Thanks Eric, the patch looks correct to me, this was the only path in that area not freeing the tcur (or freeing cur and assigning tcur to cur) before jumping to newino/alloc_inode. Reviewed-by: Christoph Hellwig <hch at lst.de>