[PATCH] Btrfs: fix race between balance recovery and root deletion

[PATCH] Btrfs: fix race between balance recovery and root deletion

[Wang Shilong]

Balance recovery is called when RW mounting or remounting from
RO to RW, it is called to finish roots merging.

When doing balance recovery, relocation root's corresponding
fs root(whose root refs is 0) might be destroyed by cleaner
thread, this will make btrfs fail to mount.

Fix this problem by holding @cleaner_mutex when doing balance
recovery.

Signed-off-by: Wang Shilong <wangsl.fnst@cn.fujitsu.com>
---
 fs/btrfs/disk-io.c | 2 ++
 fs/btrfs/super.c   | 2 ++
 2 files changed, 4 insertions(+)

diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 8bb4aa1..61d8e87 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -2904,7 +2904,9 @@ retry_root_backup:
 		if (ret)
 			goto fail_qgroup;
 
+		mutex_lock(&fs_info->cleaner_mutex);
 		ret = btrfs_recover_relocation(tree_root);
+		mutex_unlock(&fs_info->cleaner_mutex);
 		if (ret < 0) {
 			printk(KERN_WARNING
 			       "BTRFS: failed to recover relocation\n");
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 4662d92..b6ebde2 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -1467,7 +1467,9 @@ static int btrfs_remount(struct super_block *sb, int *flags, char *data)
 			goto restore;
 
 		/* recover relocation */
+		mutex_lock(&fs_info->cleaner_mutex);
 		ret = btrfs_recover_relocation(root);
+		mutex_unlock(&fs_info->cleaner_mutex);
 		if (ret)
 			goto restore;
 
-- 
1.8.2.1

Re: [PATCH] Btrfs: fix race between balance recovery and root deletion

[David Sterba]

On Thu, Jun 26, 2014 at 11:08:16AM +0800, Wang Shilong wrote:
> Balance recovery is called when RW mounting or remounting from
> RO to RW, it is called to finish roots merging.
> 
> When doing balance recovery, relocation root's corresponding
> fs root(whose root refs is 0) might be destroyed by cleaner
> thread, this will make btrfs fail to mount.
> 
> Fix this problem by holding @cleaner_mutex when doing balance
> recovery.
> 
> Signed-off-by: Wang Shilong <wangsl.fnst@cn.fujitsu.com>

I don't pretend to understand all the dark corners of relocation, but
this looks as a correct fix to me. Relocation and cleaner are mutually
exclusive in the normal case.

Reviewed-by: David Sterba <dsterba@suse.cz>