Soil phosphorus (P) improvement is one of the important aims of vegetation restoration. However, the effects of different vegetation restoration types on soil P cycling and the underlying mechanisms remain unclear. Together with vegetation structure and soil properties, we measured the soil P fractions and P transformation rates to evaluate the characteristics of ecologically relevant soil P fraction distributions and their dynamics for the most common vegetation restoration types in a subtropical mountain ecosystem (i.e., PF, Pinus forest; EF, Eucalyptus forest; SL, shrubland; and NSF, natural secondary forest). We found that water-extractable inorganic P (Pi), organic P (Po), and acid phosphatase activity (APA) were significantly higher in NSF than those in the other vegetation types (P < 0.01), together with the highest gross P mineralization rate (Pmin) and net P immobilization rate (Pimm), suggesting biological processes played a more important role in soil P cycling in the NSF. In contrast, the soil showed the lowest values of Pmin, Pimm, and APA in the EF compared to NSF and PF (P < 0.05), together with the highest P loss and the higher net P solubilization rate (Psol), indicating the greater importance of soil geochemical processes. Compared with the NSF and EF, geochemical and biological processes co-regulated soil P cycling in the PF and SL. However, the soil in the PF displayed higher P fluxes (Pmin, Pimm, Psol, and P loss) than those in the SL (P < 0.05), suggesting the soil had higher P flux magnitudes in the PF. Results of correlation analyses showed that the soil microbial community structure and activity played a more important role than plant community attributes and soil physicochemical properties in soil P fraction distribution and fluxes. In conclusion, soil P fraction distributions and their fluxes can be significantly influenced by vegetation restoration types. Land management strategies focusing on restoration of the soil microbial community may enhance soil P cycling and improve soil quality.