[PATCH] common: skip data write EIO survival tests on fatal configs

[PATCH] common: skip data write EIO survival tests on fatal configs

[Yao Sang]

generic/441 and xfs/656 inject data write errors after mounting the scratch
filesystem on dm-error. These tests do not verify whether a filesystem may
shut down after media errors. They require the filesystem to survive long
enough to report the error through the path under test: generic/441 checks
fsync(2) writeback error reporting, and xfs/656 checks healthmon per-I/O
error events.

That expectation is not valid for filesystem configurations where a data
write error is fatal. Zoned XFS is one such case because a failed data write
can leave the in-core zone allocation state untrustworthy, so shutting down
the mount is expected behavior.

Add a common helper for tests that require data write EIO to be reported
without shutting down the mounted filesystem, and use it after mounting the
dm-error device in both tests. The helper currently detects zoned XFS by
checking the XFS per-mount zoned/max_open_zones sysfs attribute for the
mounted device, so the check follows the filesystem instance rather than the
underlying block queue alone.

Fatal-on-data-write-error configurations are reported as notrun instead of
test failures.

Signed-off-by: Yao Sang <sangyao@kylinos.cn>
---
Changes since RFC:
- Replace the separate fatal predicate and survival requirement helpers with
  a single _require_fs_survives_data_write_eio helper.
- Drop the FS_DATA_WRITE_EIO_SHUTDOWN override.
- Require callers to pass the mounted device explicitly.
- Remove _scratch from the helper name.
- Keep the XFS per-mount zoned/max_open_zones detection.

RFC Link:
https://lore.kernel.org/fstests/20260618034756.615426-1-sangyao@kylinos.cn/

 common/rc         | 22 ++++++++++++++++++++++
 tests/generic/441 |  1 +
 tests/xfs/656     |  1 +
 3 files changed, 24 insertions(+)

diff --git a/common/rc b/common/rc
index 9632b211..4bf0861b 100644
--- a/common/rc
+++ b/common/rc
@@ -4162,6 +4162,28 @@ _require_scratch_shutdown_and_syncfs()
 	_require_scratch_shutdown
 }
 
+# Require that the mounted filesystem survive data write errors long enough to
+# report them through the regular error reporting path.
+_require_fs_survives_data_write_eio()
+{
+	local dev=$1
+
+	if [ ! -b "$dev" ]; then
+		_fail "Usage: _require_fs_survives_data_write_eio <mounted_device>"
+	fi
+
+	case "$FSTYP" in
+	xfs)
+		# This is an XFS per-mount sysfs attribute, not a block queue
+		# attribute.  Its existence means XFS mounted this instance with
+		# zoned filesystem semantics.
+		if _has_fs_sysfs_attr "$dev" "zoned/max_open_zones"; then
+			_notrun "zoned XFS shuts down on data write EIO"
+		fi
+		;;
+	esac
+}
+
 _check_s_dax()
 {
 	local target=$1
diff --git a/tests/generic/441 b/tests/generic/441
index 6b48fc9e..cc370bea 100755
--- a/tests/generic/441
+++ b/tests/generic/441
@@ -53,6 +53,7 @@ echo "Format and mount"
 _scratch_mkfs > $seqres.full 2>&1
 _dmerror_init no_log
 _dmerror_mount
+_require_fs_survives_data_write_eio $DMERROR_DEV
 
 _require_fs_space $SCRATCH_MNT 65536
 
diff --git a/tests/xfs/656 b/tests/xfs/656
index 3a867692..3830d719 100755
--- a/tests/xfs/656
+++ b/tests/xfs/656
@@ -48,6 +48,7 @@ echo "Format and mount"
 _scratch_mkfs > $seqres.full 2>&1
 _dmerror_init no_log
 _dmerror_mount
+_require_fs_survives_data_write_eio $DMERROR_DEV
 
 _require_fs_space $SCRATCH_MNT 65536
 
-- 
2.25.1

Re: [PATCH] common: skip data write EIO survival tests on fatal configs

[Christoph Hellwig]

Looks good:

Reviewed-by: Christoph Hellwig <hch@lst.de>