From patchwork Thu Apr 27 22:45:21 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: "Darrick J. Wong" X-Patchwork-Id: 13225790 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 EBE5EC77B61 for ; Thu, 27 Apr 2023 22:45:32 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1344237AbjD0Wpb (ORCPT ); Thu, 27 Apr 2023 18:45:31 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:51560 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1344308AbjD0Wpa (ORCPT ); Thu, 27 Apr 2023 18:45:30 -0400 Received: from dfw.source.kernel.org (dfw.source.kernel.org [139.178.84.217]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 840AB3C28 for ; Thu, 27 Apr 2023 15:45:23 -0700 (PDT) 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 05E8963FE4 for ; Thu, 27 Apr 2023 22:45:23 +0000 (UTC) Received: by smtp.kernel.org (Postfix) with ESMTPSA id 61778C433D2; Thu, 27 Apr 2023 22:45:22 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=kernel.org; s=k20201202; t=1682635522; bh=Q5W1rF60XpnNXeFoZYkq4BV4AhhVk01b7fskXEVSqoc=; h=Date:From:To:Cc:Subject:From; b=R6hA0eiMJguLQ+5HxVHIHHje3qFSMIYu4jqHVYSJddwIpI0jLLTf6je+W+Va8/tgh n/eFHhbUTpxKjiIhYfZ+V0Tyra5a/zL3Qbx5bXYTf/6K8H694PeIkfCFXkToiuEUE5 JqJqM9AXMjBQSdhIT4E6HhPauvCdHGYqQ1QVre/K6+8wHscV+zdKC6kj0VvMMk6Tr0 o71IwXuOqDiiBbHvfYivHX4jqmf2OugSJYqcUypgn+4YUC8P9IIHvBP7JC9KMiXBhQ yIYd24yjSS0xkd35VJGoeElp9rWoIoB6DgQfKNFKeXcyl/UaKNzNMxn1Y/iqZyFGjA 6aRbogVdowZtw== Date: Thu, 27 Apr 2023 15:45:21 -0700 From: "Darrick J. Wong" To: Carlos Maiolino Cc: xfs Subject: [PATCH] xfs_repair: estimate per-AG btree slack better Message-ID: <20230427224521.GD59213@frogsfrogsfrogs> MIME-Version: 1.0 Content-Disposition: inline Precedence: bulk List-ID: X-Mailing-List: linux-xfs@vger.kernel.org From: Darrick J. Wong The slack calculation for per-AG btrees is a bit inaccurate because it only disables slack space in the new btrees when the amount of free space in the AG (not counting the btrees) is less than 3/32ths of the AG. In other words, it assumes that the btrees will fit in less than 9 percent of the space. However, there's one scenario where this goes wrong -- if the rmapbt consumes a significant portion of the AG space. Say a filesystem is hosting a VM image farm that starts with perfectly shared images. As time goes by, random writes to those images will slowly cause the rmapbt to increase in size as blocks within those images get COWed. Suppose that the rmapbt now consumes 20% of the space in the AG, that the AG is nearly full, and that the blocks in the old rmapbt are mostly full. At the start of phase5_func, mk_incore_fstree will return that num_freeblocks is ~20% of the AG size. Hence the slack calculation will conclude that there's plenty of space in the AG and new btrees will be built with 25% slack in the blocks. If the size of these new expanded btrees is larger than the free space in the AG, repair will fail to allocate btree blocks and fail, causing severe filesystem damage. To combat this, estimate the worst case size of the AG btrees given the number of records we intend to put in them, subtract that worst case figure from num_freeblocks, and feed that to bulkload_estimate_ag_slack. This results in tighter packing of new btree blocks when space is dear, and hopefully fewer problems. This /can/ be reproduced with generic/333 if you hack it to keep COWing blocks until the filesystem is totally out of space, even if reflink has long since refused to share more blocks. Signed-off-by: Darrick J. Wong Reviewed-by: Carlos Maiolino --- libxfs/libxfs_api_defs.h | 4 ++ repair/agbtree.c | 90 ++++++++++++++++++++++++++++++++++++++++------ repair/agbtree.h | 3 ++ repair/phase5.c | 18 +++++++-- repair/rmap.c | 44 ++++++++++++++++++++++ repair/rmap.h | 3 ++ 6 files changed, 146 insertions(+), 16 deletions(-) diff --git a/libxfs/libxfs_api_defs.h b/libxfs/libxfs_api_defs.h index f8efcce7..d973e300 100644 --- a/libxfs/libxfs_api_defs.h +++ b/libxfs/libxfs_api_defs.h @@ -25,6 +25,7 @@ #define xfs_ag_resv_free libxfs_ag_resv_free #define xfs_alloc_ag_max_usable libxfs_alloc_ag_max_usable +#define xfs_allocbt_calc_size libxfs_allocbt_calc_size #define xfs_allocbt_maxlevels_ondisk libxfs_allocbt_maxlevels_ondisk #define xfs_allocbt_maxrecs libxfs_allocbt_maxrecs #define xfs_allocbt_stage_cursor libxfs_allocbt_stage_cursor @@ -115,6 +116,7 @@ #define xfs_highbit32 libxfs_highbit32 #define xfs_highbit64 libxfs_highbit64 #define xfs_ialloc_calc_rootino libxfs_ialloc_calc_rootino +#define xfs_iallocbt_calc_size libxfs_iallocbt_calc_size #define xfs_iallocbt_maxlevels_ondisk libxfs_iallocbt_maxlevels_ondisk #define xfs_ialloc_read_agi libxfs_ialloc_read_agi #define xfs_idata_realloc libxfs_idata_realloc @@ -146,6 +148,7 @@ #define xfs_read_agf libxfs_read_agf #define xfs_refc_block libxfs_refc_block #define xfs_refcountbt_calc_reserves libxfs_refcountbt_calc_reserves +#define xfs_refcountbt_calc_size libxfs_refcountbt_calc_size #define xfs_refcountbt_init_cursor libxfs_refcountbt_init_cursor #define xfs_refcountbt_maxlevels_ondisk libxfs_refcountbt_maxlevels_ondisk #define xfs_refcountbt_maxrecs libxfs_refcountbt_maxrecs @@ -155,6 +158,7 @@ #define xfs_rmap_alloc libxfs_rmap_alloc #define xfs_rmapbt_calc_reserves libxfs_rmapbt_calc_reserves +#define xfs_rmapbt_calc_size libxfs_rmapbt_calc_size #define xfs_rmapbt_init_cursor libxfs_rmapbt_init_cursor #define xfs_rmapbt_maxlevels_ondisk libxfs_rmapbt_maxlevels_ondisk #define xfs_rmapbt_maxrecs libxfs_rmapbt_maxrecs diff --git a/repair/agbtree.c b/repair/agbtree.c index ef001803..d5e441a3 100644 --- a/repair/agbtree.c +++ b/repair/agbtree.c @@ -17,13 +17,13 @@ static void init_rebuild( struct repair_ctx *sc, const struct xfs_owner_info *oinfo, - xfs_agblock_t free_space, + xfs_agblock_t est_agfreeblocks, struct bt_rebuild *btr) { memset(btr, 0, sizeof(struct bt_rebuild)); bulkload_init_ag(&btr->newbt, sc, oinfo); - bulkload_estimate_ag_slack(sc, &btr->bload, free_space); + bulkload_estimate_ag_slack(sc, &btr->bload, est_agfreeblocks); } /* @@ -227,7 +227,7 @@ void init_freespace_cursors( struct repair_ctx *sc, struct xfs_perag *pag, - unsigned int free_space, + unsigned int est_agfreeblocks, unsigned int *nr_extents, int *extra_blocks, struct bt_rebuild *btr_bno, @@ -239,8 +239,8 @@ init_freespace_cursors( agfl_goal = libxfs_alloc_min_freelist(sc->mp, NULL); - init_rebuild(sc, &XFS_RMAP_OINFO_AG, free_space, btr_bno); - init_rebuild(sc, &XFS_RMAP_OINFO_AG, free_space, btr_cnt); + init_rebuild(sc, &XFS_RMAP_OINFO_AG, est_agfreeblocks, btr_bno); + init_rebuild(sc, &XFS_RMAP_OINFO_AG, est_agfreeblocks, btr_cnt); btr_bno->cur = libxfs_allocbt_stage_cursor(sc->mp, &btr_bno->newbt.afake, pag, XFS_BTNUM_BNO); @@ -439,7 +439,7 @@ void init_ino_cursors( struct repair_ctx *sc, struct xfs_perag *pag, - unsigned int free_space, + unsigned int est_agfreeblocks, uint64_t *num_inos, uint64_t *num_free_inos, struct bt_rebuild *btr_ino, @@ -453,7 +453,7 @@ init_ino_cursors( int error; finobt = xfs_has_finobt(sc->mp); - init_rebuild(sc, &XFS_RMAP_OINFO_INOBT, free_space, btr_ino); + init_rebuild(sc, &XFS_RMAP_OINFO_INOBT, est_agfreeblocks, btr_ino); /* Compute inode statistics. */ *num_free_inos = 0; @@ -506,7 +506,7 @@ _("Unable to compute inode btree geometry, error %d.\n"), error); if (!finobt) return; - init_rebuild(sc, &XFS_RMAP_OINFO_INOBT, free_space, btr_fino); + init_rebuild(sc, &XFS_RMAP_OINFO_INOBT, est_agfreeblocks, btr_fino); btr_fino->cur = libxfs_inobt_stage_cursor(pag, &btr_fino->newbt.afake, XFS_BTNUM_FINO); @@ -577,7 +577,7 @@ void init_rmapbt_cursor( struct repair_ctx *sc, struct xfs_perag *pag, - unsigned int free_space, + unsigned int est_agfreeblocks, struct bt_rebuild *btr) { xfs_agnumber_t agno = pag->pag_agno; @@ -586,7 +586,7 @@ init_rmapbt_cursor( if (!xfs_has_rmapbt(sc->mp)) return; - init_rebuild(sc, &XFS_RMAP_OINFO_AG, free_space, btr); + init_rebuild(sc, &XFS_RMAP_OINFO_AG, est_agfreeblocks, btr); btr->cur = libxfs_rmapbt_stage_cursor(sc->mp, &btr->newbt.afake, pag); btr->bload.get_record = get_rmapbt_record; @@ -648,7 +648,7 @@ void init_refc_cursor( struct repair_ctx *sc, struct xfs_perag *pag, - unsigned int free_space, + unsigned int est_agfreeblocks, struct bt_rebuild *btr) { xfs_agnumber_t agno = pag->pag_agno; @@ -657,7 +657,7 @@ init_refc_cursor( if (!xfs_has_reflink(sc->mp)) return; - init_rebuild(sc, &XFS_RMAP_OINFO_REFC, free_space, btr); + init_rebuild(sc, &XFS_RMAP_OINFO_REFC, est_agfreeblocks, btr); btr->cur = libxfs_refcountbt_stage_cursor(sc->mp, &btr->newbt.afake, pag); @@ -698,3 +698,69 @@ _("Error %d while creating refcount btree for AG %u.\n"), error, agno); libxfs_btree_del_cursor(btr->cur, 0); free_slab_cursor(&btr->slab_cursor); } + +static xfs_extlen_t +estimate_allocbt_blocks( + struct xfs_perag *pag, + unsigned int nr_extents) +{ + return libxfs_allocbt_calc_size(pag->pag_mount, nr_extents) * 2; +} + +static xfs_extlen_t +estimate_inobt_blocks( + struct xfs_perag *pag) +{ + struct ino_tree_node *ino_rec; + xfs_agnumber_t agno = pag->pag_agno; + unsigned int ino_recs = 0; + unsigned int fino_recs = 0; + xfs_extlen_t ret; + + for (ino_rec = findfirst_inode_rec(agno); + ino_rec != NULL; + ino_rec = next_ino_rec(ino_rec)) { + unsigned int rec_nfinos = 0; + int i; + + for (i = 0; i < XFS_INODES_PER_CHUNK; i++) { + ASSERT(is_inode_confirmed(ino_rec, i)); + /* + * sparse inodes are not factored into superblock (free) + * inode counts + */ + if (is_inode_sparse(ino_rec, i)) + continue; + if (is_inode_free(ino_rec, i)) + rec_nfinos++; + } + + ino_recs++; + + /* finobt only considers records with free inodes */ + if (rec_nfinos) + fino_recs++; + } + + ret = libxfs_iallocbt_calc_size(pag->pag_mount, ino_recs); + if (xfs_has_finobt(pag->pag_mount)) + ret += libxfs_iallocbt_calc_size(pag->pag_mount, fino_recs); + return ret; + +} + +/* Estimate the size of the per-AG btrees. */ +xfs_extlen_t +estimate_agbtree_blocks( + struct xfs_perag *pag, + unsigned int free_extents) +{ + unsigned int ret = 0; + + ret += estimate_allocbt_blocks(pag, free_extents); + ret += estimate_inobt_blocks(pag); + ret += estimate_rmapbt_blocks(pag); + ret += estimate_refcountbt_blocks(pag); + + return ret; +} diff --git a/repair/agbtree.h b/repair/agbtree.h index 84f7083d..714d8e68 100644 --- a/repair/agbtree.h +++ b/repair/agbtree.h @@ -59,4 +59,7 @@ void init_refc_cursor(struct repair_ctx *sc, struct xfs_perag *pag, void