From patchwork Fri Nov 22 23:53:21 2019 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Andreas Gruenbacher X-Patchwork-Id: 11258671 Return-Path: Received: from mail.kernel.org (pdx-korg-mail-1.web.codeaurora.org [172.30.200.123]) by pdx-korg-patchwork-2.web.codeaurora.org (Postfix) with ESMTP id 0267514C0 for ; Fri, 22 Nov 2019 23:53:42 +0000 (UTC) Received: from kanga.kvack.org (kanga.kvack.org [205.233.56.17]) by mail.kernel.org (Postfix) with ESMTP id C328B20714 for ; Fri, 22 Nov 2019 23:53:41 +0000 (UTC) Authentication-Results: mail.kernel.org; dkim=pass (1024-bit key) header.d=redhat.com header.i=@redhat.com header.b="egY9Rh7v" DMARC-Filter: OpenDMARC Filter v1.3.2 mail.kernel.org C328B20714 Authentication-Results: mail.kernel.org; dmarc=fail (p=none dis=none) header.from=redhat.com Authentication-Results: mail.kernel.org; spf=pass smtp.mailfrom=owner-linux-mm@kvack.org Received: by kanga.kvack.org (Postfix) id E967E6B052C; Fri, 22 Nov 2019 18:53:40 -0500 (EST) Delivered-To: linux-mm-outgoing@kvack.org Received: by kanga.kvack.org (Postfix, from userid 40) id E764D6B052D; Fri, 22 Nov 2019 18:53:40 -0500 (EST) X-Original-To: int-list-linux-mm@kvack.org X-Delivered-To: int-list-linux-mm@kvack.org Received: by kanga.kvack.org (Postfix, from userid 63042) id D83106B052E; Fri, 22 Nov 2019 18:53:40 -0500 (EST) X-Original-To: linux-mm@kvack.org X-Delivered-To: linux-mm@kvack.org Received: from forelay.hostedemail.com (smtprelay0193.hostedemail.com [216.40.44.193]) by kanga.kvack.org (Postfix) with ESMTP id C30036B052C for ; Fri, 22 Nov 2019 18:53:40 -0500 (EST) Received: from smtpin24.hostedemail.com (10.5.19.251.rfc1918.com [10.5.19.251]) by forelay01.hostedemail.com (Postfix) with SMTP id 3C399180AD81F for ; Fri, 22 Nov 2019 23:53:40 +0000 (UTC) X-FDA: 76185568200.24.dad49_773912ab9ed4a X-Spam-Summary: 1,0,0,,d41d8cd98f00b204,agruenba@redhat.com,:torvalds@linux-foundation.org:swhiteho@redhat.com:khlebnikov@yandex-team.ru:kirill@shutemov.name::akpm@linux-foundation.org:linux-kernel@vger.kernel.org:linux-fsdevel@vger.kernel.org:viro@zeniv.linux.org.uk:hannes@cmpxchg.org:cluster-devel@redhat.com:lsahlber@redhat.com:sfrench@samba.org:rpeterso@redhat.com:agruenba@redhat.com,RULES_HIT:30003:30012:30054:30070,0,RBL:205.139.110.120:@redhat.com:.lbl8.mailshell.net-62.18.0.100 66.10.201.10,CacheIP:none,Bayesian:0.5,0.5,0.5,Netcheck:none,DomainCache:0,MSF:not bulk,SPF:fp,MSBL:0,DNSBL:neutral,Custom_rules:0:0:0,LFtime:24,LUA_SUMMARY:none X-HE-Tag: dad49_773912ab9ed4a X-Filterd-Recvd-Size: 5676 Received: from us-smtp-1.mimecast.com (us-smtp-delivery-1.mimecast.com [205.139.110.120]) by imf41.hostedemail.com (Postfix) with ESMTP for ; Fri, 22 Nov 2019 23:53:39 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=redhat.com; s=mimecast20190719; t=1574466818; h=from:from:reply-to:subject:subject:date:date:message-id:message-id: to:to:cc:cc:mime-version:mime-version:content-type:content-type: content-transfer-encoding:content-transfer-encoding; bh=e8LnAK96cO1Kj05+ukhpTbSenDBS0l/Ge53Izwdy1dc=; b=egY9Rh7vKwRVZES4wbxgy9mEEMmdPJggnSamzXldkT/26iUHRDd96k8Xgtx5lfTKfdszsw zt+dDgSER2FZY00jEoR9q7aMmj8fxBtPnFxU8nv9yp+AwKupd2grgCIL8tRncMzrsRXtq2 12pVxn3WFjiAUMlZ9mx6hlC7InQDlyM= Received: from mimecast-mx01.redhat.com (mimecast-mx01.redhat.com [209.132.183.4]) (Using TLS) by relay.mimecast.com with ESMTP id us-mta-94-oS30aRTpOTqH0Wo7idawag-1; Fri, 22 Nov 2019 18:53:37 -0500 Received: from smtp.corp.redhat.com (int-mx06.intmail.prod.int.phx2.redhat.com [10.5.11.16]) (using TLSv1.2 with cipher AECDH-AES256-SHA (256/256 bits)) (No client certificate requested) by mimecast-mx01.redhat.com (Postfix) with ESMTPS id 8B6DE801E74; Fri, 22 Nov 2019 23:53:34 +0000 (UTC) Received: from max.com (ovpn-204-21.brq.redhat.com [10.40.204.21]) by smtp.corp.redhat.com (Postfix) with ESMTP id CF6275C1B5; Fri, 22 Nov 2019 23:53:26 +0000 (UTC) From: Andreas Gruenbacher To: Linus Torvalds Cc: Steven Whitehouse , Konstantin Khlebnikov , "Kirill A. Shutemov" , linux-mm@kvack.org, Andrew Morton , linux-kernel@vger.kernel.org, linux-fsdevel@vger.kernel.org, Alexander Viro , Johannes Weiner , cluster-devel@redhat.com, Ronnie Sahlberg , Steve French , Bob Peterson , Andreas Gruenbacher Subject: [RFC PATCH 0/3] Rework the gfs2 read and page fault locking Date: Sat, 23 Nov 2019 00:53:21 +0100 Message-Id: <20191122235324.17245-1-agruenba@redhat.com> MIME-Version: 1.0 X-Scanned-By: MIMEDefang 2.79 on 10.5.11.16 X-MC-Unique: oS30aRTpOTqH0Wo7idawag-1 X-Mimecast-Spam-Score: 0 X-Bogosity: Ham, tests=bogofilter, spamicity=0.000000, version=1.2.4 Sender: owner-linux-mm@kvack.org Precedence: bulk X-Loop: owner-majordomo@kvack.org List-ID: Hello, this patch series moves the glock lock taking in gfs2 from the ->readpage and ->readpages inode operations to the ->read_iter file and ->fault vm operations. To achieve that, we add flags to the generic_file_read_iter and filemap_fault generic helpers. This proposal was triggered by the following discussion: https://lore.kernel.org/linux-fsdevel/157225677483.3442.4227193290486305330.stgit@buzz/ In that thread, Linus argued that filesystems should make sure the inode size is sufficiently up-to-date before calling the generic helpers, and that filesystems can do it themselves if they want more than that. That's surely doable. However, implementing those operations properly at the filesystem level quickly becomes complicated when it gets to things like readahead. In addition, those slightly modified copies of those helpers would surely diverge from their originals over time, and maintaining them properly would become hard. So I hope the relatively small changes to make the original helpers slightly more flexible will be acceptable instead. With the IOCB_CACHED flag added by one of the patches in this series, the code that Konstantin's initial patch adds to generic_file_buffered_read could be made conditional on the IOCB_CACHED flag being cleared. That way, it won't misfire on filesystems that allow a stale inode size. (I'm not sure if any filesystems other than gfs2 are actually affected.) Some additional explanation: The cache consistency model of filesystems like gfs2 is such that if pages are found in an inode's address space, those pages as well as the inode size are up to date and can be used without taking any filesystem locks. If a page is not cached, filesystem locks must be taken before the page can be read; this will also bring the inode size up to date. Thus far, gfs2 has taken the filesystem locks inside the ->readpage and ->readpages address space operations. A better approach seems to be to take those locks earlier, in the ->read_iter file and ->fault vm operations. This would also avoid a lock inversion in ->readpages. We obviously want to avoid taking the filesystem locks unnecessarily when the pages we are looking for are already cached; otherwise, we would cripple performance. So we need to check if those pages are present first. That's actually exactly what the generic_file_read_iter and filemap_fault helpers do already anyway, except that they will call into ->readpage and ->readpages when they find pages missing. Instead of that, we'd like those helpers to return with an error code that allows us to retry the operation after taking the filesystem locks. Thanks, Andreas Andreas Gruenbacher (3): fs: Add IOCB_CACHED flag for generic_file_read_iter fs: Add FAULT_FLAG_CACHED flag for filemap_fault gfs2: Rework read and page fault locking fs/gfs2/aops.c | 36 +++++-------------------- fs/gfs2/file.c | 66 ++++++++++++++++++++++++++++++++++++++++++++-- include/linux/fs.h | 1 + include/linux/mm.h | 4 ++- mm/filemap.c | 60 ++++++++++++++++++++++++++++++----------- 5 files changed, 119 insertions(+), 48 deletions(-)