btrfs: qgroups: Move transaction management inside btrfs_quota_enable/disable

Commit 5d23515be6 ("btrfs: Move qgroup rescan on quota enable to
btrfs_quota_enable") not only resulted in an easier to follow code but
it also introduced a subtle bug. It changed the timing when the initial
transaction rescan was happening:

- before the commit: it would happen after transaction commit had occured
- after the commit: it might happen before the transaction was committed

This results in failure to correctly rescan the quota since there could
be data which is still not committed on disk.

This patch aims to fix this by moving the transaction creation/commit
inside btrfs_quota_enable, which allows to schedule the quota commit
after the transaction has been committed.

Fixes: 5d23515be6 ("btrfs: Move qgroup rescan on quota enable to btrfs_quota_enable")
Reported-by: Misono Tomohiro <misono.tomohiro@jp.fujitsu.com>
Link: https://marc.info/?l=linux-btrfs&m=152999289017582
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

fs/btrfs/ioctl.c

@@ -5105,9 +5105,7 @@ static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
 	struct inode *inode = file_inode(file);
 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
 	struct btrfs_ioctl_quota_ctl_args *sa;
-	struct btrfs_trans_handle *trans = NULL;
 	int ret;
-	int err;
 
 	if (!capable(CAP_SYS_ADMIN))
 		return -EPERM;
@@ -5123,28 +5121,19 @@ static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
 	}
 
 	down_write(&fs_info->subvol_sem);
-	trans = btrfs_start_transaction(fs_info->tree_root, 2);
-	if (IS_ERR(trans)) {
-		ret = PTR_ERR(trans);
-		goto out;
-	}
 
 	switch (sa->cmd) {
 	case BTRFS_QUOTA_CTL_ENABLE:
-		ret = btrfs_quota_enable(trans, fs_info);
+		ret = btrfs_quota_enable(fs_info);
 		break;
 	case BTRFS_QUOTA_CTL_DISABLE:
-		ret = btrfs_quota_disable(trans, fs_info);
+		ret = btrfs_quota_disable(fs_info);
 		break;
 	default:
 		ret = -EINVAL;
 		break;
 	}
 
-	err = btrfs_commit_transaction(trans);
-	if (err && !ret)
-		ret = err;
-out:
 	kfree(sa);
 	up_write(&fs_info->subvol_sem);
 drop_write:

fs/btrfs/qgroup.c

@@ -875,8 +875,7 @@ static int btrfs_clean_quota_tree(struct btrfs_trans_handle *trans,
 	return ret;
 }
 
-int btrfs_quota_enable(struct btrfs_trans_handle *trans,
-		       struct btrfs_fs_info *fs_info)
+int btrfs_quota_enable(struct btrfs_fs_info *fs_info)
 {
 	struct btrfs_root *quota_root;
 	struct btrfs_root *tree_root = fs_info->tree_root;
@@ -886,6 +885,7 @@ int btrfs_quota_enable(struct btrfs_trans_handle *trans,
 	struct btrfs_key key;
 	struct btrfs_key found_key;
 	struct btrfs_qgroup *qgroup = NULL;
+	struct btrfs_trans_handle *trans = NULL;
 	int ret = 0;
 	int slot;
 
@@ -893,9 +893,25 @@ int btrfs_quota_enable(struct btrfs_trans_handle *trans,
 	if (fs_info->quota_root)
 		goto out;
 
+	/*
+	 * 1 for quota root item
+	 * 1 for BTRFS_QGROUP_STATUS item
+	 *
+	 * Yet we also need 2*n items for a QGROUP_INFO/QGROUP_LIMIT items
+	 * per subvolume. However those are not currently reserved since it
+	 * would be a lot of overkill.
+	 */
+	trans = btrfs_start_transaction(tree_root, 2);
+	if (IS_ERR(trans)) {
+		ret = PTR_ERR(trans);
+		trans = NULL;
+		goto out;
+	}
+
 	fs_info->qgroup_ulist = ulist_alloc(GFP_KERNEL);
 	if (!fs_info->qgroup_ulist) {
 		ret = -ENOMEM;
+		btrfs_abort_transaction(trans, ret);
 		goto out;
 	}
 
@@ -906,12 +922,14 @@ int btrfs_quota_enable(struct btrfs_trans_handle *trans,
 				       BTRFS_QUOTA_TREE_OBJECTID);
 	if (IS_ERR(quota_root)) {
 		ret =  PTR_ERR(quota_root);
+		btrfs_abort_transaction(trans, ret);
 		goto out;
 	}
 
 	path = btrfs_alloc_path();
 	if (!path) {
 		ret = -ENOMEM;
+		btrfs_abort_transaction(trans, ret);
 		goto out_free_root;
 	}
 
@@ -921,8 +939,10 @@ int btrfs_quota_enable(struct btrfs_trans_handle *trans,
 
 	ret = btrfs_insert_empty_item(trans, quota_root, path, &key,
 				      sizeof(*ptr));
-	if (ret)
+	if (ret) {
+		btrfs_abort_transaction(trans, ret);
 		goto out_free_path;
+	}
 
 	leaf = path->nodes[0];
 	ptr = btrfs_item_ptr(leaf, path->slots[0],
@@ -944,9 +964,10 @@ int btrfs_quota_enable(struct btrfs_trans_handle *trans,
 	ret = btrfs_search_slot_for_read(tree_root, &key, path, 1, 0);
 	if (ret > 0)
 		goto out_add_root;
-	if (ret < 0)
+	if (ret < 0) {
+		btrfs_abort_transaction(trans, ret);
 		goto out_free_path;
-
+	}
 
