Received-SPF: pass (google.com: domain of linux-kernel+bounces-29634-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:4601:e00::3 as permitted sender) client-ip=2604:1380:4601:e00::3;
Message-ID: <19ff9020-043c-ffc1-c09c-3c40814620b2@huaweicloud.com>
Date: Thu, 18 Jan 2024 09:49:48 +0800
Precedence: bulk
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:91.0) Gecko/20100101
 Thunderbird/91.10.0
Subject: Re: [PATCH] md: Don't clear MD_CLOSING when the raid is about to stop
To: Yu Kuai <yukuai1@huaweicloud.com>,
 Mariusz Tkaczyk <mariusz.tkaczyk@linux.intel.com>, linan666@huaweicloud.com
Cc: song@kernel.org, zlliu@suse.com, neilb@suse.com, shli@fb.com,
 linux-raid@vger.kernel.org, linux-kernel@vger.kernel.org,
 yi.zhang@huawei.com, houtao1@huawei.com, yangerkun@huawei.com,
 "yukuai (C)" <yukuai3@huawei.com>
References: <20231211081714.1923567-1-linan666@huaweicloud.com>
 <20231211105620.00001753@linux.intel.com>
 <f0ab24e5-eb0a-d564-19d4-b72ecedff34f@huaweicloud.com>
From: Li Nan <linan666@huaweicloud.com>
In-Reply-To: <f0ab24e5-eb0a-d564-19d4-b72ecedff34f@huaweicloud.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit


在 2023/12/12 11:21, Yu Kuai 写道:
> Hi,
> 
> 在 2023/12/11 17:56, Mariusz Tkaczyk 写道:
>> On Mon, 11 Dec 2023 16:17:14 +0800
>> linan666@huaweicloud.com wrote:
>>
>>> From: Li Nan <linan122@huawei.com>
>>>
>>> The raid should not be opened anymore when it is about to be stopped.
>>> However, other processes can open it again if the flag MD_CLOSING is
>>> cleared before exiting. From now on, this flag will not be cleared when
>>> the raid will be stopped.
>>>
>>> Fixes: 065e519e71b2 ("md: MD_CLOSING needs to be cleared after called
>>> md_set_readonly or do_md_stop") Signed-off-by: Li Nan 
>>> <linan122@huawei.com>
>>
>> Hello Li Nan,
>> I was there when I needed to fix this:
>> https://git.kernel.org/pub/scm/linux/kernel/git/song/md.git/commit/?h=md-next&id=c8870379a21fbd9ad14ca36204ccfbe9d25def43 
>>
>>
>> For sure, you have to consider applying same solution for array_store 
>> "clear".
>> Minor nit below.
>>
>> Thanks,
>> Mariusz
>>
>>> ---
>>>   drivers/md/md.c | 8 +++-----
>>>   1 file changed, 3 insertions(+), 5 deletions(-)
>>>
>>> diff --git a/drivers/md/md.c b/drivers/md/md.c
>>> index 4e9fe5cbeedc..ebdfc9068a60 100644
>>> --- a/drivers/md/md.c
>>> +++ b/drivers/md/md.c
>>> @@ -6238,7 +6238,6 @@ static void md_clean(struct mddev *mddev)
>>>       mddev->persistent = 0;
>>>       mddev->level = LEVEL_NONE;
>>>       mddev->clevel[0] = 0;
>>> -    mddev->flags = 0;
>>
>> I recommend (safety recommendation):
>>     mddev->flags = MD_CLOSING;
> 
> Taking a look I think both MD_CLOSING and MD_DELETED should not be
> cleared, however, there is no guarantee that MD_CLOSING will be set
> before md_clean, because mdadm can be removed without running. Hence I
> think just set MD_CLOSING is werid.
> 
> I think the proper way is to keep MD_CLOSING and MD_DELETED if they are
> set. However, there is no such api to clear other bits at once. Since
> we're not expecting anyone else to write flags, following maybe
> acceptable:
> 
> mddev->flags &= BIT_ULL_MASK(MD_CLOSING) | BIT_ULL_MASK(MD_DELETED);
> 

MD_DELETED is only set after mddev->active is put to 0. We need to open 
mddev and get it before stropping raid, so the active must not be 0 and
MD_DELETED  will not be set in md_clean.

> Or after making sure other flags cannot race, this patch is ok.
> 
> Thanks,
> Kuai
> 
>>
>> Unless you can prove that other flags cannot race.
>>
>>>       mddev->sb_flags = 0;
>>>       mddev->ro = MD_RDWR;
>>>       mddev->metadata_type[0] = 0;
>>
>> .
>>
> 
> 
> .

-- 
Thanks,
Nan