Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Wed, 20 Nov 2019 22:56:44 -0800
From:   Christoph Hellwig <hch@infradead.org>
To:     Dave Chinner <david@fromorbit.com>
Cc:     Shaokun Zhang <zhangshaokun@hisilicon.com>,
        linux-xfs@vger.kernel.org, linux-kernel@vger.kernel.org,
        Yang Guo <guoyang2@huawei.com>,
        "Darrick J. Wong" <darrick.wong@oracle.com>
Subject: Re: [PATCH] xfs: optimise xfs_mod_icount/ifree when delta < 0
Message-ID: <20191121065644.GA21798@infradead.org>
References: <1572866980-13001-1-git-send-email-zhangshaokun@hisilicon.com>
 <20191104204909.GB4614@dread.disaster.area>
 <dc7456d6-616d-78c5-0ac6-c5ffaf721e41@hisilicon.com>
 <20191105040325.GC4614@dread.disaster.area>
 <675693c2-8600-1cbd-ce50-5696c45c6cd9@hisilicon.com>
 <20191106212041.GF4614@dread.disaster.area>
 <d627883a-850c-1ec4-e057-cf9e9b47c50e@hisilicon.com>
 <20191118081212.GT4614@dread.disaster.area>
 <20191120210825.GB4614@dread.disaster.area>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20191120210825.GB4614@dread.disaster.area>
User-Agent: Mutt/1.12.1 (2019-06-15)
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

> +static void
>  xfs_sb_mod8(
>  	uint8_t			*field,
>  	int8_t			delta)
>  {
>  	int8_t			counter = *field;
>  
> +	if (!delta)
> +		return;
>  	counter += delta;
> +	ASSERT(counter >= 0);
>  	*field = counter;
>  }

I'd almost find it easier to keep the != 0 check in the caller, and
in fact also move the assert there and just kill these helpers, e.g.

	if (tp->t_imaxpct_delta != 0) {
		mp->m_sb.sb_imax_pct += tp->t_imaxpct_delta;
		ASSERT(mp->m_sb.sb_imax_pct >= 0);
	}
	if (tp->t_rextsize_delta != 0) {
		mp->m_sb.sb_rextsize += tp->t_rextsize_delta;
		ASSERT(mp->m_sb.sb_rextsize >= 0);
	}

While this is 2 more lines per counter it is a lot more explicit and
easier to follow.

>  	if (idelta) {
> -		error = xfs_mod_icount(mp, idelta);
> -		if (error)
> -			goto out_undo_fdblocks;
> +		percpu_counter_add_batch(&mp->m_icount, idelta,
> +					 XFS_ICOUNT_BATCH);
> +		if (idelta < 0)
> +			ASSERT(__percpu_counter_compare(&mp->m_icount, 0,
> +							XFS_ICOUNT_BATCH) >= 0)

And here I wonder if keeping single use helpers wouldn't have been a
little nicer.  Not that it really matters.