fscounters.c | Explore in Territory

// SPDX-License-Identifier: GPL-2.0-or-later
/*
 * Copyright (C) 2019-2023 Oracle.  All Rights Reserved.
 * Author: Darrick J. Wong <[email protected]>
 */
#include "xfs.h"
#include "xfs_fs.h"
#include "xfs_shared.h"
#include "xfs_format.h"
#include "xfs_trans_resv.h"
#include "xfs_log_format.h"
#include "xfs_trans.h"
#include "xfs_mount.h"
#include "xfs_alloc.h"
#include "xfs_ialloc.h"
#include "xfs_health.h"
#include "xfs_btree.h"
#include "xfs_ag.h"
#include "xfs_rtbitmap.h"
#include "xfs_inode.h"
#include "xfs_icache.h"
#include "scrub/scrub.h"
#include "scrub/common.h"
#include "scrub/trace.h"
#include "scrub/fscounters.h"

/*
 * FS Summary Counters
 * ===================
 *
 * The basics of filesystem summary counter checking are that we iterate the
 * AGs counting the number of free blocks, free space btree blocks, per-AG
 * reservations, inodes, delayed allocation reservations, and free inodes.
 * Then we compare what we computed against the in-core counters.
 *
 * However, the reality is that summary counters are a tricky beast to check.
 * While we /could/ freeze the filesystem and scramble around the AGs counting
 * the free blocks, in practice we prefer not do that for a scan because
 * freezing is costly.  To get around this, we added a per-cpu counter of the
 * delalloc reservations so that we can rotor around the AGs relatively
 * quickly, and we allow the counts to be slightly off because we're not taking
 * any locks while we do this.
 *
 * So the first thing we do is warm up the buffer cache in the setup routine by
 * walking all the AGs to make sure the incore per-AG structure has been
 * initialized.  The expected value calculation then iterates the incore per-AG
 * structures as quickly as it can.  We snapshot the percpu counters before and
 * after this operation and use the difference in counter values to guess at
 * our tolerance for mismatch between expected and actual counter values.
 */

/*
 * Since the expected value computation is lockless but only browses incore
 * values, the percpu counters should be fairly close to each other.  However,
 * we'll allow ourselves to be off by at least this (arbitrary) amount.
 */
#define XCHK_FSCOUNT_MIN_VARIANCE …

/*
 * Make sure the per-AG structure has been initialized from the on-disk header
 * contents and trust that the incore counters match the ondisk counters.  (The
 * AGF and AGI scrubbers check them, and a normal xfs_scrub run checks the
 * summary counters after checking all AG headers).  Do this from the setup
 * function so that the inner AG aggregation loop runs as quickly as possible.
 *
 * This function runs during the setup phase /before/ we start checking any
 * metadata.
 */
STATIC int
xchk_fscount_warmup(
	struct xfs_scrub	*sc)
{ … }

static inline int
xchk_fsfreeze(
	struct xfs_scrub	*sc)
{ … }

static inline int
xchk_fsthaw(
	struct xfs_scrub	*sc)
{ … }

/*
 * We couldn't stabilize the filesystem long enough to sample all the variables
 * that comprise the summary counters and compare them to the percpu counters.
 * We need to disable all writer threads, which means taking the first two
 * freeze levels to put userspace to sleep, and the third freeze level to
 * prevent background threads from starting new transactions.  Take one level
 * more to prevent other callers from unfreezing the filesystem while we run.
 */
STATIC int
xchk_fscounters_freeze(
	struct xfs_scrub	*sc)
{ … }

/* Thaw the filesystem after checking or repairing fscounters. */
STATIC void
xchk_fscounters_cleanup(
	void			*buf)
{ … }

int
xchk_setup_fscounters(
	struct xfs_scrub	*sc)
{ … }

/*
 * Part 1: Collecting filesystem summary counts.  For each AG, we add its
 * summary counts (total inodes, free inodes, free data blocks) to an incore
 * copy of the overall filesystem summary counts.
 *
 * To avoid false corruption reports in part 2, any failure in this part must
 * set the INCOMPLETE flag even when a negative errno is returned.  This care
 * must be taken with certain errno values (i.e. EFSBADCRC, EFSCORRUPTED,
 * ECANCELED) that are absorbed into a scrub state flag update by
 * xchk_*_process_error.  Scrub and repair share the same incore data
 * structures, so the INCOMPLETE flag is critical to prevent a repair based on
 * insufficient information.
 */

/* Count free space btree blocks manually for pre-lazysbcount filesystems. */
static int
xchk_fscount_btreeblks(
	struct xfs_scrub	*sc,
	struct xchk_fscounters	*fsc,
	xfs_agnumber_t		agno)
{ … }

/*
 * Calculate what the global in-core counters ought to be from the incore
 * per-AG structure.  Callers can compare this to the actual in-core counters
 * to estimate by how much both in-core and on-disk counters need to be
 * adjusted.
 */
STATIC int
xchk_fscount_aggregate_agcounts(
	struct xfs_scrub	*sc,
	struct xchk_fscounters	*fsc)
{ … }

#ifdef CONFIG_XFS_RT
STATIC int
xchk_fscount_add_frextent(
	struct xfs_mount		*mp,
	struct xfs_trans		*tp,
	const struct xfs_rtalloc_rec	*rec,
	void				*priv)
{ … }

/* Calculate the number of free realtime extents from the realtime bitmap. */
STATIC int
xchk_fscount_count_frextents(
	struct xfs_scrub	*sc,
	struct xchk_fscounters	*fsc)
{ … }
#else
STATIC int
xchk_fscount_count_frextents(
	struct xfs_scrub	*sc,
	struct xchk_fscounters	*fsc)
{
	fsc->frextents = 0;
	fsc->frextents_delayed = 0;
	return 0;
}
#endif /* CONFIG_XFS_RT */

/*
 * Part 2: Comparing filesystem summary counters.  All we have to do here is
 * sum the percpu counters and compare them to what we've observed.
 */

/*
 * Is the @counter reasonably close to the @expected value?
 *
 * We neither locked nor froze anything in the filesystem while aggregating the
 * per-AG data to compute the @expected value, which means that the counter
 * could have changed.  We know the @old_value of the summation of the counter
 * before the aggregation, and we re-sum the counter now.  If the expected
 * value falls between the two summations, we're ok.
 *
 * Otherwise, we /might/ have a problem.  If the change in the summations is
 * more than we want to tolerate, the filesystem is probably busy and we should
 * just send back INCOMPLETE and see if userspace will try again.
 *
 * If we're repairing then we require an exact match.
 */
static inline bool
xchk_fscount_within_range(
	struct xfs_scrub	*sc,
	const int64_t		old_value,
	struct percpu_counter	*counter,
	uint64_t		expected)
{ … }

/* Check the superblock counters. */
int
xchk_fscounters(
	struct xfs_scrub	*sc)
{ … }
linux/fs/xfs/scrub/fscounters.c
