Subject: Re: [PATCH 4/9] writeback: track if we're sleeping on progress in
 balance_dirty_pages()
To: Jan Kara <jack@suse.cz>
References: <1459350477-16404-1-git-send-email-axboe@fb.com>
 <1459350477-16404-5-git-send-email-axboe@fb.com>
 <20160413130846.GA23550@quack2.suse.cz>
CC: <linux-kernel@vger.kernel.org>, <linux-fsdevel@vger.kernel.org>,
        <linux-block@vger.kernel.org>
From: Jens Axboe <axboe@fb.com>
Message-ID: <570E5599.7000008@fb.com>
Date: Wed, 13 Apr 2016 08:20:09 -0600
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:38.0) Gecko/20100101
 Thunderbird/38.6.0
MIME-Version: 1.0
In-Reply-To: <20160413130846.GA23550@quack2.suse.cz>
Content-Type: text/plain; charset="windows-1252"; format=flowed
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1179
Lines: 32

On 04/13/2016 07:08 AM, Jan Kara wrote:
> On Wed 30-03-16 09:07:52, Jens Axboe wrote:
>> Note in the bdi_writeback structure if a task is currently being
>> limited in balance_dirty_pages(), waiting for writeback to
>> proceed.
> ...
>> diff --git a/mm/page-writeback.c b/mm/page-writeback.c
>> index 11ff8f758631..15e696bc5d14 100644
>> --- a/mm/page-writeback.c
>> +++ b/mm/page-writeback.c
>> @@ -1746,7 +1746,9 @@ pause:
>>   					  pause,
>>   					  start_time);
>>   		__set_current_state(TASK_KILLABLE);
>> +		wb->dirty_sleeping = 1;
>>   		io_schedule_timeout(pause);
>> +		wb->dirty_sleeping = 0;
>
> Huh, but wb->dirty_sleeping is shared by all the processes in the system.
> So this is seriously racy, isn't it? You rather need a counter for this to
> work.

Sure, but it's not _that_ important. It's like wb->dirty_exceeded, we 
have an equally relaxed relationship.

I don't mind making it more solid, but I can't make it a counter without 
making it atomic. Which is why I left it as just a basic assignment. But 
I guess since we only fiddle with it when going to sleep, we can make it 
an atomic and not have to worry about the potential impact.

-- 
Jens Axboe