This paper presents the results of a survey conducted by the working group on research software metadata within the NFDI (Nationale Forschungsdateninfrastruktur / National Research Data Infrastructure) to gain a better understanding of the requirements and practices for research software metadata in the different NFDI consortia. The survey consisted of four main parts: usage of research software, FAIRness of research software, state of metadata on research software, and artifacts developed to support FAIRness. The survey was distributed to representatives of the 26 NFDI consortia, with responses from 21 of them. The survey showed that research software is primarily used for data processing, simulation, and software prototyping. Analysis scripts, libraries, stand-alone software, and web services are the most common software types. Some consortia reported that research software could be considered a research output, while others only viewed it as a supplement to traditional publications. Proper citation of software is crucial for ensuring FAIRness. Approximately half of the domains represented by the surveyed NFDI consortia already use DOIs for citing software. Some domains had not yet fully adopted practices for documenting research software, posing challenges for both developers and users. Researchers typically search for software using generic search engines, academic papers, and community-specific marketplaces where they exist. In respect to metadata, only a few domain-specific metadata schemas for research software were identified by the consortia, with generic metadata schemas such as Codemeta, Bioschemas, schema.org, and the Citation File Format being the most commonly mentioned. Relevant metadata elements for research software include algorithms, methods, software/hardware requirements, and tasks performed by the software. The survey results will guide activities in the NFDI working group on research software metadata and will also be a relevant foundation for the base service nfdi.software.