Skip to contentSkip to conversation

btrfs: qgroup: Automatically remove qgroup item when dropping a subvolume

PATCH In review Qu Wenruo 2 participants 3 messages Oct 17, 2019 – Feb 3, 2020 Btrfs file system

lore
Share & tools
lore

Qu Wenruo

author
#0 lore raw

PATCH 1/1 btrfs: qgroup: Automatically remove qgroup item when dropping a subvolume

[BUG]
When a subvolume is created, we automatically create a level 0 qgroup for it, but don't remove it when the subvolume is dropped.

Although it's not a big deal, it can easily pollute the output of "btrfs qgroup show" and make it pretty annoying.

[FIX]
For btrfs_drop_snapshot(), if it's a valid subvolume (not a reloc tree) and qgroup is enabled, we do the following work to remove the qgroup:
- Commit transaction
This is to ensure that the qgroup numbers of that subvolume is updated properly (all number of that subvolume should be 0).

- Start a new transaction for later operation

- Call btrfs_remove_qgroup()

So that qgroup can be automatically removed when the subvolume get fully dropped.

---
 fs/btrfs/extent-tree.c | 25 +++++++++++++++++++++++++
 1 file changed, 25 insertions(+)
fs/btrfs/extent-tree.c +25 -0
index 49cb26fa7c63..5e8569cad16d 100644
@@ -5182,6 +5182,7 @@ int btrfs_drop_snapshot(struct btrfs_root *root,
 	struct btrfs_root_item *root_item = &root->root_item;
 	struct walk_control *wc;
 	struct btrfs_key key;
+	u64 rootid = root->root_key.objectid;
 	int err = 0;
 	int ret;
 	int level;
@@ -5384,6 +5385,30 @@ int btrfs_drop_snapshot(struct btrfs_root *root,
 	}
 	root_dropped = true;
 out_end_trans:
+	/* If qgroup is enabled, also try to remove the qgroup */
+	if (test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags) && root_dropped &&
+	    is_fstree(rootid) && !for_reloc) {
+		/* Commit transaction so qgroup numbers get updated to all 0 */
+		ret = btrfs_commit_transaction(trans);
+		if (ret < 0) {
+			btrfs_end_transaction_throttle(trans);
+			err = ret;
+			goto out_free;
+		}
+		trans = btrfs_start_transaction(fs_info->quota_root, 1);
+		if (IS_ERR(trans)) {
+			err = PTR_ERR(trans);
+			goto out_free;
+		}
+
+		/*
+		 * Here we ignore the return value of btrfs_remove_qgroup(),
+		 * as there is no critical error could happen.
+		 * Error like EINVAL (quota disabled) or ENOENT (no such qgroup)
+		 * are all safe to ignore.
+		 */
+		btrfs_remove_qgroup(trans, rootid);
+	}
 	btrfs_end_transaction_throttle(trans);
 out_free:
 	kfree(wc);
Signature · 1 line
2.23.0
author
#1 lore raw

Ping?

Quoted 70 lines · Qu Wenruo

[BUG]
When a subvolume is created, we automatically create a level 0 qgroup for it, but don't remove it when the subvolume is dropped.

Although it's not a big deal, it can easily pollute the output of "btrfs qgroup show" and make it pretty annoying.

[FIX]
For btrfs_drop_snapshot(), if it's a valid subvolume (not a reloc tree) and qgroup is enabled, we do the following work to remove the qgroup:
- Commit transaction
This is to ensure that the qgroup numbers of that subvolume is updated properly (all number of that subvolume should be 0).

- Start a new transaction for later operation

- Call btrfs_remove_qgroup()

So that qgroup can be automatically removed when the subvolume get fully dropped.

---
 fs/btrfs/extent-tree.c | 25 +++++++++++++++++++++++++
 1 file changed, 25 insertions(+)
fs/btrfs/extent-tree.c +25 -0
index 49cb26fa7c63..5e8569cad16d 100644
@@ -5182,6 +5182,7 @@ int btrfs_drop_snapshot(struct btrfs_root *root,
 	struct btrfs_root_item *root_item = &root->root_item;
 	struct walk_control *wc;
 	struct btrfs_key key;
+	u64 rootid = root->root_key.objectid;
 	int err = 0;
 	int ret;
 	int level;
@@ -5384,6 +5385,30 @@ int btrfs_drop_snapshot(struct btrfs_root *root,
 	}
 	root_dropped = true;
 out_end_trans:
+	/* If qgroup is enabled, also try to remove the qgroup */
+	if (test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags) && root_dropped &&
+	    is_fstree(rootid) && !for_reloc) {
+		/* Commit transaction so qgroup numbers get updated to all 0 */
+		ret = btrfs_commit_transaction(trans);
+		if (ret < 0) {
+			btrfs_end_transaction_throttle(trans);
+			err = ret;
+			goto out_free;
+		}
+		trans = btrfs_start_transaction(fs_info->quota_root, 1);
+		if (IS_ERR(trans)) {
+			err = PTR_ERR(trans);
+			goto out_free;
+		}
+
+		/*
+		 * Here we ignore the return value of btrfs_remove_qgroup(),
+		 * as there is no critical error could happen.
+		 * Error like EINVAL (quota disabled) or ENOENT (no such qgroup)
+		 * are all safe to ignore.
+		 */
+		btrfs_remove_qgroup(trans, rootid);
+	}
 	btrfs_end_transaction_throttle(trans);
 out_free:
 	kfree(wc);
#2 lore raw
In reply to Qu Wenruo
Quoted 39 lines · Qu Wenruo

[BUG]
When a subvolume is created, we automatically create a level 0 qgroup for it, but don't remove it when the subvolume is dropped.

Although it's not a big deal, it can easily pollute the output of "btrfs qgroup show" and make it pretty annoying.

[FIX]
For btrfs_drop_snapshot(), if it's a valid subvolume (not a reloc tree) and qgroup is enabled, we do the following work to remove the qgroup:
- Commit transaction
This is to ensure that the qgroup numbers of that subvolume is updated properly (all number of that subvolume should be 0).

- Start a new transaction for later operation

- Call btrfs_remove_qgroup()

So that qgroup can be automatically removed when the subvolume get fully dropped.

---
  fs/btrfs/extent-tree.c | 25 +++++++++++++++++++++++++
  1 file changed, 25 insertions(+)
fs/btrfs/extent-tree.c +1 -0
index 49cb26fa7c63..5e8569cad16d 100644
@@ -5182,6 +5182,7 @@ int btrfs_drop_snapshot(struct btrfs_root *root,
  	struct btrfs_root_item *root_item = &root->root_item;
  	struct walk_control *wc;
  	struct btrfs_key key;
+	u64 rootid = root->root_key.objectid;
  	int err = 0;
  	int ret;
  	int level;

fs/btrfs/extent-tree.c int btrfs_drop_snapshot(struct btrfs_root *root, · line 5390

Quoted 6 lines
fs/btrfs/extent-tree.c +3 -0
@@ -5384,6 +5385,30 @@ int btrfs_drop_snapshot(struct btrfs_root *root,
  	}
  	root_dropped = true;
  out_end_trans:
+	/* If qgroup is enabled, also try to remove the qgroup */
+	if (test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags) && root_dropped &&
+	    is_fstree(rootid) && !for_reloc) {

Is the !for_reloc check superflous here? is_fstree() should be enough. Thanks,

Josef