Btrfs: improve performance on fsync against new inode after rename/unlink

[cascardo/linux.git] / fs / btrfs / inode.c

diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index f968654..453b9d0 100644 (file)
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -4203,6 +4203,7 @@ static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
        int err = 0;
        struct btrfs_root *root = BTRFS_I(dir)->root;
        struct btrfs_trans_handle *trans;
+       u64 last_unlink_trans;
 
        if (inode->i_size > BTRFS_EMPTY_DIR_SIZE)
                return -ENOTEMPTY;
@@ -4225,11 +4226,27 @@ static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
        if (err)
                goto out;
 
+       last_unlink_trans = BTRFS_I(inode)->last_unlink_trans;
+
        /* now the directory is empty */
        err = btrfs_unlink_inode(trans, root, dir, d_inode(dentry),
                                 dentry->d_name.name, dentry->d_name.len);
-       if (!err)
+       if (!err) {
                btrfs_i_size_write(inode, 0);
+               /*
+                * Propagate the last_unlink_trans value of the deleted dir to
+                * its parent directory. This is to prevent an unrecoverable
+                * log tree in the case we do something like this:
+                * 1) create dir foo
+                * 2) create snapshot under dir foo
+                * 3) delete the snapshot
+                * 4) rmdir foo
+                * 5) mkdir foo
+                * 6) fsync foo or some file inside foo
+                */
+               if (last_unlink_trans >= trans->transid)
+                       BTRFS_I(dir)->last_unlink_trans = last_unlink_trans;
+       }
 out:
        btrfs_end_transaction(trans, root);
        btrfs_btree_balance_dirty(root);