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, 11 Mar 2020 17:59:19 -0700
From:   Andrew Morton <akpm@linux-foundation.org>
To:     Jules Irenge <jbi.octave@gmail.com>
Cc:     boqun.feng@gmail.com, linux-mm@kvack.org,
        linux-kernel@vger.kernel.org
Subject: Re: [PATCH 1/1] backing-dev: refactor wb_congested_put()
Message-Id: <20200311175919.30523d55b2e5307ba22bbdc0@linux-foundation.org>
In-Reply-To: <20200312002156.49023-2-jbi.octave@gmail.com>
References: <20200312002156.49023-1-jbi.octave@gmail.com>
        <20200312002156.49023-2-jbi.octave@gmail.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Thu, 12 Mar 2020 00:21:56 +0000 Jules Irenge <jbi.octave@gmail.com> wrote:

> wb_congested_put() was written in such a way that made it difficult
> 	for Sparse tool not to complain
> Expanding the function locking block in the if statement improves on
> the readability of the code. Rewritting it  comes with one add-on:
> 
> It fixes a warning reported by Sparse tool at wb_congested_put()
> 
> warning: context imbalance in wb_congested_put() - unexpected unlock
> 
> Refactor the function wb_congested_put()
> 
> ...
>
> --- a/mm/backing-dev.c
> +++ b/mm/backing-dev.c
> @@ -464,18 +464,21 @@ void wb_congested_put(struct bdi_writeback_congested *congested)
>  {
>  	unsigned long flags;
>  
> -	if (!refcount_dec_and_lock_irqsave(&congested->refcnt, &cgwb_lock, &flags))
> -		return;
> -
> +	if (!refcount_dec_not_one(&congested->refcnt)) {
> +		spin_lock_irqsave(&cgwb_lock, flags);
> +		if (!refcount_dec_and_test(&congested->refcnt)) {
> +			spin_unlock_irqrestore(&cgwb_lock, flags);
> +			return;
> +		}
>  	/* bdi might already have been destroyed leaving @congested unlinked */
> -	if (congested->__bdi) {
> -		rb_erase(&congested->rb_node,
> -			 &congested->__bdi->cgwb_congested_tree);
> -		congested->__bdi = NULL;
> +		if (congested->__bdi) {
> +			rb_erase(&congested->rb_node,
> +				 &congested->__bdi->cgwb_congested_tree);
> +			congested->__bdi = NULL;
> +		}
> +		spin_unlock_irqrestore(&cgwb_lock, flags);
> +		kfree(congested);
>  	}
> -
> -	spin_unlock_irqrestore(&cgwb_lock, flags);
> -	kfree(congested);
>  }

hm, it's hard to get excited over this.  Open-coding the
refcount_dec_and_lock_irqsave() internals at a callsite in order to
make sparse happy.

Is there some other way, using __acquires (for example)?