From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id 8E716C4332F for ; Tue, 8 Nov 2022 14:31:04 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S234159AbiKHObC (ORCPT ); Tue, 8 Nov 2022 09:31:02 -0500 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:36338 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S234246AbiKHOap (ORCPT ); Tue, 8 Nov 2022 09:30:45 -0500 Received: from dfw.source.kernel.org (dfw.source.kernel.org [139.178.84.217]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 6956454B08 for ; Tue, 8 Nov 2022 06:30:18 -0800 (PST) Received: from smtp.kernel.org (relay.kernel.org [52.25.139.140]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by dfw.source.kernel.org (Postfix) with ESMTPS id EABB8615F2 for ; Tue, 8 Nov 2022 14:30:17 +0000 (UTC) Received: by smtp.kernel.org (Postfix) with ESMTPSA id 19E4FC433D7; Tue, 8 Nov 2022 14:30:15 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=kernel.org; s=k20201202; t=1667917817; bh=HfYyn2zfWoXVMiL04ajYFQAOGugFAO+1dw/DJMyUUU8=; h=Date:Subject:To:Cc:References:From:In-Reply-To:From; b=Ue6Zs8utYdJmMhWRb1grLj7XVg1S4ZS8CcHg3jCMIwRfqQfu9PhbaGVHqA5LGmlZG mB7kHOoOgL5NIJNOWQrNNVPKj/lHdaRCFYrBsybYiaAFsHyvzOas6o6UpgHbElIx50 3/xKG4SS/2vgmkH8ZKZWw1sLsN6TMstkH+2COUbTf9BaCUy8OKPwRZuJJUoFkZKkIe funsPjlipwr1tnbBiNgke/m3jZnmmCFYz2i53hooT1xc3vRt7jbM2ppaYJd+z3Qu1v 9h0GfrD8lnSmFOyOU7RQAWqu4lrTO75AS4qymZ53zP7JUurtgj7e0TgxJ95P4f4rQQ gJ01DEKAJZrEg== Message-ID: <8368702c-ae33-b810-a6c3-ac8fa29998df@kernel.org> Date: Tue, 8 Nov 2022 22:30:13 +0800 MIME-Version: 1.0 User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:102.0) Gecko/20100101 Thunderbird/102.4.1 Subject: Re: [PATCH] f2fs: separate IPU policy for fdatasync from F2FS_IPU_FSYNC To: qixiaoyu Cc: Jaegeuk Kim , linux-kernel@vger.kernel.org, linux-f2fs-devel@lists.sourceforge.net, qixiaoyu1 References: <20221021023136.22863-1-qixiaoyu1@xiaomi.com> <20221102122518.GB22857@mi-HP-ProDesk-680-G4-MT> <3d2b1141-995a-4bfb-4bf0-5227be25105a@kernel.org> <20221108123218.GC22857@mi-HP-ProDesk-680-G4-MT> Content-Language: en-US From: Chao Yu In-Reply-To: <20221108123218.GC22857@mi-HP-ProDesk-680-G4-MT> Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: 7bit Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On 2022/11/8 20:32, qixiaoyu wrote: > On Sun, Nov 06, 2022 at 09:54:59PM +0800, Chao Yu wrote: >> On 2022/11/2 20:25, qixiaoyu wrote: >>> Hi Chao, >>> >>> fdatasync do in-place-update to avoid additional node writes, but currently >>> it only do that with F2FS_IPU_FSYNC as: >>> >>> f2fs_do_sync_file: >>> if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks) >>> set_inode_flag(inode, FI_NEED_IPU); >>> >>> check_inplace_update_policy: >>> /* this is only set during fdatasync */ >>> if (policy & (0x1 << F2FS_IPU_FSYNC) && >>> is_inode_flag_set(inode, FI_NEED_IPU)) >>> return true; >>> >>> So this patch separate in-place-update of fdatasync from F2FS_IPU_FSYNC to >>> apply it to all IPU policy. >>> >>> BTW, we found small performance improvement with this patch on AndroBench app >>> using F2FS_IPU_SSR_UTIL on our product: >> >> How this