Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Subject: Re: [PATCH] Revert "bfq: Fix computation of shallow depth"
To:     Jens Axboe <axboe@kernel.dk>, Jan Kara <jack@suse.cz>
Cc:     paolo.valente@linaro.org, linux-kernel@vger.kernel.org,
        linux-block@vger.kernel.org
References: <20210129111808.45796-1-linf@wangsu.com>
 <20210202122836.GC17147@quack2.suse.cz>
 <f92a55b4-0743-a700-8296-e739f4fcf093@kernel.dk>
From:   Lin Feng <linf@wangsu.com>
Message-ID: <9747bada-05f8-b08a-ced6-f14874b89024@wangsu.com>
Date:   Wed, 3 Feb 2021 10:36:43 +0800
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:78.0) Gecko/20100101
 Thunderbird/78.4.0
MIME-Version: 1.0
In-Reply-To: <f92a55b4-0743-a700-8296-e739f4fcf093@kernel.dk>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Precedence: bulk

Hi all,

On 2/2/21 22:20, Jens Axboe wrote:
> On 2/2/21 5:28 AM, Jan Kara wrote:
>> Hello!
>>
>> On Fri 29-01-21 19:18:08, Lin Feng wrote:
>>> This reverts commit 6d4d273588378c65915acaf7b2ee74e9dd9c130a.
>>>
>>> bfq.limit_depth passes word_depths[] as shallow_depth down to sbitmap core
>>> sbitmap_get_shallow, which uses just the number to limit the scan depth of
>>> each bitmap word, formula:
>>> scan_percentage_for_each_word = shallow_depth / (1 << sbimap->shift) * 100%
>>
>> Looking at sbitmap_get_shallow() again more carefully, I agree that I
>> misunderstood how shallow_depth argument gets used and the original code
>> was correct and I broke it. Thanks for spotting this!
>>
>> What I didn't notice is that shallow_depth indeed gets used for each bitmap
>> word separately and not for bitmap as a whole. I'd say this could use some
>> more documentation but that's unrelated to your revert. So feel free to add:
>>
>> Reviewed-by: Jan Kara <jack@suse.cz>
> 
> I don't have the original patch (neither directly nor in the archive), so
> I had to hand-apply it. In any case, applied for 5.11, thanks.
> 

Take a look at linux-block.git tree, the hand-applied commit for this patch
is broken, the following changing line is left out:
-	bfqd->word_depths[0][1] = max((bt->sb.depth * 3) >> 2, 1U);
+	bfqd->word_depths[0][1] = max(((1U << bt->sb.shift) * 3) >> 2, 1U);

Sorry for making troubles to you, I will resend this patch with tiny commit
log typo fix(sbimap -> sbitmap) and attaching Jan's Reviewed-by, also thanks
his time for reviewing.

Hope this time lkml server will not block my patch.

Thanks,
linfeng