From: Darrick J. Wong <djwong@xxxxxxxxxx>
If metadata directory trees are enabled, the superblock inode pointers
to quota and rt free space metadata must all be zero. The only inode
pointers in the superblock are sb_rootino and sb_metadirino.
Found by running xfs/418.
Fixes: b790ab2a303d58 ("xfs_repair: support quota inodes in the metadata directory")
Signed-off-by: "Darrick J. Wong" <djwong@xxxxxxxxxx>
---
Note: I found this omission while reading the fuzz tests results, and
it's a fix for new code in 6.13. Could we please get this in before
the release of xfsprogs 6.13?
---
repair/agheader.c | 49 ++++++++++++++++++++++++++++++++++++++++++++++++-
1 file changed, 48 insertions(+), 1 deletion(-)
diff --git a/repair/agheader.c b/repair/agheader.c
index 89a23a869a02e4..327ba041671f9f 100644
--- a/repair/agheader.c
+++ b/repair/agheader.c
@@ -319,6 +319,51 @@ check_v5_feature_mismatch(
return XR_AG_SB_SEC;
}
+static inline int
+require_zeroed_ino(
+ struct xfs_mount *mp,
+ __be64 *inop,
+ const char *tag,
+ xfs_agnumber_t agno,
+ int do_bzero)
+{
+ if (*inop == 0)
+ return 0;
+ if (!no_modify)
+ *inop = 0;
+ if (do_bzero)
+ return XR_AG_SB_SEC;
+
+ do_warn(_("non-zero %s inode field in superblock %d\n"),
+ tag, agno);
+ return XR_AG_SB;
+}
+
+/* With metadir, quota and rt metadata inums in the sb must all be zero. */
+static int
+check_pre_metadir_sb_inodes(
+ struct xfs_mount *mp,
+ struct xfs_buf *sbuf,
+ xfs_agnumber_t agno,
+ int do_bzero)
+{
+ struct xfs_dsb *dsb = sbuf->b_addr;
+ int rval = 0;
+
+ rval |= require_zeroed_ino(mp, &dsb->sb_uquotino,
+ _("user quota"), agno, do_bzero);
+ rval |= require_zeroed_ino(mp, &dsb->sb_gquotino,
+ _("group quota"), agno, do_bzero);
+ rval |= require_zeroed_ino(mp, &dsb->sb_pquotino,
+ _("project quota"), agno, do_bzero);
+
+ rval |= require_zeroed_ino(mp, &dsb->sb_rbmino,
+ _("realtime bitmap"), agno, do_bzero);
+ rval |= require_zeroed_ino(mp, &dsb->sb_rsumino,
+ _("realtime summary"), agno, do_bzero);
+ return rval;
+}
+
/*
* quota inodes and flags in secondary superblocks are never set by mkfs.
* However, they could be set in a secondary if a fs with quotas was growfs'ed
@@ -509,7 +554,9 @@ secondary_sb_whack(
rval |= XR_AG_SB_SEC;
}
- if (!xfs_has_metadir(mp))
+ if (xfs_has_metadir(mp))
+ rval |= check_pre_metadir_sb_inodes(mp, sbuf, i, do_bzero);
+ else
rval |= secondary_sb_quota(mp, sbuf, sb, i, do_bzero);
/*
