Re: [PATCH 06/12] xfs_repair: create a new class of btree rebuild cursors

["Darrick J. Wong" <darrick.wong@oracle.com>]

On Wed, Jun 17, 2020 at 08:10:01AM -0400, Brian Foster wrote:
> On Mon, Jun 01, 2020 at 09:27:31PM -0700, Darrick J. Wong wrote:
> > From: Darrick J. Wong <darrick.wong@oracle.com>
> > 
> > Create some new support structures and functions to assist phase5 in
> > using the btree bulk loader to reconstruct metadata btrees.  This is the
> > first step in removing the open-coded AG btree rebuilding code.
> > 
> > Note: The code in this patch will not be used anywhere until the next
> > patch, so warnings about unused symbols are expected.
> > 
> > Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
> > ---
> 
> I still find it odd to include the phase5.c changes in this patch when
> it amounts to the addition of a single unused parameter, but I'll defer
> to the maintainer on that. Otherwise LGTM:
> 
> Reviewed-by: Brian Foster <bfoster@redhat.com>
> 
> >  repair/Makefile   |    4 +
> >  repair/agbtree.c  |  152 +++++++++++++++++++++++++++++++++++++++++++++++++++++
> >  repair/agbtree.h  |   29 ++++++++++
> >  repair/bulkload.c |   37 +++++++++++++
> >  repair/bulkload.h |    2 +
> >  repair/phase5.c   |   41 ++++++++------
> >  6 files changed, 244 insertions(+), 21 deletions(-)
> >  create mode 100644 repair/agbtree.c
> >  create mode 100644 repair/agbtree.h
> > 
> > 
> > diff --git a/repair/Makefile b/repair/Makefile
> > index 62d84bbf..f6a6e3f9 100644
> > --- a/repair/Makefile
> > +++ b/repair/Makefile
> > @@ -9,11 +9,11 @@ LSRCFILES = README
> >  
> >  LTCOMMAND = xfs_repair
> >  
> > -HFILES = agheader.h attr_repair.h avl.h bulkload.h bmap.h btree.h \
> > +HFILES = agheader.h agbtree.h attr_repair.h avl.h bulkload.h bmap.h btree.h \
> >  	da_util.h dinode.h dir2.h err_protos.h globals.h incore.h protos.h \
> >  	rt.h progress.h scan.h versions.h prefetch.h rmap.h slab.h threads.h
> >  
> > -CFILES = agheader.c attr_repair.c avl.c bulkload.c bmap.c btree.c \
> > +CFILES = agheader.c agbtree.c attr_repair.c avl.c bulkload.c bmap.c btree.c \
> >  	da_util.c dino_chunks.c dinode.c dir2.c globals.c incore.c \
> >  	incore_bmc.c init.c incore_ext.c incore_ino.c phase1.c \
> >  	phase2.c phase3.c phase4.c phase5.c phase6.c phase7.c \
> > diff --git a/repair/agbtree.c b/repair/agbtree.c
> > new file mode 100644
> > index 00000000..e4179a44
> > --- /dev/null
> > +++ b/repair/agbtree.c
> > @@ -0,0 +1,152 @@
> > +// SPDX-License-Identifier: GPL-2.0-or-later
> > +/*
> > + * Copyright (C) 2020 Oracle.  All Rights Reserved.
> > + * Author: Darrick J. Wong <darrick.wong@oracle.com>
> > + */
> > +#include <libxfs.h>
> > +#include "err_protos.h"
> > +#include "slab.h"
> > +#include "rmap.h"
> > +#include "incore.h"
> > +#include "bulkload.h"
> > +#include "agbtree.h"
> > +
> > +/* Initialize a btree rebuild context. */
> > +static void
> > +init_rebuild(
> > +	struct repair_ctx		*sc,
> > +	const struct xfs_owner_info	*oinfo,
> > +	xfs_agblock_t			free_space,
> > +	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);
> > +}
> > +
> > +/*
> > + * Update this free space record to reflect the blocks we stole from the
> > + * beginning of the record.
> > + */
> > +static void
> > +consume_freespace(
> > +	xfs_agnumber_t		agno,
> > +	struct extent_tree_node	*ext_ptr,
> > +	uint32_t		len)
> > +{
> > +	struct extent_tree_node	*bno_ext_ptr;
> > +	xfs_agblock_t		new_start = ext_ptr->ex_startblock + len;
> > +	xfs_extlen_t		new_len = ext_ptr->ex_blockcount - len;
> > +
> > +	/* Delete the used-up extent from both extent trees. */
> > +#ifdef XR_BLD_FREE_TRACE
> > +	fprintf(stderr, "releasing extent: %u [%u %u]\n", agno,
> > +			ext_ptr->ex_startblock, ext_ptr->ex_blockcount);
> > +#endif
> > +	bno_ext_ptr = find_bno_extent(agno, ext_ptr->ex_startblock);
> > +	ASSERT(bno_ext_ptr != NULL);
> > +	get_bno_extent(agno, bno_ext_ptr);
> > +	release_extent_tree_node(bno_ext_ptr);
> > +
> > +	ext_ptr = get_bcnt_extent(agno, ext_ptr->ex_startblock,
> > +			ext_ptr->ex_blockcount);
> > +	release_extent_tree_node(ext_ptr);
> > +
> > +	/*
> > +	 * If we only used part of this last extent, then we must reinsert the
> > +	 * extent to maintain proper sorting order.
> > +	 */
> > +	if (new_len > 0) {
> > +		add_bno_extent(agno, new_start, new_len);
> > +		add_bcnt_extent(agno, new_start, new_len);
> > +	}
> > +}
> > +
> > +/* Reserve blocks for the new btree. */
> > +static void
> > +reserve_btblocks(
> > +	struct xfs_mount	*mp,
> > +	xfs_agnumber_t		agno,
> > +	struct bt_rebuild	*btr,
> > +	uint32_t		nr_blocks)
> > +{
> > +	struct extent_tree_node	*ext_ptr;
> > +	uint32_t		blocks_allocated = 0;
> > +	uint32_t		len;
> > +	int			error;
> > +
> > +	while (blocks_allocated < nr_blocks)  {
> > +		xfs_fsblock_t	fsbno;
> > +
> > +		/*
> > +		 * Grab the smallest extent and use it up, then get the
> > +		 * next smallest.  This mimics the init_*_cursor code.
> > +		 */
> > +		ext_ptr = findfirst_bcnt_extent(agno);
> > +		if (!ext_ptr)
> > +			do_error(
> > +_("error - not enough free space in filesystem\n"));
> > +
> > +		/* Use up the extent we've got. */
> > +		len = min(ext_ptr->ex_blockcount, nr_blocks - blocks_allocated);
> > +		fsbno = XFS_AGB_TO_FSB(mp, agno, ext_ptr->ex_startblock);
> > +		error = bulkload_add_blocks(&btr->newbt, fsbno, len);
> > +		if (error)
> > +			do_error(_("could not set up btree reservation: %s\n"),
> > +				strerror(-error));
> > +
> > +		error = rmap_add_ag_rec(mp, agno, ext_ptr->ex_startblock, len,
> > +				btr->newbt.oinfo.oi_owner);
> > +		if (error)
> > +			do_error(_("could not set up btree rmaps: %s\n"),
> > +				strerror(-error));
> > +
> > +		consume_freespace(agno, ext_ptr, len);
> > +		blocks_allocated += len;
> > +	}
> > +#ifdef XR_BLD_FREE_TRACE
> > +	fprintf(stderr, "blocks_allocated = %d\n",
> > +		blocks_allocated);
> > +#endif
> > +}
> > +
> > +/* Feed one of the new btree blocks to the bulk loader. */
> > +static int
> > +rebuild_claim_block(
> > +	struct xfs_btree_cur	*cur,
> > +	union xfs_btree_ptr	*ptr,
> > +	void			*priv)
> > +{
> > +	struct bt_rebuild	*btr = priv;
> > +
> > +	return bulkload_claim_block(cur, &btr->newbt, ptr);
> > +}
> > +
> > +/*
> > + * Scoop up leftovers from a rebuild cursor for later freeing, then free the
> > + * rebuild context.