build_refcount_tree(struct repair_ctx *sc, xfs_agnumber_t agno, struct bt_rebuild *btr); +xfs_extlen_t estimate_agbtree_blocks(struct xfs_perag *pag, + unsigned int free_extents); + #endif /* __XFS_REPAIR_AG_BTREE_H__ */ diff --git a/repair/phase5.c b/repair/phase5.c index b04912d8..0d14c354 100644 --- a/repair/phase5.c +++ b/repair/phase5.c @@ -447,6 +447,8 @@ phase5_func( int extra_blocks = 0; uint num_freeblocks; xfs_agblock_t num_extents; + unsigned int est_agfreeblocks = 0; + unsigned int total_btblocks; if (verbose) do_log(_(" - agno = %d\n"), agno); @@ -474,12 +476,20 @@ _("unable to rebuild AG %u. Not enough free space in on-disk AG.\n"), agno); } - init_ino_cursors(&sc, pag, num_freeblocks, &sb_icount_ag[agno], + /* + * Estimate the number of free blocks in this AG after rebuilding + * all btrees. + */ + total_btblocks = estimate_agbtree_blocks(pag, num_extents); + if (num_freeblocks > total_btblocks) + est_agfreeblocks = num_freeblocks - total_btblocks; + + init_ino_cursors(&sc, pag, est_agfreeblocks, &sb_icount_ag[agno], &sb_ifree_ag[agno], &btr_ino, &btr_fino); - init_rmapbt_cursor(&sc, pag, num_freeblocks, &btr_rmap); + init_rmapbt_cursor(&sc, pag, est_agfreeblocks, &btr_rmap); - init_refc_cursor(&sc, pag, num_freeblocks, &btr_refc); + init_refc_cursor(&sc, pag, est_agfreeblocks, &btr_refc); num_extents = count_bno_extents_blocks(agno, &num_freeblocks); /* @@ -507,7 +517,7 @@ _("unable to rebuild AG %u. Not enough free space in on-disk AG.\n"), /* * track blocks that we might really lose */ - init_freespace_cursors(&sc, pag, num_freeblocks, &num_extents, + init_freespace_cursors(&sc, pag, est_agfreeblocks, &num_extents, &extra_blocks, &btr_bno, &btr_cnt); /* diff --git a/repair/rmap.c b/repair/rmap.c index 9013daa2..6bb77e08 100644 --- a/repair/rmap.c +++ b/repair/rmap.c @@ -1531,3 +1531,47 @@ rmap_store_agflcount( ag_rmaps[agno].ar_flcount = count; } + +/* Estimate the size of the ondisk rmapbt from the incore data. */ +xfs_extlen_t +estimate_rmapbt_blocks( + struct xfs_perag *pag) +{ + struct xfs_mount *mp = pag->pag_mount; + struct xfs_ag_rmap *x; + unsigned long long nr_recs = 0; + + if (!rmap_needs_work(mp) || !xfs_has_rmapbt(mp)) + return 0; + + /* + * Overestimate the amount of space needed by pretending that every + * record in the incore slab will become rmapbt records. + */ + x = &ag_rmaps[pag->pag_agno]; + if (x->ar_rmaps) + nr_recs += slab_count(x->ar_rmaps); + if (x->ar_raw_rmaps) + nr_recs += slab_count(x->ar_raw_rmaps); + + return libxfs_rmapbt_calc_size(mp, nr_recs); +} + +/* Estimate the size of the ondisk refcountbt from the incore data. */ +xfs_extlen_t +estimate_refcountbt_blocks( + struct xfs_perag *pag) +{ + struct xfs_mount *mp = pag->pag_mount; + struct xfs_ag_rmap *x; + + if (!rmap_needs_work(mp) || !xfs_has_reflink(mp)) + return 0; + + x = &ag_rmaps[pag->pag_agno]; + if (!x->ar_refcount_items) + return 0; + + return libxfs_refcountbt_calc_size(mp, + slab_count(x->ar_refcount_items)); +} diff --git a/repair/rmap.h b/repair/rmap.h index 8d176cb3..6004e9f6 100644 --- a/repair/rmap.h +++ b/repair/rmap.h @@ -48,4 +48,7 @@ extern int fix_inode_reflink_flags(struct xfs_mount *, xfs_agnumber_t); extern void fix_freelist(struct xfs_mount *, xfs_agnumber_t, bool); extern void rmap_store_agflcount(struct xfs_mount *, xfs_agnumber_t, int); +xfs_extlen_t estimate_rmapbt_blocks(struct xfs_perag *pag); +xfs_extlen_t estimate_refcountbt_blocks(struct xfs_perag *pag); + #endif /* RMAP_H_ */