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 X-Spam-Level: X-Spam-Status: No, score=-12.3 required=3.0 tests=BAYES_00,DKIM_SIGNED, DKIM_VALID,DKIM_VALID_AU,HEADER_FROM_DIFFERENT_DOMAINS,INCLUDES_CR_TRAILER, MAILING_LIST_MULTI,NICE_REPLY_A,SPF_HELO_NONE,SPF_PASS,URIBL_BLOCKED, USER_AGENT_SANE_1 autolearn=ham autolearn_force=no version=3.4.0 Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id C4892C433E0 for ; Thu, 18 Feb 2021 18:16:44 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by mail.kernel.org (Postfix) with ESMTP id 834BB64EAD for ; Thu, 18 Feb 2021 18:16:44 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S231326AbhBRSQa (ORCPT ); Thu, 18 Feb 2021 13:16:30 -0500 Received: from mx2.suse.de ([195.135.220.15]:42176 "EHLO mx2.suse.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S232389AbhBRP3N (ORCPT ); Thu, 18 Feb 2021 10:29:13 -0500 X-Virus-Scanned: by amavisd-new at test-mx.suse.de DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=suse.com; s=susede1; t=1613662104; h=from:from:reply-to:date:date:message-id:message-id:to:to:cc: mime-version:mime-version:content-type:content-type: content-transfer-encoding:content-transfer-encoding: in-reply-to:in-reply-to:references:references:autocrypt:autocrypt; bh=zGcDkIOvJjW3mlFMtSYMMqs3yYFC8kl+WhnBz2WVBvU=; b=ubAclM7bBnVo+rlWSSTsKhk5FVewTgHHnVHDoUt+5B7uRLWlOUDIQQxcvyItOK9F7np/x+ GfUvqDbd4BOfXzKLcsGxCg3FaFwCwR9YeV+gDjo1VWHmQe3lFHTEHlIkQw8lUGQXX50V+q NrM3VxKIvUowhMv35jjyZGGh3KR8yLM= Received: from relay2.suse.de (unknown [195.135.221.27]) by mx2.suse.de (Postfix) with ESMTP id 3D40AAD57; Thu, 18 Feb 2021 15:28:24 +0000 (UTC) Subject: Re: [PATCH] btrfs: make btrfs_dirty_inode() to always reserve metadata space To: Qu Wenruo , linux-btrfs@vger.kernel.org References: <20210108053659.87728-1-wqu@suse.com> From: Nikolay Borisov Autocrypt: addr=nborisov@suse.com; prefer-encrypt=mutual; keydata= mQINBFiKBz4BEADNHZmqwhuN6EAzXj9SpPpH/nSSP8YgfwoOqwrP+JR4pIqRK0AWWeWCSwmZ T7g+RbfPFlmQp+EwFWOtABXlKC54zgSf+uulGwx5JAUFVUIRBmnHOYi/lUiE0yhpnb1KCA7f u/W+DkwGerXqhhe9TvQoGwgCKNfzFPZoM+gZrm+kWv03QLUCr210n4cwaCPJ0Nr9Z3c582xc bCUVbsjt7BN0CFa2BByulrx5xD9sDAYIqfLCcZetAqsTRGxM7LD0kh5WlKzOeAXj5r8DOrU2 GdZS33uKZI/kZJZVytSmZpswDsKhnGzRN1BANGP8sC+WD4eRXajOmNh2HL4P+meO1TlM3GLl EQd2shHFY0qjEo7wxKZI1RyZZ5AgJnSmehrPCyuIyVY210CbMaIKHUIsTqRgY5GaNME24w7h TyyVCy2qAM8fLJ4Vw5bycM/u5xfWm7gyTb9V1TkZ3o1MTrEsrcqFiRrBY94Rs0oQkZvunqia c+NprYSaOG1Cta14o94eMH271Kka/reEwSZkC7T+o9hZ4zi2CcLcY0DXj0qdId7vUKSJjEep c++s8ncFekh1MPhkOgNj8pk17OAESanmDwksmzh1j12lgA5lTFPrJeRNu6/isC2zyZhTwMWs k3LkcTa8ZXxh0RfWAqgx/ogKPk4ZxOXQEZetkEyTFghbRH2BIwARAQABtCNOaWtvbGF5IEJv cmlzb3YgPG5ib3Jpc292QHN1c2UuY29tPokCOAQTAQIAIgUCWIo48QIbAwYLCQgHAwIGFQgC