Message-ID: <bc8f48ce-7280-c70a-db7d-825923f08b48@bytedance.com>
Date:   Fri, 17 Feb 2023 00:10:42 +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: [External] Re: [PATCH v2] ext4: make dioread_nolock consistent in
 each mapping round
To:     "Ritesh Harjani (IBM)" <ritesh.list@gmail.com>, tytso@mit.edu,
        adilger.kernel@dilger.ca
Cc:     linux-ext4@vger.kernel.org, linux-kernel@vger.kernel.org,
        yi.zhang@huawei.com
References: <87fsb5lpn1.fsf@doe.com>
From:   hanjinke <hanjinke.666@bytedance.com>
In-Reply-To: <87fsb5lpn1.fsf@doe.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
Precedence: bulk


在 2023/2/16 下午9:13, Ritesh Harjani (IBM) 写道:
> Jinke Han <hanjinke.666@bytedance.com> writes:
> 
>> From: Jinke Han <hanjinke.666@bytedance.com>
>>
>> When disable and enable dioread_nolock by remount, we may see
>> dioread_lock in ext4_do_writepages while see dioread_nolock in
>> mpage_map_one_extent. This inconsistency may triger the warning
>> in ext4_add_complete_io when the io_end->handle is NULL. Although
>> this warning is harmless in most cases, there is still a risk of
>> insufficient log reservation in conversion of unwritten extents.
>>
> 
> Sorry, I haven't completely gone through the patch yet. But this idea of
> caching the initial value of mount parameter and passing it do different
> functions while an I/O request completes, is not looking right to me.
> 
> If that's the case shouldn't we disallow this mount option to change
> until all the outstanding I/O's are done or complete?
> Then we need not cache the value of dioread_nolock at the start of
> writepages and continue to pass it down in case it it changes.
> 
> Just my initial thoughts.
> 
> -ritesh
> 

Fair enough, thanks.