Skip to contentSkip to conversation

generic/361: disable xfs metadata io error retries

PATCH New Darrick J. Wong 1 participant 1 message Jun 24, 2020

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 generic/361: disable xfs metadata io error retries

From: Darrick J. Wong <[email protected]>

This test examines the behavior of xfs when the underlying filesystem is a sparse image on the scratch filesystem when the scratch fs is about to run out of space. Unfortunately, the test assumes that the scratch fs will ENOSPC on the large data write. It's possible that metadata writeback will hit ENOSPC instead, and if we do, the test will hang forever while xfs retries the write. Make sure we're set up to fail fast so that we don't hang the test appliance.

---
 tests/generic/361 |    5 +++++
 1 file changed, 5 insertions(+)
tests/generic/361 +5 -0
index e28aaf46..403bbe41 100755
@@ -54,6 +54,11 @@ mkdir -p $fs_mnt
 loop_dev=$(_create_loop_device $fs_img)
 _mkfs_dev $loop_dev
 _mount $loop_dev $fs_mnt
+if [ "$FSTYP" = "xfs" ]; then
+	# Turn off all XFS metadata IO error retries
+	dname=$(_short_dev $loop_dev)
+	echo 0 | tee /sys/fs/xfs/$dname/error/*/*/* > /dev/null
+fi
 $XFS_IO_PROG -fc "pwrite 0 520m" $fs_mnt/testfile >>$seqres.full 2>&1
 
 # remount should not hang