[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <36f1f74b-4376-89f3-0690-09f36e05d6e1@huaweicloud.com>
Date: Tue, 25 Mar 2025 15:27:06 +0800
From: Yu Kuai <yukuai1@...weicloud.com>
To: Coly Li <colyli@...nel.org>, Yu Kuai <yukuai1@...weicloud.com>
Cc: song@...nel.org, jgq516@...il.com, linux-raid@...r.kernel.org,
linux-kernel@...r.kernel.org, yi.zhang@...wei.com, yangerkun@...wei.com,
"yukuai (C)" <yukuai3@...wei.com>
Subject: Re: [PATCH] md/raid10: fix missing discard IO accounting
Hi,
在 2025/03/25 15:04, Coly Li 写道:
> On Tue, Mar 25, 2025 at 09:57:46AM +0800, Yu Kuai wrote:
>> From: Yu Kuai <yukuai3@...wei.com>
>>
>> md_account_bio() is not called from raid10_handle_discard(), now that we
>> handle bitmap inside md_account_bio(), also fix missing
>> bitmap_startwrite for discard.
>>
>> Test whole disk discard for 20G raid10:
>>
>> Before:
>> Device d/s dMB/s drqm/s %drqm d_await dareq-sz
>> md0 48.00 16.00 0.00 0.00 5.42 341.33
>>
>> After:
>> Device d/s dMB/s drqm/s %drqm d_await dareq-sz
>> md0 68.00 20462.00 0.00 0.00 2.65 308133.65
>>
>> Fixes: 528bc2cf2fcc ("md/raid10: enable io accounting")
>> Signed-off-by: Yu Kuai <yukuai3@...wei.com>
>
> Should we treat discard request as real I/O?
>
> Normally IMHO discard request should not be counted as real data transfer,
> correct me if I am wrong.
Normally it's not, that's why discard IOs are accounted separately in
the block layer.
Also notice that discard should be treated as write, because after
discard, reading will get zero data.
Thanks,
Kuai
>
> Thanks.
>
>
>> ---
>> drivers/md/raid10.c | 1 +
>> 1 file changed, 1 insertion(+)
>>
>> diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c
>> index 9d8516acf2fd..6ef65b4d1093 100644
>> --- a/drivers/md/raid10.c
>> +++ b/drivers/md/raid10.c
>> @@ -1735,6 +1735,7 @@ static int raid10_handle_discard(struct mddev *mddev, struct bio *bio)
>> * The discard bio returns only first r10bio finishes
>> */
>> if (first_copy) {
>> + md_account_bio(mddev, &bio);
>> r10_bio->master_bio = bio;
>> set_bit(R10BIO_Discard, &r10_bio->state);
>> first_copy = false;
>> --
>> 2.39.2
>>
>>
>
Powered by blists - more mailing lists