 	while (1) {
 		slot = path->slots[0];
@@ -956,18 +977,23 @@ int btrfs_quota_enable(struct btrfs_trans_handle *trans,
 		if (found_key.type == BTRFS_ROOT_REF_KEY) {
 			ret = add_qgroup_item(trans, quota_root,
 					      found_key.offset);
-			if (ret)
+			if (ret) {
+				btrfs_abort_transaction(trans, ret);
 				goto out_free_path;
+			}
 
 			qgroup = add_qgroup_rb(fs_info, found_key.offset);
 			if (IS_ERR(qgroup)) {
 				ret = PTR_ERR(qgroup);
+				btrfs_abort_transaction(trans, ret);
 				goto out_free_path;
 			}
 		}
 		ret = btrfs_next_item(tree_root, path);
-		if (ret < 0)
+		if (ret < 0) {
+			btrfs_abort_transaction(trans, ret);
 			goto out_free_path;
+		}
 		if (ret)
 			break;
 	}
@@ -975,18 +1001,28 @@ int btrfs_quota_enable(struct btrfs_trans_handle *trans,
 out_add_root:
 	btrfs_release_path(path);
 	ret = add_qgroup_item(trans, quota_root, BTRFS_FS_TREE_OBJECTID);
-	if (ret)
+	if (ret) {
+		btrfs_abort_transaction(trans, ret);
 		goto out_free_path;
+	}
 
 	qgroup = add_qgroup_rb(fs_info, BTRFS_FS_TREE_OBJECTID);
 	if (IS_ERR(qgroup)) {
 		ret = PTR_ERR(qgroup);
+		btrfs_abort_transaction(trans, ret);
 		goto out_free_path;
 	}
 	spin_lock(&fs_info->qgroup_lock);
 	fs_info->quota_root = quota_root;
 	set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags);
 	spin_unlock(&fs_info->qgroup_lock);
+
+	ret = btrfs_commit_transaction(trans);
+	if (ret) {
+		trans = NULL;
+		goto out_free_path;
+	}
+
 	ret = qgroup_rescan_init(fs_info, 0, 1);
 	if (!ret) {
 	        qgroup_rescan_zero_tracking(fs_info);
@@ -1006,20 +1042,35 @@ int btrfs_quota_enable(struct btrfs_trans_handle *trans,
 	if (ret) {
 		ulist_free(fs_info->qgroup_ulist);
 		fs_info->qgroup_ulist = NULL;
+		if (trans)
+			btrfs_end_transaction(trans);
 	}
 	mutex_unlock(&fs_info->qgroup_ioctl_lock);
 	return ret;
 }
 
-int btrfs_quota_disable(struct btrfs_trans_handle *trans,
-			struct btrfs_fs_info *fs_info)
+int btrfs_quota_disable(struct btrfs_fs_info *fs_info)
 {
 	struct btrfs_root *quota_root;
+	struct btrfs_trans_handle *trans = NULL;
 	int ret = 0;
 
 	mutex_lock(&fs_info->qgroup_ioctl_lock);
 	if (!fs_info->quota_root)
 		goto out;
+
+	/*
+	 * 1 For the root item
+	 *
+	 * We should also reserve enough items for the quota tree deletion in
+	 * btrfs_clean_quota_tree but this is not done.
+	 */
+	trans = btrfs_start_transaction(fs_info->tree_root, 1);
+	if (IS_ERR(trans)) {
+		ret = PTR_ERR(trans);
+		goto out;
+	}
+
 	clear_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags);
 	btrfs_qgroup_wait_for_completion(fs_info, false);
 	spin_lock(&fs_info->qgroup_lock);
@@ -1031,12 +1082,16 @@ int btrfs_quota_disable(struct btrfs_trans_handle *trans,
 	btrfs_free_qgroup_config(fs_info);
 
 	ret = btrfs_clean_quota_tree(trans, quota_root);
-	if (ret)
-		goto out;
+	if (ret) {
+		btrfs_abort_transaction(trans, ret);
+		goto end_trans;
+	}
 
 	ret = btrfs_del_root(trans, fs_info, &quota_root->root_key);
-	if (ret)
-		goto out;
+	if (ret) {
+		btrfs_abort_transaction(trans, ret);
+		goto end_trans;
+	}
 
 	list_del(&quota_root->dirty_list);
 
@@ -1048,6 +1103,9 @@ int btrfs_quota_disable(struct btrfs_trans_handle *trans,
 	free_extent_buffer(quota_root->node);
 	free_extent_buffer(quota_root->commit_root);
 	kfree(quota_root);
+
+end_trans:
+	ret = btrfs_end_transaction(trans);
 out:
 	mutex_unlock(&fs_info->qgroup_ioctl_lock);
 	return ret;

fs/btrfs/qgroup.h

@@ -141,10 +141,8 @@ struct btrfs_qgroup {
 #define QGROUP_RELEASE		(1<<1)
 #define QGROUP_FREE		(1<<2)
 
-int btrfs_quota_enable(struct btrfs_trans_handle *trans,
-		       struct btrfs_fs_info *fs_info);
-int btrfs_quota_disable(struct btrfs_trans_handle *trans,
-			struct btrfs_fs_info *fs_info);
+int btrfs_quota_enable(struct btrfs_fs_info *fs_info);
+int btrfs_quota_disable(struct btrfs_fs_info *fs_info);
 int btrfs_qgroup_rescan(struct btrfs_fs_info *fs_info);
 void btrfs_qgroup_rescan_resume(struct btrfs_fs_info *fs_info);
 int btrfs_qgroup_wait_for_completion(struct btrfs_fs_info *fs_info,