In this paper, we propose a peer-to-peer approach to derive and obtain additional context data from lowlevel context data that may be spread over multiple domains in pervasive computing environments. In this system, peers are self-organized into a semantic peerto-peer network as the underlying communication substrate. Context reasoning is done in a distributed fashion through logical reasoning according to a set of user-defined rules. Both pull and push services are supported in the system to enable message exchange during the reasoning process. We present our design concepts, and prove the effectiveness of our system through the prototype evaluation.