patch affects performance when F2FS_IPU_SSR_UTIL is on? >> >> Thanks, >> > > SQLite test in AndroBench app use fdatasync to sync file to the disk. > When switch to F2FS_IPU_SSR_UTIL ipu_policy, it will use out-of-place-update > even though SQLite calls fdatasync, which will introduce extra meta data write. Why not using F2FS_IPU_SSR_UTIL | F2FS_IPU_FSYNC, I guess these two flags cover different scenarios, F2FS_IPU_SSR_UTIL for ssr case, and F2FS_IPU_FSYNC for f{data,}sync case. Thanks, > > Thanks. > >>> >>> F2FS_IPU_FSYNC F2FS_IPU_SSR_UTIL F2FS_IPU_SSR_UTIL(with patch) >>> SQLite Insert(QPS) 6818.08 6327.09(-7.20%) 6757.72 >>> SQLite Update(QPS) 6528.81 6336.57(-2.94%) 6490.77 >>> SQLite Delete(QPS) 9724.68 9378.37(-3.56%) 9622.27 >>> >>> Thanks >>> >>> On Tue, Nov 01, 2022 at 11:14:55PM +0800, Chao Yu wrote: >>>> On 2022/10/21 10:31, qixiaoyu1 wrote: >>>>> Currently IPU policy for fdatasync is coupled with F2FS_IPU_FSYNC. >>>>> Fix to apply it to all IPU policy. >>>> >>>> Xiaoyu, >>>> >>>> Sorry for the delay. >>>> >>>> I didn't get the point, can you please explain more about the >>>> issue? >>>> >>>> Thanks, >>>> >>>>> >>>>> Signed-off-by: qixiaoyu1 >>>>> --- >>>>> fs/f2fs/data.c | 8 +++----- >>>>> fs/f2fs/file.c | 4 +++- >>>>> 2 files changed, 6 insertions(+), 6 deletions(-) >>>>> >>>>> diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c >>>>> index a71e818cd67b..fec8e15fe820 100644 >>>>> --- a/fs/f2fs/data.c >>>>> +++ b/fs/f2fs/data.c >>>>> @@ -2518,6 +2518,9 @@ static inline bool check_inplace_update_policy(struct inode *inode, >>>>> if (policy & (0x1 << F2FS_IPU_HONOR_OPU_WRITE) && >>>>> is_inode_flag_set(inode, FI_OPU_WRITE)) >>>>> return false; >>>>> + /* this is set by fdatasync or F2FS_IPU_FSYNC policy */ >>>>> + if (is_inode_flag_set(inode, FI_NEED_IPU)) >>>>> + return true; >>>>> if (policy & (0x1 << F2FS_IPU_FORCE)) >>>>> return true; >>>>> if (policy & (0x1 << F2FS_IPU_SSR) && f2fs_need_SSR(sbi)) >>>>> @@ -2538,11 +2541,6 @@ static inline bool check_inplace_update_policy(struct inode *inode, >>>>> !IS_ENCRYPTED(inode)) >>>>> return true; >>>>> - /* this is only set during fdatasync */ >>>>> - if (policy & (0x1 << F2FS_IPU_FSYNC) && >>>>> - is_inode_flag_set(inode, FI_NEED_IPU)) >>>>> - return true; >>>>> - >>>>> if (unlikely(fio && is_sbi_flag_set(sbi, SBI_CP_DISABLED) && >>>>> !f2fs_is_checkpointed_data(sbi, fio->old_blkaddr))) >>>>> return true; >>>>> diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c >>>>> index 82cda1258227..08091550cdf2 100644 >>>>> --- a/fs/f2fs/file.c >>>>> +++ b/fs/f2fs/file.c >>>>> @@ -270,8 +270,10 @@ static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end, >>>>> goto go_write; >>>>> /* if fdatasync is triggered, let's do in-place-update */ >>>>> - if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks) >>>>> + if (datasync || (SM_I(sbi)->ipu_policy & (0x1 << F2FS_IPU_FSYNC) && >>>>> + get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)) >>>>> set_inode_flag(inode, FI_NEED_IPU); >>>>> + >>>>> ret = file_write_and_wait_range(file, start, end); >>>>> clear_inode_flag(inode, FI_NEED_IPU);