Merge tag 'scrub-fscounters-enhancements-6.2_2022-11-16' of...

Merge tag 'scrub-fscounters-enhancements-6.2_2022-11-16' of git://git.kernel.org/pub/scm/linux/kernel/git/djwong/xfs-linux into xfs-6.2-mergeA

xfs: enhance fs summary counter scrubber

This series makes two changes to the fs summary counter scrubber: first,
we should mark the scrub incomplete when we can't read the AG headers.
Second, it fixes a functionality gap where we don't actually check the
free rt extent count.

v23.2: fix pointless inline
Signed-off-by: Darrick J. Wong <djwong@kernel.org>

* tag 'scrub-fscounters-enhancements-6.2_2022-11-16' of git://git.kernel.org/pub/scm/linux/kernel/git/djwong/xfs-linux:
  xfs: online checking of the free rt extent count
  xfs: skip fscounters comparisons when the scan is incomplete

fs/xfs/scrub/fscounters.c

@@ -14,6 +14,8 @@
 #include "xfs_health.h"
 #include "xfs_btree.h"
 #include "xfs_ag.h"
+#include "xfs_rtalloc.h"
+#include "xfs_inode.h"
 #include "scrub/scrub.h"
 #include "scrub/common.h"
 #include "scrub/trace.h"
@@ -43,6 +45,16 @@
  * our tolerance for mismatch between expected and actual counter values.
  */
 
+struct xchk_fscounters {
+	struct xfs_scrub	*sc;
+	uint64_t		icount;
+	uint64_t		ifree;
+	uint64_t		fdblocks;
+	uint64_t		frextents;
+	unsigned long long	icount_min;
+	unsigned long long	icount_max;
+};
+
 /*
  * Since the expected value computation is lockless but only browses incore
  * values, the percpu counters should be fairly close to each other.  However,
@@ -120,6 +132,7 @@ xchk_setup_fscounters(
 	if (!sc->buf)
 		return -ENOMEM;
 	fsc = sc->buf;
+	fsc->sc = sc;
 
 	xfs_icount_range(sc->mp, &fsc->icount_min, &fsc->icount_max);
 
@@ -138,6 +151,18 @@ xchk_setup_fscounters(
 	return xchk_trans_alloc(sc, 0);
 }
 
+/*
+ * 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.
+ */
+
 /* Count free space btree blocks manually for pre-lazysbcount filesystems. */
 static int
 xchk_fscount_btreeblks(
@@ -225,8 +250,10 @@ xchk_fscount_aggregate_agcounts(
 	}
 	if (pag)
 		xfs_perag_put(pag);
-	if (error)
+	if (error) {
+		xchk_set_incomplete(sc);
 		return error;
+	}
 
 	/*
 	 * The global incore space reservation is taken from the incore
@@ -267,6 +294,64 @@ xchk_fscount_aggregate_agcounts(
 	return 0;
 }
 
+#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)
+{
+	struct xchk_fscounters		*fsc = priv;
+	int				error = 0;
+
+	fsc->frextents += rec->ar_extcount;
+
+	xchk_should_terminate(fsc->sc, &error);
+	return error;
+}
+
+/* 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)
+{
+	struct xfs_mount	*mp = sc->mp;
+	int			error;
+
+	fsc->frextents = 0;
+	if (!xfs_has_realtime(mp))
+		return 0;
+
+	xfs_ilock(sc->mp->m_rbmip, XFS_ILOCK_SHARED | XFS_ILOCK_RTBITMAP);
+	error = xfs_rtalloc_query_all(sc->mp, sc->tp,
+			xchk_fscount_add_frextent, fsc);
+	if (error) {
+		xchk_set_incomplete(sc);
+		goto out_unlock;
+	}
+
+out_unlock:
+	xfs_iunlock(sc->mp->m_rbmip, XFS_ILOCK_SHARED | XFS_ILOCK_RTBITMAP);
+	return error;
+}
+#else
+STATIC int
+xchk_fscount_count_frextents(
+	struct xfs_scrub	*sc,
+	struct xchk_fscounters	*fsc)
+{
+	fsc->frextents = 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?
  *
@@ -333,16 +418,17 @@ xchk_fscounters(
 {
 	struct xfs_mount	*mp = sc->mp;
 	struct xchk_fscounters	*fsc = sc->buf;
-	int64_t			icount, ifree, fdblocks;
+	int64_t			icount, ifree, fdblocks, frextents;
 	int			error;
 
 	/* Snapshot the percpu counters. */
 	icount = percpu_counter_sum(&mp->m_icount);
 	ifree = percpu_counter_sum(&mp->m_ifree);
 	fdblocks = percpu_counter_sum(&mp->m_fdblocks);
+	frextents = percpu_counter_sum(&mp->m_frextents);
 
 	/* No negative values, please! */
-	if (icount < 0 || ifree < 0 || fdblocks < 0)
+	if (icount < 0 || ifree < 0 || fdblocks < 0 || frextents < 0)
 		xchk_set_corrupt(sc);
 
 	/* See if icount is obviously wrong. */
@@ -353,6 +439,10 @@ xchk_fscounters(
 	if (fdblocks > mp->m_sb.sb_dblocks)
 		xchk_set_corrupt(sc);
 
+	/* See if frextents is obviously wrong. */
+	if (frextents > mp->m_sb.sb_rextents)
+		xchk_set_corrupt(sc);
+
 	/*
 	 * If ifree exceeds icount by more than the minimum variance then
 	 * something's probably wrong with the counters.
@@ -367,6 +457,13 @@ xchk_fscounters(
 	if (sc->sm->sm_flags & XFS_SCRUB_OFLAG_INCOMPLETE)
 		return 0;
 
+	/* Count the free extents counter for rt volumes. */
+	error = xchk_fscount_count_frextents(sc, fsc);
+	if (!xchk_process_error(sc, 0, XFS_SB_BLOCK(mp), &error))
+		return error;
+	if (sc->sm->sm_flags & XFS_SCRUB_OFLAG_INCOMPLETE)
+		return 0;
+
 	/* Compare the in-core counters with whatever we counted. */
 	if (!xchk_fscount_within_range(sc, icount, &mp->m_icount, fsc->icount))
 		xchk_set_corrupt(sc);
@@ -378,5 +475,9 @@ xchk_fscounters(
 			fsc->fdblocks))
 		xchk_set_corrupt(sc);
 
+	if (!xchk_fscount_within_range(sc, frextents, &mp->m_frextents,
+			fsc->frextents))
+		xchk_set_corrupt(sc);
+
 	return 0;
 }

fs/xfs/scrub/scrub.h

@@ -169,12 +169,4 @@ void xchk_xref_is_used_rt_space(struct xfs_scrub *sc, xfs_rtblock_t rtbno,
 # define xchk_xref_is_used_rt_space(sc, rtbno, len) do { } while (0)
 #endif
 
-struct xchk_fscounters {
-	uint64_t		icount;
-	uint64_t		ifree;
-	uint64_t		fdblocks;
-	unsigned long long	icount_min;
-	unsigned long long	icount_max;
-};
-
 #endif	/* __XFS_SCRUB_SCRUB_H__ */