Test case btrfs/125 previous has a very random pass/fail ratio.
This is due to the fact that previously btrfs RAID56 recovery path will
always trust the cached data.
Thus if there is some operation reading part/all of the full stripe,
that full stripe can be cached, making later recovery to generate bad
data.
As the cached data can be stale (intentionally caused by the test case).
Upstream commit d4e28d9b5f04 ("btrfs: raid56: make steal_rbio() subpage
compatible") incidentally disable all the cache re-use for btrfs RAID56,
thus make btrfs/125 to always pass.
Although we will later re-enable the cache behavior for btrfs RAID56, we
will not let recovery path to use any cache at all, thus this will give
us the best things from both worlds, making btrfs/125 to pass
consistently, while still enable cache for regular read/write.
Signed-off-by: Qu Wenruo <wqu@xxxxxxxx>
---
tests/btrfs/125 | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/tests/btrfs/125 b/tests/btrfs/125
index b58f2aa282bd..ced812cd921a 100755
--- a/tests/btrfs/125
+++ b/tests/btrfs/125
@@ -22,7 +22,7 @@
# Verify if all three checkpoints match
#
. ./common/preamble
-_begin_fstest replace volume balance
+_begin_fstest auto replace volume balance
# Override the default cleanup function.
_cleanup()
--
2.36.1
