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:   Thu, 10 Nov 2022 12:16:36 +0100
From:   Jan Kara <jack@suse.cz>
To:     Yu Kuai <yukuai1@huaweicloud.com>
Cc:     Gabriel Krisman Bertazi <krisman@suse.de>, axboe@kernel.dk,
        linux-kernel@vger.kernel.org, linux-block@vger.kernel.org,
        Hugh Dickins <hughd@google.com>,
        Keith Busch <kbusch@kernel.org>,
        Liu Song <liusong@linux.alibaba.com>, Jan Kara <jack@suse.cz>,
        "yukuai (C)" <yukuai3@huawei.com>
Subject: Re: [PATCH] sbitmap: Use single per-bitmap counting to wake up
 queued tags
Message-ID: <20221110111636.ufgyp4tkbzexugk2@quack3>
References: <20221105231055.25953-1-krisman@suse.de>
 <2a445c5c-fd15-c0bf-8655-2fb5bde3fe67@huaweicloud.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: <2a445c5c-fd15-c0bf-8655-2fb5bde3fe67@huaweicloud.com>
Precedence: bulk

Hi!

On Thu 10-11-22 17:42:49, Yu Kuai wrote:
> 在 2022/11/06 7:10, Gabriel Krisman Bertazi 写道:
> > +void sbitmap_queue_wake_up(struct sbitmap_queue *sbq, int nr)
> >   {
> > -	struct sbq_wait_state *ws;
> > -	unsigned int wake_batch;
> > -	int wait_cnt, cur, sub;
> > -	bool ret;
> > +	unsigned int wake_batch = READ_ONCE(sbq->wake_batch);
> > +	struct sbq_wait_state *ws = NULL;
> > +	unsigned int wakeups;
> > -	if (*nr <= 0)
> > -		return false;
> > +	if (!atomic_read(&sbq->ws_active))
> > +		return;
> > -	ws = sbq_wake_ptr(sbq);
> > -	if (!ws)
> > -		return false;
> > +	atomic_add(nr, &sbq->completion_cnt);
> > +	wakeups = atomic_read(&sbq->wakeup_cnt);
> > -	cur = atomic_read(&ws->wait_cnt);
> >   	do {
> > -		/*
> > -		 * For concurrent callers of this, callers should call this
> > -		 * function again to wakeup a new batch on a different 'ws'.
> > -		 */
> > -		if (cur == 0)
> > -			return true;
> > -		sub = min(*nr, cur);
> > -		wait_cnt = cur - sub;
> > -	} while (!atomic_try_cmpxchg(&ws->wait_cnt, &cur, wait_cnt));
> > -
> > -	/*
> > -	 * If we decremented queue without waiters, retry to avoid lost
> > -	 * wakeups.
> > -	 */
> > -	if (wait_cnt > 0)
> > -		return !waitqueue_active(&ws->wait);
> > +		if (atomic_read(&sbq->completion_cnt) - wakeups < wake_batch)
> > +			return;
> 
> Should it be considered that completion_cnt overflow and becomes
> negtive?

Yes, the counters can (and will) certainly overflow but since we only care
about (completion_cnt - wakeups), we should be fine - this number is always
sane (and relatively small) and in the kernel we do compile with signed
overflows being well defined.

								Honza
-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR