btrfs-progs: check: Make btrfs check return error for qgroup mismatch

PATCH Applied Qu Wenruo 3 participants 6 messages Apr 30, 2018 – May 10, 2018 Btrfs file system

lore

Share & tools

lore

Qu Wenruo

author

2018-04-30 06:37 UTC #0 lore raw

PATCH 1/2 btrfs-progs: check: Make btrfs check return error for qgroup mismatch

Current btrfs-check will check qgroup consistency, but even when it finds something wrong, the return value is still 0.

Fix it by allowing report_qgroups() to return int to indicate qgroup mismatch, and also add extra logical to return no error if qgroup repair is successful.

Without this patch, fstests can't detect qgroup corruption by its fsck alone.

Signed-off-by Qu Wenruo <[email protected]>

---
 check/main.c    | 10 ++++++----
 qgroup-verify.c | 20 +++++++++++++++++---
 qgroup-verify.h |  2 +-
 3 files changed, 24 insertions(+), 8 deletions(-)

check/main.c +6 -4

index c4a1801fb0ef..d49b4326915e 100644
@@ -9467,6 +9467,7 @@ int cmd_check(int argc, char **argv)
 	int clear_space_cache = 0;
 	int qgroup_report = 0;
 	int qgroups_repaired = 0;
+	int qgroup_report_ret;
 	unsigned ctree_flags = OPEN_CTREE_EXCLUSIVE;
 	int force = 0;

@@ -9709,7 +9710,7 @@ int cmd_check(int argc, char **argv)
 		ret = qgroup_verify_all(info);
 		err |= !!ret;
 		if (ret == 0)
-			report_qgroups(1);
+			err |= !!report_qgroups(1);
 		goto close_out;
 	}
 	if (subvolid) {

@@ -9894,13 +9895,14 @@ int cmd_check(int argc, char **argv)
 			error("failed to check quota groups");
 			goto out;
 		}
-		report_qgroups(0);
+		qgroup_report_ret = report_qgroups(0);
 		ret = repair_qgroups(info, &qgroups_repaired);
-		err |= !!ret;
-		if (err) {
+		if (ret) {
 			error("failed to repair quota groups");
 			goto out;
 		}
+		if (qgroup_report_ret && (!qgroups_repaired || ret))
+			err |= qgroup_report_ret;
 		ret = 0;
 	}

qgroup-verify.c +17 -3

index 571b4d4f7171..4deb9879fb35 100644
@@ -1298,10 +1298,19 @@ static int report_qgroup_difference(struct qgroup_count *count, int verbose)
 	return is_different;
 }
 
-void report_qgroups(int all)
+/*
+ * Report qgroups errors
+ * Return 0 if nothing wrong.
+ * Return <0 if any qgroup is inconsistent.
+ *
+ * @all:	if set, all qgroup will be checked and reported even already
+ * 		inconsistent or under rescan.
+ */
+int report_qgroups(int all)
 {
 	struct rb_node *node;
 	struct qgroup_count *c;
+	bool found_err = false;
 
 	if (!repair && counts.rescan_running) {
 		if (all) {

@@ -1310,7 +1319,7 @@ void report_qgroups(int all)
 		} else {
 			printf(
 	"Qgroup rescan is running, qgroups will not be printed.\n");
-			return;
+			return 0;
 		}
 	}
 	if (counts.qgroup_inconsist && !counts.rescan_running)

@@ -1319,11 +1328,16 @@ void report_qgroups(int all)
 	while (node) {
 		c = rb_entry(node, struct qgroup_count, rb_node);
 
-		if (report_qgroup_difference(c, all))
+		if (report_qgroup_difference(c, all)) {
 			list_add_tail(&c->bad_list, &bad_qgroups);
+			found_err = true;
+		}
 
 		node = rb_next(node);
 	}
+	if (found_err)
+		return -EUCLEAN;
+	return 0;
 }
 
 void free_qgroup_counts(void)

qgroup-verify.h +1 -1

index d7d83a46ed5a..14d36bbf81d1 100644
@@ -23,7 +23,7 @@
 #include "ctree.h"
 
 int qgroup_verify_all(struct btrfs_fs_info *info);
-void report_qgroups(int all);
+int report_qgroups(int all);
 int repair_qgroups(struct btrfs_fs_info *info, int *repaired);
 
 int print_extent_state(struct btrfs_fs_info *info, u64 subvol);

Signature · 1 line

2.17.0

Qu Wenruo

author

2018-04-30 06:17 UTC #1 lore raw

PATCH 2/2 btrfs-progs: fsck-tests: Add test case to ensure btrfs check return error for corrupted qgroups

Since the test case uses run_mustfail(), which is pretty easy pass due to other unexpected problems, so here an extra run_check() is added to ensure we don't only report qgroup error, but also fix it without problem.

Signed-off-by Qu Wenruo <[email protected]>

---
 .../032-corrupted-qgroup/qgroup_corrupted.img | Bin 0 -> 3072 bytes
 tests/fsck-tests/032-corrupted-qgroup/test.sh |  19 ++++++++++++++++++
 2 files changed, 19 insertions(+)
 create mode 100644 tests/fsck-tests/032-corrupted-qgroup/qgroup_corrupted.img
 create mode 100755 tests/fsck-tests/032-corrupted-qgroup/test.sh

tests/fsck-tests/032-corrupted-qgroup/qgroup_corrupted.img new file binary +0 -0

new file mode 100644
index 0000000000000000000000000000000000000000..a19a303534d419fa7660615d0dcf25ad314e7b20
GIT binary patch

tests/fsck-tests/032-corrupted-qgroup/test.sh new file +19 -0

new file mode 100755
index 000000000000..ccecfeac3ca3
@@ -0,0 +1,19 @@
+#!/bin/bash
+# To check if btrfs check can handle valid orphan items.
+# Orphan item is a marker for deleted inodes that were open at the time of
+# deletion.  # Orphan inode/root will is not referenced and will have an orphan
+# item, which should not be reported as error.
+
+source "$TEST_TOP/common"
+
+check_prereq btrfs
+
+check_image() {
+	run_mustfail "btrfs check failed to detect qgroup corruption" \
+		     "$TOP/btrfs" check "$1"
+	# Above command can fail due to other bugs, so add extra safe net to
+	# ensure we can fix qgroup without problem.
+	run_check "$TOP/btrfs" check --repair "$1"
+}
+
+check_all_images

Signature · 1 line

2.17.0

David Sterba

Applied

2018-05-08 17:47 UTC #2 lore raw

In reply to Qu Wenruo

Quoted 11 lines · Qu Wenruo

Current btrfs-check will check qgroup consistency, but even when it finds something wrong, the return value is still 0.
Fix it by allowing report_qgroups() to return int to indicate qgroup mismatch, and also add extra logical to return no error if qgroup repair is successful.
Without this patch, fstests can't detect qgroup corruption by its fsck alone.
Signed-off-by Qu Wenruo <[email protected]>

Applied, thanks. Please send the 0/N (cover letter) mail even for 2-patch series.

Qu Wenruo

2018-05-10 01:43 UTC #3 lore raw

Quoted 8 lines · David Sterba

Quoted 7 lines · Qu Wenruo
Current btrfs-check will check qgroup consistency, but even when it finds something wrong, the return value is still 0.
Fix it by allowing report_qgroups() to return int to indicate qgroup mismatch, and also add extra logical to return no error if qgroup repair is successful.

Quoted 7 lines

Quoted 4 lines
Without this patch, fstests can't detect qgroup corruption by its fsck alone.
Signed-off-by Qu Wenruo <[email protected]>
Applied, thanks. Please send the 0/N (cover letter) mail even for 2-patch series.

Any advice on the necessity of the cover letter?

Is cover letter highly recommended for any patches which have any dependency like the 2nd test case?

Thanks,
Qu

Quoted 5 lines

--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

David Sterba

2018-05-10 11:48 UTC #4 lore raw

Quoted 16 lines · Qu Wenruo

Quoted 13 lines · David Sterba
Quoted 11 lines · Qu Wenruo
Current btrfs-check will check qgroup consistency, but even when it finds something wrong, the return value is still 0.
Fix it by allowing report_qgroups() to return int to indicate qgroup mismatch, and also add extra logical to return no error if qgroup repair is successful.
Without this patch, fstests can't detect qgroup corruption by its fsck alone.
Signed-off-by Qu Wenruo <[email protected]>

Quoted 7 lines

Quoted 2 lines
Applied, thanks. Please send the 0/N (cover letter) mail even for 2-patch series.
Any advice on the necessity of the cover letter?
Is cover letter highly recommended for any patches which have any dependency like the 2nd test case?

The cover letter is a place where you can get feedback on the whole patchset, individual patches get the specific comments to the code.

A patchset makes a logical chunk of patches so the cover letter also serves as the high-level overview of the changes before I or others even start looking further.

I've seen other maintainers to prefer cover letters, but as this is not documented and not generally appreciated by developers, it needs to be stated. Developers who post patches frequently could help to improve the patch flow process, but lack of cover letter is not a big deal otherwise.

Qu Wenruo

2018-05-10 12:17 UTC #5 lore raw

Quoted 32 lines · David Sterba

Quoted 16 lines · Qu Wenruo
Quoted 13 lines · David Sterba
Quoted 11 lines · Qu Wenruo
Current btrfs-check will check qgroup consistency, but even when it finds something wrong, the return value is still 0.
Fix it by allowing report_qgroups() to return int to indicate qgroup mismatch, and also add extra logical to return no error if qgroup repair is successful.
Without this patch, fstests can't detect qgroup corruption by its fsck alone.
Signed-off-by Qu Wenruo <[email protected]>
Quoted 7 lines
Quoted 2 lines
Applied, thanks. Please send the 0/N (cover letter) mail even for 2-patch series.
Any advice on the necessity of the cover letter?
Is cover letter highly recommended for any patches which have any dependency like the 2nd test case?
The cover letter is a place where you can get feedback on the whole patchset, individual patches get the specific comments to the code.
A patchset makes a logical chunk of patches so the cover letter also serves as the high-level overview of the changes before I or others even start looking further.

Quoted 5 lines

I've seen other maintainers to prefer cover letters, but as this is not documented and not generally appreciated by developers, it needs to be stated. Developers who post patches frequently could help to improve the patch flow process, but lack of cover letter is not a big deal otherwise.

Much appreciated about the explain on this topic.

Personally I'm completely fine for cover letters, so I would take this method for later logically related patches.

Thanks,
Qu