Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <a364cb12-3241-50a7-fdd5-323825d0a911@kernel.org>
Date:   Tue, 13 Dec 2022 09:32:09 +0800
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:102.0) Gecko/20100101
 Thunderbird/102.5.0
Subject: Re: [PATCH v2] f2fs: don't call f2fs_issue_discard_timeout() when
 discard_cmd_cnt is 0 in f2fs_put_super()
Content-Language: en-US
To:     Jaegeuk Kim <jaegeuk@kernel.org>
Cc:     Yangtao Li <frank.li@vivo.com>,
        linux-f2fs-devel@lists.sourceforge.net,
        linux-kernel@vger.kernel.org
References: <974f5013-b6af-a39e-0b0f-2ce86253eaeb@kernel.org>
 <20221212141429.6329-1-frank.li@vivo.com>
 <2f100a4a-592d-f098-b204-efeef58341ee@kernel.org>
 <Y5evHVvzGC/8lMfK@google.com>
From:   Chao Yu <chao@kernel.org>
In-Reply-To: <Y5evHVvzGC/8lMfK@google.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 2022/12/13 6:45, Jaegeuk Kim wrote:
> On 12/12, Chao Yu wrote:
>> On 2022/12/12 22:14, Yangtao Li wrote:
>>> Hi Chao,
>>>
>>>> The difference here is, if we use f2fs_realtime_discard_enable() in
>>>> f2fs_put_super(), we will only write checkpoint w/ CP_TRIMMED flag
>>>> when discard option is enable and device supports discard.
>>>
>>>> But actually, if discard option is disabled, we still needs to give
>>>> put_super() a chance to write checkpoint w/ CP_TRIMMED flag.
>>>
>>> Why do we still have to set the CP_TRIMMED flag when the discard opt is not set.
>>> Did I miss something?
>>
>> Hi Yangtao,
>>
>> I guess it's up to scenario. e.g.
>>
>> mount w/ nodiscard and use FITRIM to trigger in-batch discard,
>> if we set CP_TRIMMED flag during umount, next time, after mount
>> w/ discard, it doesn't to issue redundant discard.
> 
> If fitrim was called with a range, we can get a wrong FI_TRIMMED flag. Isn't it

We can set CP_TRIMMED flag only if fitrim was called on full range w/ 4k granularity,
due to it will check sbi->discard_blks variable to make sure there is no range we
haven't trimmed.

> better to get a full discard range after remount even though some are redundant?

If nodiscard is set, and sbi->discard_blks becomes zero, it says a full range fitrim
was been triggered.

So, previous check condition has no problem, right?

	if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) &&
					!sbi->discard_blks && !dropped) {

Thanks,

> 
>>
>> Thanks,
>>
>>>
>>> Thx,
>>> Yangtao