Modern vehicular networks support various services with variable QoS (Quality of Service) constraints. Two major classes of vehicular applications are identified, namely, safety and non safety services. The former are delay-sensitive while the latter depends mainly on throughput. However, many regions are suffering from a shortage of network resources while an increasing number of vehicle users need to be satisfied. Thus, the design of efficient allocation schemes of available resources is necessary. In this regard, one of the promising technologies is network slicing, a next-generation 5G perspective, that creates multiple logical networks on a common physical infrastructure. This paradigm enables efficient exploitation of shared physical infrastructure resources to meet the diverse needs of different use cases.In this paper we design a framework for a road-state-based and adaptive network slicing scheme for vehicular networks.The goal is to temporarily prioritize emergency traffic in incident situations while maintaining acceptable QoS for non-safety related sevices in a resource constrained environment. Our proposal adds to native slicing the ability to take into account road conditions besides of customer’s specifications in terms of QoS. Moreover, our adaptive scheme makes it possible to judiciously exploit the available resources even if they are limited according to the rigor of the application. SDN (Software Defined Networking), NFV (Network Function Virtualization) and Fog computing paradigms are the key enablers of our proposed solution. We implemented the proposed architecture based on the Mininet-Wifi emulator to create a vehicular network, the ONOS SDN controller, and the network slicing tool OpenVirteX. Experimental results prove that our suggested adaptive resource allocation scheme enhances the performance of the emergency services in terms of end-to-end delay while keeping acceptable throughput for non-safety traffic in stressed situations.