CQoLBBYCAwECHgECF4AACgkQcb6CRuU/KFc0eg/9GLD3wTQz9iZHMFbjiqTCitD7B6dTLV1C ddZVlC8Hm/TophPts1bWZORAmYIihHHI1EIF19+bfIr46pvfTu0yFrJDLOADMDH+Ufzsfy2v HSqqWV/nOSWGXzh8bgg/ncLwrIdEwBQBN9SDS6aqsglagvwFD91UCg/TshLlRxD5BOnuzfzI Leyx2c6YmH7Oa1R4MX9Jo79SaKwdHt2yRN3SochVtxCyafDlZsE/efp21pMiaK1HoCOZTBp5 VzrIP85GATh18pN7YR9CuPxxN0V6IzT7IlhS4Jgj0NXh6vi1DlmKspr+FOevu4RVXqqcNTSS E2rycB2v6cttH21UUdu/0FtMBKh+rv8+yD49FxMYnTi1jwVzr208vDdRU2v7Ij/TxYt/v4O8 V+jNRKy5Fevca/1xroQBICXsNoFLr10X5IjmhAhqIH8Atpz/89ItS3+HWuE4BHB6RRLM0gy8 T7rN6ja+KegOGikp/VTwBlszhvfLhyoyjXI44Tf3oLSFM+8+qG3B7MNBHOt60CQlMkq0fGXd mm4xENl/SSeHsiomdveeq7cNGpHi6i6ntZK33XJLwvyf00PD7tip/GUj0Dic/ZUsoPSTF/mG EpuQiUZs8X2xjK/AS/l3wa4Kz2tlcOKSKpIpna7V1+CMNkNzaCOlbv7QwprAerKYywPCoOSC 7P25Ag0EWIoHPgEQAMiUqvRBZNvPvki34O/dcTodvLSyOmK/MMBDrzN8Cnk302XfnGlW/YAQ csMWISKKSpStc6tmD+2Y0z9WjyRqFr3EGfH1RXSv9Z1vmfPzU42jsdZn667UxrRcVQXUgoKg QYx055Q2FdUeaZSaivoIBD9WtJq/66UPXRRr4H/+Y5FaUZx+gWNGmBT6a0S/GQnHb9g3nonD jmDKGw+YO4P6aEMxyy3k9PstaoiyBXnzQASzdOi39BgWQuZfIQjN0aW+Dm8kOAfT5i/yk59h VV6v3NLHBjHVw9kHli3jwvsizIX9X2W8tb1SefaVxqvqO1132AO8V9CbE1DcVT8fzICvGi42 FoV/k0QOGwq+LmLf0t04Q0csEl+h69ZcqeBSQcIMm/Ir+NorfCr6HjrB6lW7giBkQl6hhomn l1mtDP6MTdbyYzEiBFcwQD4terc7S/8ELRRybWQHQp7sxQM/Lnuhs77MgY/e6c5AVWnMKd/z MKm4ru7A8+8gdHeydrRQSWDaVbfy3Hup0Ia76J9FaolnjB8YLUOJPdhI2vbvNCQ2ipxw3Y3c KhVIpGYqwdvFIiz0Fej7wnJICIrpJs/+XLQHyqcmERn3s/iWwBpeogrx2Lf8AGezqnv9woq7 OSoWlwXDJiUdaqPEB/HmGfqoRRN20jx+OOvuaBMPAPb+aKJyle8zABEBAAGJAh8EGAECAAkF AliKBz4CGwwACgkQcb6CRuU/KFdacg/+M3V3Ti9JYZEiIyVhqs+yHb6NMI1R0kkAmzsGQ1jU zSQUz9AVMR6T7v2fIETTT/f5Oout0+Hi9cY8uLpk8CWno9V9eR/B7Ifs2pAA8lh2nW43FFwp IDiSuDbH6oTLmiGCB206IvSuaQCp1fed8U6yuqGFcnf0ZpJm/sILG2ECdFK9RYnMIaeqlNQm iZicBY2lmlYFBEaMXHoy+K7nbOuizPWdUKoKHq+tmZ3iA+qL5s6Qlm4trH28/fPpFuOmgP8P K+7LpYLNSl1oQUr+WlqilPAuLcCo5Vdl7M7VFLMq4xxY/dY99aZx0ZJQYFx0w/6UkbDdFLzN upT7NIN68lZRucImffiWyN7CjH23X3Tni8bS9ubo7OON68NbPz1YIaYaHmnVQCjDyDXkQoKC R82Vf9mf5slj0Vlpf+/Wpsv/TH8X32ajva37oEQTkWNMsDxyw3aPSps6MaMafcN7k60y2Wk/ TCiLsRHFfMHFY6/lq/c0ZdOsGjgpIK0G0z6et9YU6MaPuKwNY4kBdjPNBwHreucrQVUdqRRm RcxmGC6ohvpqVGfhT48ZPZKZEWM+tZky0mO7bhZYxMXyVjBn4EoNTsXy1et9Y1dU3HVJ8fod 5UqrNrzIQFbdeM0/JqSLrtlTcXKJ7cYFa9ZM2AP7UIN9n1UWxq+OPY9YMOewVfYtL8M= Message-ID: Date: Thu, 18 Feb 2021 17:28:23 +0200 User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:68.0) Gecko/20100101 Thunderbird/68.10.0 MIME-Version: 1.0 In-Reply-To: <20210108053659.87728-1-wqu@suse.com> Content-Type: text/plain; charset=utf-8 Content-Language: en-US Content-Transfer-Encoding: 8bit Precedence: bulk List-ID: X-Mailing-List: linux-btrfs@vger.kernel.org On 8.01.21 г. 7:36 ч., Qu Wenruo wrote: > There are several qgroup flush related bugs fixed