From mboxrd@z Thu Jan 1 00:00:00 1970 From: Dave Chinner Subject: Re: [PATCH] vfs: make inode_sb_list_lock per sb Date: Sat, 20 Dec 2014 10:16:37 +1100 Message-ID: <20141219231637.GG15665@dastard> References: <1419018033-26704-1-git-send-email-jbacik@fb.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: viro@ZenIV.linux.org.uk, linux-fsdevel@vger.kernel.org To: Josef Bacik Return-path: Received: from ipmail04.adl6.internode.on.net ([150.101.137.141]:17618 "EHLO ipmail04.adl6.internode.on.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752604AbaLSXQk (ORCPT ); Fri, 19 Dec 2014 18:16:40 -0500 Content-Disposition: inline In-Reply-To: <1419018033-26704-1-git-send-email-jbacik@fb.com> Sender: linux-fsdevel-owner@vger.kernel.org List-ID: On Fri, Dec 19, 2014 at 02:40:33PM -0500, Josef Bacik wrote: > When doing fs_mark tests I was noticing other things getting starved out of > doing operations while the fs was unmounting. This is because we protect all > super_block's s_inodes list with a global lock, which is kind of a bummer. > There doesn't seem to be any reason we do this so make it a per-sb lock. This > makes sure that we don't add latency to anybody trying to add/remove inodes from > the per sb list while somebody else is unmounting or evicting inodes. Thanks, > > Signed-off-by: Josef Bacik I'll have a look at this on Monday, but in the mean time I proposed this 18 months ago for different reasons, and never got back to it: https://lkml.org/lkml/2013/7/31/15 Maybe you want to pick up someof the other sb inode list scalability mods out of that series as well? Cheers, Dave. -- Dave Chinner david@fromorbit.com