This paper discusses aspects of the redocumentation of legacy systems and proposes a model oriented approach to generating documentation, which is to produce models from existing systems and to generate the documentation based on the models. Since the software models can bridge the gap of a legacy system and an evolved system, the generated documentation covers all the information of system evolution. A prototype software redocumentation tool is presented to semi-automate this process and a case study of a system in IBM assembler is used for experiments with the approach and the prototype tool.