Thread (4 messages) 4 messages, 3 authors, 2021-10-14

Re: [PATCH v2] btrfs-progs: test-misc: search the backup slot to use at runtime

From: Qu Wenruo <hidden>
Date: 2021-10-14 22:36:45


On 2021/10/14 23:36, Josef Bacik wrote:
On Wed, Oct 13, 2021 at 09:12:33AM +0800, Qu Wenruo wrote:
quoted
Test case misc/038 uses hardcoded backup slot number, this means if we
change how many transactions we commit during mkfs, it will immediately
break the tests.

Such hardcoded tests will be a big pain for later btrfs-progs updates.

Update it with runtime backup slot search.

Such search is done by using current filesystem generation as a search
target and grab the slot number.

By this, no matter how many transactions we commit during mkfs, the test
case should be able to handle it.

Signed-off-by: Qu Wenruo <redacted>
---
Changelog:
v2:
- Use run_check() instead of manually redirect output to "$RESULT"
- Quote "$main_root_ptr"
---
  .../038-backup-root-corruption/test.sh        | 47 ++++++++++++-------
  1 file changed, 29 insertions(+), 18 deletions(-)
diff --git a/tests/misc-tests/038-backup-root-corruption/test.sh b/tests/misc-tests/038-backup-root-corruption/test.sh
index b6c3671f2c3a..bf41f1e0952b 100755
--- a/tests/misc-tests/038-backup-root-corruption/test.sh
+++ b/tests/misc-tests/038-backup-root-corruption/test.sh
@@ -17,25 +17,35 @@ run_check $SUDO_HELPER touch "$TEST_MNT/file"
  run_check_umount_test_dev

  dump_super() {
-	run_check_stdout $SUDO_HELPER "$TOP/btrfs" inspect-internal dump-super -f "$TEST_DEV"
+	# In this test, we will dump super block multiple times, while the
+	# existing run_check*() helpers will always dump all the output into
+	# the log, flooding the log and hide real important info.
+	# Thus here we call "btrfs" directly.
+	$SUDO_HELPER "$TOP/btrfs" inspect-internal dump-super -f "$TEST_DEV"
  }

-# Ensure currently active backup slot is the expected one (slot 3)
-backup2_root_ptr=$(dump_super | grep -A1 "backup 2" | grep backup_tree_root | awk '{print $2}')
-
  main_root_ptr=$(dump_super | awk '/^root\t/{print $2}')
+# Grab current fs generation, and it will be used to determine which backup
+# slot to use
+cur_gen=$(dump_super | grep ^generation | awk '{print $2}')
+backup_gen=$(($cur_gen - 1))
+
+# Grab the slot which matches @backup_gen
+found=$(dump_super | grep backup_tree_root | grep -n "gen: $backup_gen")

-if [ "$backup2_root_ptr" -ne "$main_root_ptr" ]; then
-	_log "Backup slot 2 not in use, trying slot 3"
-	# Or use the next slot in case of free-space-tree
-	backup3_root_ptr=$(dump_super | grep -A1 "backup 3" | grep backup_tree_root | awk '{print $2}')
-	if [ "$backup3_root_ptr" -ne "$main_root_ptr" ]; then
-		_fail "Neither backup slot 2 nor slot 3 are in use"
-	fi
-	_log "Backup slot 3 in use"
+if [ -z "$found" ]; then
+	_fail "Unable to find a backup slot with generation $backup_gen"
  fi

-run_check "$INTERNAL_BIN/btrfs-corrupt-block" -m $main_root_ptr -f generation "$TEST_DEV"
+slot_num=$(echo $found | cut -f1 -d:)
+# To follow the dump-super output, where backup slot starts at 0.
+slot_num=$(($slot_num - 1))
What happens if we're on $slot_num == 0?  Seems like this would mess up, right?
Thanks,
Note that, the $found is from "grep -n" which starts its line number at 1.

Thus $slot_num will always be >= 1, and nothing will be wrong.

Thanks,
Qu
Josef
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help