Btrfs: Fix the problem that the replace destroys the seed filesystem

The seed filesystem was destroyed by the device replace, the reproduce
method is:
 # mkfs.btrfs -f <dev0>
 # btrfstune -S 1 <dev0>
 # mount <dev0> <mnt>
 # btrfs device add <dev1> <mnt>
 # umount <mnt>
 # mount <dev1> <mnt>
 # btrfs replace start -f <dev0> <dev2> <mnt>
 # umount <mnt>
 # mount <dev0> <mnt>

It is because we erase the super block on the seed device. It is wrong,
we should not change anything on the seed device.

Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Chris Mason <clm@fb.com>

1 files changed, 6 insertions, 2 deletions

diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
index da0e632a21fc..00c8efdcd1e5 100644
--- a/fs/btrfs/volumes.c
+++ b/fs/btrfs/volumes.c
@@ -1848,8 +1848,12 @@ void btrfs_rm_dev_replace_srcdev(struct btrfs_fs_info *fs_info,
         if (srcdev->bdev) {
                 fs_info->fs_devices->open_devices--;
 
-                /* zero out the old super */
-                btrfs_scratch_superblock(srcdev);
+                /*
+                 * zero out the old super if it is not writable
+                 * (e.g. seed device)
+                 */
+                if (srcdev->writeable)
+                        btrfs_scratch_superblock(srcdev);
         }
 
         call_rcu(&srcdev->rcu, free_device);