recently, all of them > are caused by the fact that we can trigger qgroup metadata space > reservation holding a transaction handle. > > Thankfully the only situation to trigger above reservation is > btrfs_dirty_inode(). > > Currently btrfs_dirty_inode() will try join transactio first, then > update the inode. > If btrfs_update_inode() fails with -ENOSPC, then it retry to start > transaction to reserve metadata space. > > This not only forces us to reserve metadata space with a transaction > handle hold, but can't handle other errors like -EDQUOT. > > This patch will make btrfs_dirty_inode() to call > btrfs_start_transaction() directly without first try joining then > starting, so that in try_flush_qgroup() we won't hold a trans handle. > > This will slow down btrfs_dirty_inode() but my fstests doesn't show too > much different for most test cases, thus it may be worthy to skip such > performance "optimization". > > Signed-off-by: Qu Wenruo Ok I actually run 2 tests against this patch. The first one is a 10 second run of stress-ng's utime test (stress-ng --temp-path /media/scratch --utime 4 -M -t 10 ; done) to see if I can reproduce intel's results and here's what I found: bogo ops/s real (Before-patch) bogo ops/s real (After Patch) 35993 32968 35712 33146 35369 32996 35544 33159 35623 33000 35939 33016 35693 32829 35562 32685 35675 32815 Std dev 182.161981912585 146.829034703967 HMean 35677.9600871036 32957.1111111111 Diff%: -7.626 So there's a 7.6% decrease in the rate of utime() calls we can make, given that we now start a transaction I'd say that's expected. The other test was a randwrite with fio as I was mostly worried that making btrfs_dirty_inode more expensive would hit write performance since file_update_times is called from the generic iter. But inspecting the code btrfs uses update_time_for_write which doesn't dirty the inode per-se as this is deferred to endio completion time. I also measured the impact during buffered read time as file_accessed is called a lot of times but the following bpftrace script: BEGIN {@execs = 0; } kprobe:btrfs_dirty_inode { @test[kstack] = count(); @execs++; } kprobe:touch_atime { @test[kstack] = count(); } END{ printf("Total btrfs_dirty_inode calls: %llu\n", @execs); } confirmed we only ever execute around 8 btrfs_dirty_inode out of 1048773 execution of touch_atimes from generic_file_buffered_read with the following fio workload: fio --name=random-readers --thread --ioengine=sync --iodepth=4 --rw=randread --bs=4k --direct=0 --size=1g --numjobs=4 --directory=/media/scratch --filename_format=FioWorkloads.\$jobnum --new_group --group_reporting=1 So performance-wise I'm inclined to give it a "pass".