The service life of memory cards is to be substantially elongated against the occurrence of faulty blocks. A control logic searches blocks in a nonvolatile memory cell array for any acquired fault on the basis of a fault-inviting code in a management information section. If any faulty block is detected, the faulty block will be subjected to write/read comparison of data to judge whether or not the data in the block are normal. Any block determined to be normal will undergo rewriting of its fault-inviting code and registered as a normal block. Further, the registered block is stored into a write management table in the management area as a writable block. This enables an essentially normal block judged faulty on account of an erratic error or some other reason to be restored.