]> Pileus Git - ~andy/linux/commitdiff
Btrfs: remove some unnecessary spin_lock usages
authorWang Shilong <wangsl-fnst@cn.fujitsu.com>
Sun, 7 Apr 2013 10:50:17 +0000 (10:50 +0000)
committerJosef Bacik <jbacik@fusionio.com>
Mon, 6 May 2013 19:54:39 +0000 (15:54 -0400)
We use mutex lock to protect all the user change operations.
So when we are calling find_qgroup_rb() to check whether qgroup
exists, we don't have to hold spin_lock.

Besides, when enabling/disabling quota, it must be single thread
when operations come here. spin lock must be firstly used to
clear quota_root when disabling quota, while enabling quota, spin
lock must be used to complete the last assign work.

Signed-off-by: Wang Shilong <wangsl-fnst@cn.fujitsu.com>
Reviewed-by: Miao Xie <miaox@cn.fujitsu.com>
Signed-off-by: Josef Bacik <jbacik@fusionio.com>
fs/btrfs/qgroup.c

index 0a1f6861ae9adb974c44e46a801f5b4686e9a82f..49c4e6398f18210cb6dbd28da6e8c35ed7864cd7 100644 (file)
@@ -98,7 +98,7 @@ struct btrfs_qgroup_list {
        struct btrfs_qgroup *member;
 };
 
-/* must be called with qgroup_lock held */
+/* must be called with qgroup_ioctl_lock held */
 static struct btrfs_qgroup *find_qgroup_rb(struct btrfs_fs_info *fs_info,
                                           u64 qgroupid)
 {
@@ -792,13 +792,10 @@ int btrfs_quota_enable(struct btrfs_trans_handle *trans,
        int slot;
 
        mutex_lock(&fs_info->qgroup_ioctl_lock);
-       spin_lock(&fs_info->qgroup_lock);
        if (fs_info->quota_root) {
                fs_info->pending_quota_state = 1;
-               spin_unlock(&fs_info->qgroup_lock);
                goto out;
        }
-       spin_unlock(&fs_info->qgroup_lock);
 
        /*
         * initially create the quota tree
@@ -860,14 +857,11 @@ int btrfs_quota_enable(struct btrfs_trans_handle *trans,
                        if (ret)
                                goto out_free_path;
 
-                       spin_lock(&fs_info->qgroup_lock);
                        qgroup = add_qgroup_rb(fs_info, found_key.offset);
                        if (IS_ERR(qgroup)) {
-                               spin_unlock(&fs_info->qgroup_lock);
                                ret = PTR_ERR(qgroup);
                                goto out_free_path;
                        }
-                       spin_unlock(&fs_info->qgroup_lock);
                }
                ret = btrfs_next_item(tree_root, path);
                if (ret < 0)
@@ -882,13 +876,12 @@ out_add_root:
        if (ret)
                goto out_free_path;
 
-       spin_lock(&fs_info->qgroup_lock);
        qgroup = add_qgroup_rb(fs_info, BTRFS_FS_TREE_OBJECTID);
        if (IS_ERR(qgroup)) {
-               spin_unlock(&fs_info->qgroup_lock);
                ret = PTR_ERR(qgroup);
                goto out_free_path;
        }
+       spin_lock(&fs_info->qgroup_lock);
        fs_info->quota_root = quota_root;
        fs_info->pending_quota_state = 1;
        spin_unlock(&fs_info->qgroup_lock);
@@ -913,11 +906,9 @@ int btrfs_quota_disable(struct btrfs_trans_handle *trans,
        int ret = 0;
 
        mutex_lock(&fs_info->qgroup_ioctl_lock);
-       spin_lock(&fs_info->qgroup_lock);
-       if (!fs_info->quota_root) {
-               spin_unlock(&fs_info->qgroup_lock);
+       if (!fs_info->quota_root)
                goto out;
-       }
+       spin_lock(&fs_info->qgroup_lock);
        fs_info->quota_enabled = 0;
        fs_info->pending_quota_state = 0;
        quota_root = fs_info->quota_root;
@@ -1060,16 +1051,13 @@ int btrfs_remove_qgroup(struct btrfs_trans_handle *trans,
        }
 
        /* check if there are no relations to this qgroup */
-       spin_lock(&fs_info->qgroup_lock);
        qgroup = find_qgroup_rb(fs_info, qgroupid);
        if (qgroup) {
                if (!list_empty(&qgroup->groups) || !list_empty(&qgroup->members)) {
-                       spin_unlock(&fs_info->qgroup_lock);
                        ret = -EBUSY;
                        goto out;
                }
        }
-       spin_unlock(&fs_info->qgroup_lock);
 
        ret = del_qgroup_item(trans, quota_root, qgroupid);
 
@@ -1106,20 +1094,17 @@ int btrfs_limit_qgroup(struct btrfs_trans_handle *trans,
                       (unsigned long long)qgroupid);
        }
 
-       spin_lock(&fs_info->qgroup_lock);
-
        qgroup = find_qgroup_rb(fs_info, qgroupid);
        if (!qgroup) {
                ret = -ENOENT;
-               goto unlock;
+               goto out;
        }
+       spin_lock(&fs_info->qgroup_lock);
        qgroup->lim_flags = limit->flags;
        qgroup->max_rfer = limit->max_rfer;
        qgroup->max_excl = limit->max_excl;
        qgroup->rsv_rfer = limit->rsv_rfer;
        qgroup->rsv_excl = limit->rsv_excl;
-
-unlock:
        spin_unlock(&fs_info->qgroup_lock);
 out:
        mutex_unlock(&fs_info->qgroup_ioctl_lock);