From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from ipmail06.adl2.internode.on.net ([150.101.137.129]:37766 "EHLO ipmail06.adl2.internode.on.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750904AbdK1W0k (ORCPT ); Tue, 28 Nov 2017 17:26:40 -0500 Date: Wed, 29 Nov 2017 09:26:37 +1100 From: Dave Chinner Subject: Re: [PATCH RFC 4/4] xfs: include an allocfree res for inobt modifications Message-ID: <20171128222637.GG5858@dastard> References: <20171127202434.43125-1-bfoster@redhat.com> <20171127202434.43125-5-bfoster@redhat.com> <20171127232719.GC5858@dastard> <20171128140459.GD45759@bfoster.bfoster> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20171128140459.GD45759@bfoster.bfoster> Sender: linux-xfs-owner@vger.kernel.org List-ID: List-Id: xfs To: Brian Foster Cc: linux-xfs@vger.kernel.org On Tue, Nov 28, 2017 at 09:04:59AM -0500, Brian Foster wrote: > On Tue, Nov 28, 2017 at 10:27:19AM +1100, Dave Chinner wrote: > > On Mon, Nov 27, 2017 at 03:24:34PM -0500, Brian Foster wrote: > > > STATIC uint > > > @@ -415,8 +414,8 @@ __xfs_calc_create_reservation( > > > * For icreate we can allocate some inodes giving: > > > * the agi and agf of the ag getting the new inodes: 2 * sectorsize > > > * the superblock for the nlink flag: sector size > > > - * the inode btree: max depth * blocksize > > > * the allocation btrees: 2 trees * (max depth - 1) * block size > > > + * the inobt (record insertion) > > > * the finobt (record insertion) > > > */ > > > STATIC uint > > > @@ -425,10 +424,10 @@ xfs_calc_icreate_resv_alloc( > > > { > > > return xfs_calc_buf_res(2, mp->m_sb.sb_sectsize) + > > > mp->m_sb.sb_sectsize + > > > - xfs_calc_buf_res(mp->m_in_maxlevels, XFS_FSB_TO_B(mp, 1)) + > > > xfs_calc_buf_res(xfs_allocfree_log_count(mp, 1), > > > XFS_FSB_TO_B(mp, 1)) + > > > - xfs_calc_finobt_res(mp, 0, 0); > > > + xfs_calc_inobt_res(mp) + > > > + xfs_calc_finobt_res(mp); > > > } > > > > return xfs_calc_buf_res(2, mp->m_sb.sb_sectsize) + > > mp->m_sb.sb_sectsize + > > xfs_calc_inode_chunk_res(mp, false) + > > xfs_calc_inobt_res(mp) + > > xfs_calc_finobt_res(mp); > > > > The icreate reservation doesn't currently include m_ialloc_blks at all. > The helper, as defined above, adds a reservation for associated headers. > Is that really necessary? My understanding is that icreate doesn't log > the inode chunk. Right, it uses ordered buffers to avoid needing to log them. > I suppose we could get around that by tweaking the > parameter to take the size to reserve instead of a bool and pass a dummy > value (i.e., negative) to avoid logging the chunk at all. A little ugly, > but I could live with it. I don't think that having an extra few hundred bytes of overhead in the reservation is going to be noticable by anyone. I'd just ignore the problem (as I did when suggesting this). > > > STATIC uint > > > @@ -494,9 +493,14 @@ xfs_calc_symlink_reservation( > > > * the agi hash list and counters: sector size > > > * the on disk inode before ours in the agi hash list: inode cluster size > > > * the inode chunk is marked stale (headers only) > > > - * the inode btree: max depth * blocksize > > > - * the allocation btrees: 2 trees * (max depth - 1) * block size > > > + * the inode btree > > > * the finobt (record insertion, removal or modification) > > > + * > > > + * Note that the allocfree res. for the inode chunk itself is not included > > > + * because the extent free occurs after a transaction roll. We could take the > > > + * maximum of the pre/post roll operations, but the pre-roll reservation already > > > + * includes at least one allocfree res. for the inobt and is thus guaranteed to > > > + * be larger. > > > */ > > > STATIC uint > > > xfs_calc_ifree_reservation( > > > @@ -508,10 +512,8 @@ xfs_calc_ifree_reservation( > > > xfs_calc_iunlink_remove_reservation(mp) + > > > xfs_calc_buf_res(1, 0) + > > > xfs_calc_buf_res(mp->m_ialloc_blks, 0) + > > > - xfs_calc_buf_res(mp->m_in_maxlevels, XFS_FSB_TO_B(mp, 1)) + > > > - xfs_calc_buf_res(xfs_allocfree_log_count(mp, 1), > > > - XFS_FSB_TO_B(mp, 1)) + > > > - xfs_calc_finobt_res(mp, 0, 1); > > > + xfs_calc_inobt_res(mp) + > > > + xfs_calc_finobt_res(mp); > > > } > > > > ..... > > xfs_calc_iunlink_remove_reservation(mp) + > > xfs_calc_buf_res(1, 0) + > > xfs_calc_inode_chunk_res(mp, false) + > > xfs_calc_inobt_res(mp) + > > xfs_calc_finobt_res(mp); > > > > This covers the inode chunk invalidation, but also adds the allocfree > res. for the chunk free where we technically don't need it (because the > free is deferred, re: the comment above). > > I guess I'm fine with just adding that one, but I'd update the comment > above to point out that it's technically unecessary. Hm? *nod* Though with sparse inodes, we might be freeing multiple extents, right? which means we probably need all the allocfree reservations we can get.... > > This seems to make more sense to me - it's clear what the individual > > components of the reservation are, and we can ensure that the > > individual components have the necessary reservation independently > > of the overall reservations that need them. > > > > I agree in principle. I think the underlying helpers (and pushing down > some of the associated documentation) help clearly declare the intent of > the reservations, particularly when we include multiple factors of a > single reservation and/or have situations where we don't technically > have a definition of worst case, but want to define something logically > reasonable (like the whole allocfree per inode tree thing). Yup, that's pretty much what I was thinking. Cheers, Dave. -- Dave Chinner david@fromorbit.com