> > + */
> > +void
> > +finish_rebuild(
> > +	struct xfs_mount	*mp,
> > +	struct bt_rebuild	*btr,
> > +	struct xfs_slab		*lost_fsb)
> > +{
> > +	struct bulkload_resv	*resv, *n;
> > +
> > +	for_each_bulkload_reservation(&btr->newbt, resv, n) {
> > +		while (resv->used < resv->len) {
> > +			xfs_fsblock_t	fsb = resv->fsbno + resv->used;
> > +			int		error;
> > +
> > +			error = slab_add(lost_fsb, &fsb);
> > +			if (error)
> > +				do_error(
> > +_("Insufficient memory saving lost blocks.\n"));
> > +			resv->used++;
> > +		}
> > +	}
> > +
> > +	bulkload_destroy(&btr->newbt, 0);
> > +}
> > diff --git a/repair/agbtree.h b/repair/agbtree.h
> > new file mode 100644
> > index 00000000..50ea3c60
> > --- /dev/null
> > +++ b/repair/agbtree.h
> > @@ -0,0 +1,29 @@
> > +/* SPDX-License-Identifier: GPL-2.0-or-later */
> > +/*
> > + * Copyright (C) 2020 Oracle.  All Rights Reserved.
> > + * Author: Darrick J. Wong <darrick.wong@oracle.com>
> > + */
> > +#ifndef __XFS_REPAIR_AG_BTREE_H__
> > +#define __XFS_REPAIR_AG_BTREE_H__
> > +
> > +/* Context for rebuilding a per-AG btree. */
> > +struct bt_rebuild {
> > +	/* Fake root for staging and space preallocations. */
> > +	struct bulkload	newbt;
> > +
> > +	/* Geometry of the new btree. */
> > +	struct xfs_btree_bload	bload;
> > +
> > +	/* Staging btree cursor for the new tree. */
> > +	struct xfs_btree_cur	*cur;
> > +
> > +	/* Tree-specific data. */
> > +	union {
> > +		struct xfs_slab_cursor	*slab_cursor;
> > +	};
> > +};
> > +
> > +void finish_rebuild(struct xfs_mount *mp, struct bt_rebuild *btr,
> > +		struct xfs_slab *lost_fsb);
> > +
> > +#endif /* __XFS_REPAIR_AG_BTREE_H__ */
> > diff --git a/repair/bulkload.c b/repair/bulkload.c
> > index 4c69fe0d..9a6ca0c2 100644
> > --- a/repair/bulkload.c
> > +++ b/repair/bulkload.c
> > @@ -95,3 +95,40 @@ bulkload_claim_block(
> >  		ptr->s = cpu_to_be32(XFS_FSB_TO_AGBNO(cur->bc_mp, fsb));
> >  	return 0;
> >  }
> > +
> > +/*
> > + * Estimate proper slack values for a btree that's being reloaded.
> > + *
> > + * Under most circumstances, we'll take whatever default loading value the
> > + * btree bulk loading code calculates for us.  However, there are some
> > + * exceptions to this rule:
> > + *
> > + * (1) If someone turned one of the debug knobs.
> > + * (2) The AG has less than ~9% space free.
> > + *
> > + * Note that we actually use 3/32 for the comparison to avoid division.
> > + */
> > +void
> > +bulkload_estimate_ag_slack(
> > +	struct repair_ctx	*sc,
> > +	struct xfs_btree_bload	*bload,
> > +	unsigned int		free)
> > +{
> > +	/*
> > +	 * The global values are set to -1 (i.e. take the bload defaults)
> > +	 * unless someone has set them otherwise, so we just pull the values
> > +	 * here.
> > +	 */
> > +	bload->leaf_slack = bload_leaf_slack;
> > +	bload->node_slack = bload_node_slack;
> > +
> > +	/* No further changes if there's more than 3/32ths space left. */
> > +	if (free >= ((sc->mp->m_sb.sb_agblocks * 3) >> 5))
> > +		return;
> > +
> > +	/* We're low on space; load the btrees as tightly as possible. */
> > +	if (bload->leaf_slack < 0)
> > +		bload->leaf_slack = 0;
> > +	if (bload->node_slack < 0)
> > +		bload->node_slack = 0;

Heh.  It turns out that this caused infrequent warnings in
generic/333 because adding the extra rmap records for the AG btrees at
the end of phase 5 could cause enough rmapbt splits such that we
wouldn't have enough space left in the AG to satisfy the per-AG
reservation at the next mount.

I /think/ the solution here is to set the slack values to 2 (instead of
zero) like we did in xfs_repair before this patch.

--D

> > +}
> > diff --git a/repair/bulkload.h b/repair/bulkload.h
> > index 79f81cb0..01f67279 100644
> > --- a/repair/bulkload.h
> > +++ b/repair/bulkload.h
> > @@ -53,5 +53,7 @@ int bulkload_add_blocks(struct bulkload *bkl, xfs_fsblock_t fsbno,
> >  void bulkload_destroy(struct bulkload *bkl, int error);
> >  int bulkload_claim_block(struct xfs_btree_cur *cur, struct bulkload *bkl,
> >  		union xfs_btree_ptr *ptr);
> > +void bulkload_estimate_ag_slack(struct repair_ctx *sc,
> > +		struct xfs_btree_bload *bload, unsigned int free);
> >  
> >  #endif /* __XFS_REPAIR_BULKLOAD_H__ */
> > diff --git a/repair/phase5.c b/repair/phase5.c
> > index 75c480fd..8175aa6f 100644
> > --- a/repair/phase5.c
> > +++ b/repair/phase5.c
> > @@ -18,6 +18,8 @@
> >  #include "progress.h"
> >  #include "slab.h"
> >  #include "rmap.h"
> > +#include "bulkload.h"
> > +#include "agbtree.h"
> >  
> >  /*
> >   * we maintain the current slice (path from root to leaf)
> > @@ -2288,28 +2290,29 @@ keep_fsinos(xfs_mount_t *mp)
> >  
> >  static void
> >  phase5_func(
> > -	xfs_mount_t	*mp,
> > -	xfs_agnumber_t	agno,
> > -	struct xfs_slab	*lost_fsb)
> > +	struct xfs_mount	*mp,
> > +	xfs_agnumber_t		agno,
> > +	struct xfs_slab		*lost_fsb)
> >  {
> > -	uint64_t	num_inos;
> > -	uint64_t	num_free_inos;
> > -	uint64_t	finobt_num_inos;
> > -	uint64_t	finobt_num_free_inos;
> > -	bt_status_t	bno_btree_curs;
> > -	bt_status_t	bcnt_btree_curs;
> > -	bt_status_t	ino_btree_curs;
> > -	bt_status_t	fino_btree_curs;
> > -	bt_status_t	rmap_btree_curs;
> > -	bt_status_t	refcnt_btree_curs;
> > -	int		extra_blocks = 0;
> > -	uint		num_freeblocks;
> > -	xfs_extlen_t	freeblks1;
> > +	struct repair_ctx	sc = { .mp = mp, };
> > +	struct agi_stat		agi_stat = {0,};
> > +	uint64_t		num_inos;
> > +	uint64_t		num_free_inos;
> > +	uint64_t		finobt_num_inos;
> > +	uint64_t		finobt_num_free_inos;
> > +	bt_status_t		bno_btree_curs;
> > +	bt_status_t		bcnt_btree_curs;
> > +	bt_status_t		ino_btree_curs;
> > +	bt_status_t		fino_btree_curs;
> > +	bt_status_t		rmap_btree_curs;
> > +	bt_status_t		refcnt_btree_curs;
> > +	int			extra_blocks = 0;
> > +	uint			num_freeblocks;
> > +	xfs_extlen_t		freeblks1;
> >  #ifdef DEBUG
> > -	xfs_extlen_t	freeblks2;
> > +	xfs_extlen_t		freeblks2;
> >  #endif
> > -	xfs_agblock_t	num_extents;
> > -	struct agi_stat	agi_stat = {0,};
> > +	xfs_agblock_t		num_extents;
> >  
> >  	if (verbose)
> >  		do_log(_("        - agno = %d\n"), agno);
> > 
>