Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Wed, 22 Jun 2022 18:15:37 -0700
From:   "Darrick J. Wong" <djwong@kernel.org>
To:     Stephen Zhang <starzhangzsd@gmail.com>
Cc:     dchinner@redhat.com, zhangshida <zhangshida@kylinos.cn>,
        linux-kernel@vger.kernel.org, linux-xfs@vger.kernel.org
Subject: Re: [PATCH v3] xfs: add check before calling xfs_mod_fdblocks
Message-ID: <YrO+uap429yrLKD6@magnolia>
References: <20220621084238.1235880-1-zhangshida@kylinos.cn>
 <YrKdZc0fBBDWGjld@magnolia>
 <CANubcdXyM2t0M0Ks3=6Hm12_SVm50vGaTBboj5CBF2LygOkUog@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <CANubcdXyM2t0M0Ks3=6Hm12_SVm50vGaTBboj5CBF2LygOkUog@mail.gmail.com>
Precedence: bulk

On Wed, Jun 22, 2022 at 01:57:31PM +0800, Stephen Zhang wrote:
> > I understand that calling __xfs_ag_resv_init on an AG with a maximally
> > sized data structure can result in @hidden_space being zero here, but
> > why does that matter enough to change the code?  Are you experiencing
> > problems when this happens?  Unnecessary slowdowns at mount time?
> > Something else?
> >
> > This is v3 of a patch and I still can't tell why I should care ...?
> 
> After applying this patch, we can avoid to call xfs_mod_fdblocks when
> delta equals 0. So we can reduce unnecessary operations here.

Yeah, I get that, but what is the real world impact of those unnecessary
operations?  Have you run fstests to make sure this change doesn't trip
over some weird subtlety in the code?  Do the anticipated benefits
justify diverting my time to figuring out if we've really covered all
the corner cases?

IOWS: don't waste our time on theoretical improvements.  There are
/plenty/ of things in 5.19 that need real attention, like generic/522
corrupting things and recoveryloop tests that trip over log recovery.

--D