[v2] xfs_db: don't hardcode 'type data' size at 512b

Commit Message

Darrick J. Wong Feb. 22, 2024, 10:04 p.m. UTC

From: Darrick J. Wong <[e-mail]>

On a disk with 4096-byte LBAs, the xfs_db 'type data' subcommand doesn't
work:

# xfs_io -c 'sb' -c 'type data' /dev/sda
xfs_db: read failed: Invalid argument
no current object

The cause of this is the hardcoded initialization of bb_count when we're
setting type data -- it should be the filesystem sector size, not just 1.

Signed-off-by: Darrick J. Wong <[e-mail]>
Reviewed-by: Christoph Hellwig <[e-mail]>
Reviewed-by: Carlos Maiolino <[e-mail]>
---
v2: rebase to for-next
---
 db/block.c |    3 ++-
 db/io.c    |    3 ++-
 2 files changed, 4 insertions(+), 2 deletions(-)

Patch

diff --git a/db/block.c b/db/block.c
index d730c7796710..22930e5a287e 100644
--- a/db/block.c
+++ b/db/block.c
@@ -124,6 +124,7 @@  daddr_f(
 {
 	int64_t		d;
 	char		*p;
+	int		bb_count = BTOBB(mp->m_sb.sb_sectsize);
 
 	if (argc == 1) {
 		xfs_daddr_t	daddr = iocur_top->off >> BBSHIFT;
@@ -144,7 +145,7 @@  daddr_f(
 		return 0;
 	}
 	ASSERT(typtab[TYP_DATA].typnm == TYP_DATA);
-	set_cur(&typtab[TYP_DATA], d, 1, DB_RING_ADD, NULL);
+	set_cur(&typtab[TYP_DATA], d, bb_count, DB_RING_ADD, NULL);
 	return 0;
 }
 
diff --git a/db/io.c b/db/io.c
index 590dd1f82f7b..9b2c6b4cf7e9 100644
--- a/db/io.c
+++ b/db/io.c
@@ -652,7 +652,8 @@  void
 set_iocur_type(
 	const typ_t	*type)
 {
-	int		bb_count = 1;	/* type's size in basic blocks */
+	/* type's size in basic blocks */
+	int		bb_count = BTOBB(mp->m_sb.sb_sectsize);
 	int		boff = iocur_top->boff;
 
 	/*