Rice (Oryza sativa L.) straw is commonly incorporated into paddy soil after harvest, although it is not fully understood how the straw return strategy and its co‐incorporation with legumes affect soil productivity and rice performance. A 5‐yr experiment was conducted in a rice–rice–fallow or rice–rice–Chinese milk vetch (Mv, Astragalus sinicus L.) rotation in South China. Six treatments were included (i) a control with no fertilization and residue return, (ii) 100% of fertilizer recommendation (F100), (iii) 80% of F100 plus conventional return of rice straw (F80Rs), (iv) F80 plus Mv planting and incorporation (F80Mv), (v) co‐incorporation of F80Rs and F80Mv (F80RsMv), (vi) similar to F80RsMv except late rice straw was retained as high stubble (∼35 cm, F80RhMv). The results showed that, under reduced inorganic fertilizer input, grain yields in treatments with straw and/or Mv return were similar with those of F100. However, N uptake in early and late rice markedly increased by 8 to 14% and 30 to 53% in the co‐incorporation treatment F80RsMv compared with F100, respectively, as well as in the F80RhMv treatment for late rice. Significantly greater fertilizer use efficiency was obtained in the co‐incorporation practices. Although negative cumulative effects were observed for soil organic matter, total N, and available N after 5‐yr rotation, the magnitude of decline in the co‐incorporation treatments was much smaller than in others. Soil available P and K were accumulated in co‐incorporation practice, whereas pH was less affected. Overall, the study suggested that the combination of rice straw and leguminous green manure is a promising practice to produce more grain yield with less fertilizer inputs in subtropical double‐rice production.