Image super-resolution (SR) aims to reconstruct high-quality high-resolution images from low-resolution inputs and has been widely applied in visual tasks such as medical imaging and remote sensing. Despite recent advances, existing methods still suffer from limited receptive field, insufficient spatial interaction, and a trade-off between resolution accuracy and lightweight design. To solve these issues, we propose a lightweight super-resolution framework based on dynamic recursive state-space modeling called DyRSRNet. Specifically, we design a lightweight dynamic state-space block (DyRWSSB) integrating the receptance weighted key value model for efficient long-sequence dependency modeling, thereby enhancing global structure and local texture restoration. In addition, we propose a visual recurrent shift encoding (VRSE) module to capture cross-directional spatial dependencies and channel interactions, enhancing the modeling of texture details. Furthermore, we propose a recursive context-aware state-space (RCSS) module, which enhances long-range context representation and structural restoration through bidirectional state-space modeling and direction-aware dynamic modulation mechanisms. Experimental results demonstrate that DyRSRNet achieves superior resolution performance on multiple SR benchmarks datasets, while significantly reducing parameter count and computational overhead. Our code is available on https://github.com/vpsg-research/DyRSRNet