From patchwork Thu Oct 24 07:50:34 2019 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Ian Kent X-Patchwork-Id: 11208411 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 41C4914ED for ; Thu, 24 Oct 2019 07:50:39 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.kernel.org (Postfix) with ESMTP id 204462166E for ; Thu, 24 Oct 2019 07:50:39 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S2407220AbfJXHui (ORCPT ); Thu, 24 Oct 2019 03:50:38 -0400 Received: from icp-osb-irony-out1.external.iinet.net.au ([203.59.1.210]:26989 "EHLO icp-osb-irony-out1.external.iinet.net.au" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S2407212AbfJXHui (ORCPT ); Thu, 24 Oct 2019 03:50:38 -0400 X-IronPort-Anti-Spam-Filtered: true X-IronPort-Anti-Spam-Result: A2CoAADRVrFd/0e30XYNWBwBAQEBAQcBAREBBAQBAYFqBAEBCwEBgXKBGIEwhCiPSQaBEYEkiH6FIAGKJ4FnCQEBAQEBAQEBAScQAQGEOwMCg1s3Bg4CDAEBAQQBAQEBAQUDAYVYTE4BAQQHAYUDJwRSKA0CJgJJFgqDLIJSsXR1fzMahB8BCwGGBIEOKAGBZIpCeIEHgREzgipzgmIBA4FEA4Mogl4EjGokgi83hV9hQ5Zsgi6HD442DI4fA4sklmKMJocNgXtNLgqDJwkWMYM2F4EEAQKHIjuFTGeHPYMZgn4BK4IuAQE X-IPAS-Result: A2CoAADRVrFd/0e30XYNWBwBAQEBAQcBAREBBAQBAYFqBAEBCwEBgXKBGIEwhCiPSQaBEYEkiH6FIAGKJ4FnCQEBAQEBAQEBAScQAQGEOwMCg1s3Bg4CDAEBAQQBAQEBAQUDAYVYTE4BAQQHAYUDJwRSKA0CJgJJFgqDLIJSsXR1fzMahB8BCwGGBIEOKAGBZIpCeIEHgREzgipzgmIBA4FEA4Mogl4EjGokgi83hV9hQ5Zsgi6HD442DI4fA4sklmKMJocNgXtNLgqDJwkWMYM2F4EEAQKHIjuFTGeHPYMZgn4BK4IuAQE X-IronPort-AV: E=Sophos;i="5.68,224,1569254400"; d="scan'208";a="250043657" Received: from unknown (HELO [192.168.1.222]) ([118.209.183.71]) by icp-osb-irony-out1.iinet.net.au with ESMTP; 24 Oct 2019 15:50:34 +0800 Subject: [PATCH v7 00/17] xfs: mount API patch series From: Ian Kent To: linux-xfs Cc: Christoph Hellwig , "Darrick J. Wong" , Brian Foster , Eric Sandeen , David Howells , Dave Chinner , Al Viro Date: Thu, 24 Oct 2019 15:50:34 +0800 Message-ID: <157190333868.27074.13987695222060552856.stgit@fedora-28> User-Agent: StGit/unknown-version MIME-Version: 1.0 Sender: linux-xfs-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-xfs@vger.kernel.org This patch series add support to xfs for the new kernel mount API as described in the LWN article at https://lwn.net/Articles/780267/. In the article there's a lengthy description of the reasons for adopting the API and problems expected to be resolved by using it. The series has been applied to the repository located at git://git.kernel.org/pub/scm/fs/xfs/xfs-linux.git, and built and some simple tests run on it along with the generic xfstests. Note: the patches "xfs: add xfs_remount_rw() helper" and "xfs: add xfs_remount_ro() helper" that have Reviewed-by attributions each needed a forward declartion added due grouping all the mount related code together. Reviewers may want to check the attribution is still acceptable. The first patch in the series is needed only for testing. It is now present in the current upstream rc kernel so it will need to be dropped before merging. Other things that continue to cause me concern: - Message logging. There is error logging done in the VFS by the mount-api code, some is VFS specific while some is file system specific. This can lead to duplicated and sometimes inconsistent logging. The mount-api feature of saving error message text to the mount context for later retrieval by fsopen()/fsconfig()/fsmount() users is the reason the mount-api log macros are present. But, at the moment (last time I looked), these macros will either log the error message or save it to the mount context. There's not yet a way to modify this behaviour so it which can lead to messages, possibly needed for debug purposes, not being sent to the kernel log. There's also the pr_xxx() log functions (not a problem for xfs AFAICS) that aren't aware of the mount context at all. In the xfs patches I've used the same method that is used in gfs2 and was suggested by Al Viro (essentially return the error if fs_parse() returns one) except that I've also not used the mount api log macros to minimise the possibility of lost log messages. This isn't the best so follow up patches for RFC (with a slightly wider audience) will be needed to try and improve this aspect of the mount api. Changes for v7: - fix s/mount info/struct xfs_mount/g in the patches that remove the m_fsname_len and m_fsname struct xfs_mount fields. - also use IS_ENABLED() macro for the quota config check. - don't use typedef'ed struct definitions in new or chencged code. - extend comment lines to use full 80 characters and to 72 characters for commit log entries. - fix function parameter and variable declaration style inconsistencies. - use "return 0" instead of "break" in xfs_parse_param() as recommended. - avoid redundant option checks for the case where no options are given. - merge freeing of mp names and mp. - change name field of struct fs_parameter_description to lower case. - group mount related code together as much as possible to enhance readability. - changed mount api related functions to all use an "xfs_fc" prefix. Changes for v6: - drop get_tree_bdev() patch since it's now present in the xfs-linux repo. - eliminate unused mount info. field m_fsname_len. - eliminate mount info field m_fsname, use super block field s_id instead. - dont use XFS_IS_QUOTA_RUNNING() when checking if quota options have been specified. - combine mount-api specific patches into a single patch to help with bi-sectability. Changes for v5: - give error exit label in xfs_fill_super() a sensible name. - use original comment about options handling in xfs_fs_remount() for comment above xfs_reconfigure(). - use a much simpler comment in xfs_fc_free(), thanks to Brian Foster. - move cover letter comment about the first two patches above the revision comentary so it isn't missed. Changes for v4: - changed xfs_fill_super() cleanup back to what it was in v2, until I can work out what's causing the problem had previously seen (I can't reproduce it myself), since it looks like it was right from the start. - use get_tree_bdev() instead of vfs_get_block_super() in xfs_get_tree() as requested by Al Viro. - removed redundant initialisation in xfs_fs_fill_super(). - fix long line in xfs_validate_params(). - no need to validate if parameter parsing fails, just return the error. - summarise reconfigure comment about option handling, transfer bulk of comment to commit log message. - use minimal change in xfs_parse_param(), deffer discussion of mount api logging improvements until later and with a wider audience. Changes for v3: - fix struct xfs_fs_context initialisation in xfs_parseargs(). - move call to xfs_validate_params() below label "done". - if allocation of xfs_mount fails return ENOMEM immediately. - remove erroneous kfree(mp) in xfs_fill_super(). - move the removal of xfs_fs_remount() and xfs_test_remount_options() to the switch to mount api patch. - retain original usage of distinct