009 03:34:46 -0400 Content-Disposition: inline In-Reply-To: <20090506064622.GQ8633@ZenIV.linux.org.uk> Sender: linux-fsdevel-owner@vger.kernel.org List-ID: On Wed, May 06, 2009 at 07:46:22AM +0100, Al Viro wrote: > > FS blocking writeback may very well be, though. > > In any case, the point is that we should separate get_fs_excl() from > exclusion there and kill exclusion part for everything except ext4. Note that I already have patches queued for ext4 that removes the use of lock_super() for everything other than write_super() exclusion. (i.e., we were using it to protect oneline resize and the orphan list.) Similar patches are needed for ext3, which I'll backport before the next merge window. BTW, I'm *not* at all convinced that get_fs_excl() is the right interface for boosting I/O priority, since it only boosts priority for idle processes. So high priority processes will still get screwed by normal I/O. Hence, in the common case, where there are no I/O class nince processes, get_fs_excl() is a no-op anyway. And the case which I'm personally most interested in, which is real-time processes that want to do I/O and so have an elevated I/O priority, get_fs_excl() does nothing for them. - Ted