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;
Message-ID: <afe5b403-4e37-80fd-643d-79e0876a7047@linux.alibaba.com>
Date:   Sat, 17 Sep 2022 17:10:29 +0800
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.15; rv:102.0)
 Gecko/20100101 Thunderbird/102.2.2
Subject: Re: [PATCH] sbitmap: fix permanent io blocking caused by insufficient
 wakeup times
To:     Hillf Danton <hdanton@sina.com>
Cc:     axboe@kernel.dk, linux-kernel@vger.kernel.org
References: <20220917035831.735-1-hdanton@sina.com>
From:   Liu Song <liusong@linux.alibaba.com>
In-Reply-To: <20220917035831.735-1-hdanton@sina.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk


On 2022/9/17 11:58, Hillf Danton wrote:
> On 17 Sep 2022 10:33:01 +0800 Liu Song <liusong@linux.alibaba.com> wrote:
>> @@ -632,10 +633,14 @@ static bool __sbq_wake_up(struct sbitmap_queue *sbq)
>>   		if (ret == wait_cnt) {
>>   			sbq_index_atomic_inc(&sbq->wake_index);
>>   			wake_up_nr(&ws->wait, wake_batch);
>> -			return false;
>> +			if (!nr || *nr <= 0)
>> +				return false;
>>   		}
>>   
>>   		return true;
>> +	} else if (nr && *nr) {
>> +		(*nr)--;
>> +		goto again;
>>   	}
>>   
>>   	return false;
> Hi Song,
>
> See if advancing wake_index can survive your tests.

Hi

Thanks for your suggestion, this problem not only needs to switch a wait 
queue, but also needs to consume "wait_cnt" correctly.


Thanks


>
> Only for thoughts now.
>
> Hillf
> --- a/lib/sbitmap.c
> +++ b/lib/sbitmap.c
> @@ -611,6 +611,7 @@ static bool __sbq_wake_up(struct sbitmap
>   	if (wait_cnt <= 0) {
>   		int ret;
>   
> +		sbq_index_atomic_inc(&sbq->wake_index);
>   		wake_batch = READ_ONCE(sbq->wake_batch);
>   
>   		/*
> @@ -627,7 +628,6 @@ static bool __sbq_wake_up(struct sbitmap
>   		 */
>   		ret = atomic_cmpxchg(&ws->wait_cnt, wait_cnt, wake_batch);
>   		if (ret == wait_cnt) {
> -			sbq_index_atomic_inc(&sbq->wake_index);
>   			wake_up_nr(&ws->wait, wake_batch);
>   			return false;
>   		}