On 11/11/21 4:24 PM, Dave Chinner wrote:
On Thu, Nov 11, 2021 at 12:17:16AM +0000, Catherine Hoang wrote:
Add an error tag on xfs_attr3_leaf_to_node to test log attribute
recovery and replay.
Signed-off-by: Catherine Hoang <catherine.hoang@xxxxxxxxxx>
---
fs/xfs/libxfs/xfs_attr_leaf.c | 6 ++++++
fs/xfs/libxfs/xfs_errortag.h | 4 +++-
fs/xfs/xfs_error.c | 3 +++
3 files changed, 12 insertions(+), 1 deletion(-)
diff --git a/fs/xfs/libxfs/xfs_attr_leaf.c b/fs/xfs/libxfs/xfs_attr_leaf.c
index 74b76b09509f..fdeb09de74ca 100644
--- a/fs/xfs/libxfs/xfs_attr_leaf.c
+++ b/fs/xfs/libxfs/xfs_attr_leaf.c
@@ -28,6 +28,7 @@
#include "xfs_dir2.h"
#include "xfs_log.h"
#include "xfs_ag.h"
+#include "xfs_errortag.h"
/*
@@ -1189,6 +1190,11 @@ xfs_attr3_leaf_to_node(
trace_xfs_attr_leaf_to_node(args);
+ if (XFS_TEST_ERROR(false, mp, XFS_ERRTAG_LEAF_TO_NODE)) {
+ error = -EIO;
+ goto out;
+ }
+
error = xfs_da_grow_inode(args, &blkno);
if (error)
goto out;
diff --git a/fs/xfs/libxfs/xfs_errortag.h b/fs/xfs/libxfs/xfs_errortag.h
index 31aeeb94dd5b..cc1650b58723 100644
--- a/fs/xfs/libxfs/xfs_errortag.h
+++ b/fs/xfs/libxfs/xfs_errortag.h
@@ -61,7 +61,8 @@
#define XFS_ERRTAG_AG_RESV_FAIL 38
#define c 39
#define XFS_ERRTAG_LEAF_SPLIT 40
-#define XFS_ERRTAG_MAX 41
+#define XFS_ERRTAG_LEAF_TO_NODE 41
+#define XFS_ERRTAG_MAX 42
Same again about naming. THis is an attribute fork injection point,
not a generic "leaf-to-node" error injection point.
Whihc makes me wonder: this is testing just the initial leaf split
shape change. There are other shape changes - inline -> leaf, leaf
-> multi-level tree (xfs_da3_split()) and multi-level -> single
level leaf, leaf -> inline - for the attr tree, and there are even
more shape changes for the directory tree (inline, leaf, leafN,
node)
As a general infrastructure question, are we really going to now add
separate error injections for each different shape change on each
different type of btree? That explodes the number of injection
points quite quickly....
Well, that certainly seem like a bit much. But I don't think that just
because we've added a few error tags that we necessarily have to put one
on every transformation. I think it's a fair compromise to simply
identify a few points of interest, and if we later find a need to make
the tests more rigorous we can expand the infrastructure we've set up here?
Allison
Cheers,
Dave.
