Message-ID: <fbbb46dc-ac07-9a99-dece-f0077a9fd491@linux.dev>
Date:   Thu, 9 Mar 2023 15:27:04 +0800
MIME-Version: 1.0
Subject: Re: [PATCH -next] raid10: fix leak of io accounting
Content-Language: en-US
To:     Yu Kuai <yukuai1@huaweicloud.com>, song@kernel.org
Cc:     linux-raid@vger.kernel.org, linux-kernel@vger.kernel.org,
        yi.zhang@huawei.com, yangerkun@huawei.com,
        "yukuai (C)" <yukuai3@huawei.com>
References: <20230304070133.1134975-1-yukuai1@huaweicloud.com>
 <a2551c50-feea-bcbe-00ed-802456b5a19f@linux.dev>
 <76d32496-641e-c93a-df77-9ce9d9c1a1e1@huaweicloud.com>
From:   Guoqing Jiang <guoqing.jiang@linux.dev>
In-Reply-To: <76d32496-641e-c93a-df77-9ce9d9c1a1e1@huaweicloud.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
Precedence: bulk


On 3/9/23 14:56, Yu Kuai wrote:
> Hi,
>
> 在 2023/03/09 14:36, Guoqing Jiang 写道:
>> Hi,
>>
>> What do you mean 'leak' here?
>
> I try to mean that inflight counting is leaked, because it's increased
> twice for one io.

How about change the subject to something like?

'md/raid10: Don't call bio_start_io_acct twice for bio which experienced 
read error'

>
>>
>> On 3/4/23 15:01, Yu Kuai wrote:
>>> From: Yu Kuai <yukuai3@huawei.com>
>>>
>>> handle_read_error() will resumit r10_bio by raid10_read_request(), 
>>> which
>>> will call bio_start_io_acct() again, while bio_end_io_acct() will only
>>> be called once.
>>>
>>> Fix the problem by don't account io again from handle_read_error().
>>
>> My understanding is it caused inaccurate io stats for bio which had a 
>> read
>> error.
>>
>>> Fixes: 528bc2cf2fcc ("md/raid10: enable io accounting")
>>> Signed-off-by: Yu Kuai <yukuai3@huawei.com>
>>> ---
>>>   drivers/md/raid10.c | 8 ++++----
>>>   1 file changed, 4 insertions(+), 4 deletions(-)
>>>
>>> diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c
>>> index 6c66357f92f5..4f8edb6ea3e2 100644
>>> --- a/drivers/md/raid10.c
>>> +++ b/drivers/md/raid10.c
>>> @@ -1173,7 +1173,7 @@ static bool regular_request_wait(struct mddev 
>>> *mddev, struct r10conf *conf,
>>>   }
>>>   static void raid10_read_request(struct mddev *mddev, struct bio *bio,
>>> -                struct r10bio *r10_bio)
>>> +                struct r10bio *r10_bio, bool handle_error)
>>>   {
>>>       struct r10conf *conf = mddev->private;
>>>       struct bio *read_bio;
>>> @@ -1244,7 +1244,7 @@ static void raid10_read_request(struct mddev 
>>> *mddev, struct bio *bio,
>>>       }
>>>       slot = r10_bio->read_slot;
>>> -    if (blk_queue_io_stat(bio->bi_bdev->bd_disk->queue))
>>> +    if (!handle_error && 
>>> blk_queue_io_stat(bio->bi_bdev->bd_disk->queue))
>>>           r10_bio->start_time = bio_start_io_acct(bio);
>>
>> I think a simpler way is just check R10BIO_ReadError here.
>
> No, I'm afraid this is incorrect because handle_read_error clears the
> state before resubmiting the r10bio.

Right,

Acked-by: Guoqing Jiang <guoqing.jiang@linux.dev>

Thanks,
Guoqing