This paper proposes a fully decentralized peer-to-peer overlay structure GeoP2P, to facilitate geographic location based search and retrieval of information. Certain limitations of centralized geographic indexes favor peer-to-peer organization of the information, which, in addition to avoiding performance bottleneck, allows autonomy over local information. Peer-to-peer systems for geographic or multidimensional range queries built on existing DHTs suffer from the inaccuracy in linearization of the multidimensional space. Other overlay structures that are based on hierarchical partitioning of the search space are not scalable because they use special super-peers to represent the nodes in the hierarchy. GeoP2P partitions the search space hierarchically, maintains the overlay structure and performs the routing without the need of any super-peers. Although similar fully-decentralized overlays have been previously proposed, they lack the ability to dynamically grow and retract the partition hierarchy when the number of peers change. GeoP2P provides such adaptive features with minimum perturbation of the system state. Such adaptation makes both the routing delay and the state size of each peer logarithmic to the total number of peers, irrespective of the size of the multidimensional space. Our analysis also reveals that the overlay structure and the routing algorithm are generic and independent of several aspects of the partitioning hierarchy, such as the geometric shape of the zones or the dimensionality of the search space.