From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id CAA2FC433F5 for ; Tue, 24 May 2022 23:17:19 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S238044AbiEXXRR (ORCPT ); Tue, 24 May 2022 19:17:17 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:40198 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S237123AbiEXXRR (ORCPT ); Tue, 24 May 2022 19:17:17 -0400 Received: from mail104.syd.optusnet.com.au (mail104.syd.optusnet.com.au [211.29.132.246]) by lindbergh.monkeyblade.net (Postfix) with ESMTP id 8E7103ED04 for ; Tue, 24 May 2022 16:17:15 -0700 (PDT) Received: from dread.disaster.area (pa49-181-2-147.pa.nsw.optusnet.com.au [49.181.2.147]) by mail104.syd.optusnet.com.au (Postfix) with ESMTPS id 9BA735347B5; Wed, 25 May 2022 09:17:14 +1000 (AEST) Received: from dave by dread.disaster.area with local (Exim 4.92.3) (envelope-from ) id 1ntdm5-00G0bF-Hr; Wed, 25 May 2022 09:17:13 +1000 Date: Wed, 25 May 2022 09:17:13 +1000 From: Dave Chinner To: Amir Goldstein Cc: linux-xfs , Chris Dunlop Subject: Re: [PATCH 2/2] xfs: introduce xfs_inodegc_push() Message-ID: <20220524231713.GD1098723@dread.disaster.area> References: <20220524063802.1938505-1-david@fromorbit.com> <20220524063802.1938505-3-david@fromorbit.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: X-Optus-CM-Score: 0 X-Optus-CM-Analysis: v=2.4 cv=deDjYVbe c=1 sm=1 tr=0 ts=628d677a a=ivVLWpVy4j68lT4lJFbQgw==:117 a=ivVLWpVy4j68lT4lJFbQgw==:17 a=kj9zAlcOel0A:10 a=oZkIemNP1mAA:10 a=7-415B0cAAAA:8 a=20KFwNOVAAAA:8 a=N8xf_ed3AAAA:8 a=93vLQDYR59KmNjmNFe0A:9 a=CjuIK1q_8ugA:10 a=biEYGPWJfzWAr4FL6Ov7:22 a=sE4t997d3Q9FUvws1cBB:22 Precedence: bulk List-ID: X-Mailing-List: linux-xfs@vger.kernel.org On Tue, May 24, 2022 at 01:47:36PM +0300, Amir Goldstein wrote: > On Tue, May 24, 2022 at 1:37 PM Dave Chinner wrote: > > > > From: Dave Chinner > > > > The current blocking mechanism for pushing the inodegc queue out to > > disk can result in systems becoming unusable when there is a long > > running inodegc operation. This is because the statfs() > > implementation currently issues a blocking flush of the inodegc > > queue and a significant number of common system utilities will call > > statfs() to discover something about the underlying filesystem. > > > > This can result in userspace operations getting stuck on inodegc > > progress, and when trying to remove a heavily reflinked file on slow > > storage with a full journal, this can result in delays measuring in > > hours. > > > > Avoid this problem by adding "push" function that expedites the > > flushing of the inodegc queue, but doesn't wait for it to complete. > > > > Convert xfs_fs_statfs() to use this mechanism so it doesn't block > > but it does ensure that queued operations are expedited. > > > > Fixes: ab23a7768739 ("xfs: per-cpu deferred inode inactivation queues") > > Reported-by: Chris Dunlop > > Signed-off-by: Dave Chinner > > --- > > fs/xfs/xfs_icache.c | 20 +++++++++++++++----- > > fs/xfs/xfs_icache.h | 1 + > > fs/xfs/xfs_super.c | 7 +++++-- > > fs/xfs/xfs_trace.h | 1 + > > 4 files changed, 22 insertions(+), 7 deletions(-) > > > > diff --git a/fs/xfs/xfs_icache.c b/fs/xfs/xfs_icache.c > > index 786702273621..2609825d53ee 100644 > > --- a/fs/xfs/xfs_icache.c > > +++ b/fs/xfs/xfs_icache.c > > @@ -1862,19 +1862,29 @@ xfs_inodegc_worker( > > } > > > > /* > > - * Force all currently queued inode inactivation work to run immediately and > > - * wait for the work to finish. > > + * Expedite all pending inodegc work to run immediately. This does not wait for > > + * completion of the work. > > */ > > void > > -xfs_inodegc_flush( > > +xfs_inodegc_push( > > struct xfs_mount *mp) > > { > > if (!xfs_is_inodegc_enabled(mp)) > > return; > > + trace_xfs_inodegc_push(mp, __return_address); > > + xfs_inodegc_queue_all(mp); > > +} > > > > +/* > > + * Force all currently queued inode inactivation work to run immediately and > > + * wait for the work to finish. > > + */ > > +void > > +xfs_inodegc_flush( > > + struct xfs_mount *mp) > > +{ > > + xfs_inodegc_push(mp); > > trace_xfs_inodegc_flush(mp, __return_address); > > Unintentional(?) change of behavior: > trace_xfs_inodegc_flush() will be called in > (!xfs_is_inodegc_enabled(mp)) case. Intentional. inodegc can be disabled/enabled at runtime (e.g. freeze). This behaviour is to allow asynchronous push to minimise wait time when needing to expedite inodegc and wait for it to complete. P1 P2 ---- ----- xfs_inodegc_push() xfs_inodegc_stop() ..... inodegc disabled xfs_inodegc_queue_all() flush_workqueue() ..... ..... xfs_inodegc_flush() xfs_inodegc_push() (does nothing) flush_workqueue() ..... ..... > I also wonder if trace_xfs_inodegc_flush() > should not be before trace_xfs_inodegc_push() in this flow, > but this is just a matter of tracing conventions and you should > know best how it will be convenient for xfs developers to be > reading the trace events stream. push and flush have their own tracepoints and so we can easily see how flush has interacted with enable/disable - push+flush = xfs_inodegc_flush when enabled, flush = xfs_inodegc_flush when disabled, push = xfs_inodegc_push. Cheers, Dave. -- Dave Chinner david@fromorbit.com